The BEST bandwidth? [Archive]

FHDave

01-07-2003, 10:23 PM

A. NAC has a regional eastern network, Internap does not

Hm ... what does "Regional eastern network" means? Internap (in US) consists (buy transit) of nine major tier-one nationwide bandwidth providers, UUNet, Sprint, Verio, ATT, Cable and Wireless, Global Crossing, BBN/Genuity, QWest, and Digex. Internap has datacenters and/or POP from West Coast to East Coast, from Japan to Europe.

B. NAC has direct uncongested connections to London (LINX). Internap does not.

Does NAC own the fiber connecting NYC and London? No! NAC only has a router/POP in LINX and nothing more. InterNAP also has datacenter in London, not only a POP.

C. NAC has always been ultra responsive to us. We have not had to leave a voice mail in years. I am not sure about Internap and this.

Speaking about ultra-responsive, nothing beats Internap's support. Believe me in this!

D. NAC is able to offer many services you want. For example if your their customer, you can say "Dont send my traffic out to Cogent" or announce me only to your peers not your transit

There will be no cogent on Internap, so why bother compares this point. Besides, Internap routing is much more than just simple BGP4. Its intelligient routing goes beyond BGP4. I guess you are clueless about this too :)

E. NAC is a regional (now with London you could say international) backbone. Internap is not a backbone, and does not want to be.

NAC is a backbone? hehe ... since when?

If I personally would be looking for new bandwidth, and could afford it, NAC would be my choice simply because I would get more personalized service from them

And more downtime ... sure :)

I don't say NAC is bad, I just will argue Internap is operating at different level than NAC and there is no comparison between Internap and NAC. Look on it this way, tell me one of NAC's client that's comparable in size and reputation to NasDaq, American Airlines, etc :)

NAC is OK (will be good if there are no some DDOS attacks, some network outages, etc). Internap, like it or not, is the best ... well at least better than NAC, IMHO :)

FHDave

01-07-2003, 10:25 PM

Originally posted by N9ne
How much do interNAP and NAC cost?

Not sure how much NAC costs. But Internap can be very aggresive too, and with 100% network uptime guaranteed on SLA (does NAC offer this, btw?), even if Internap costs $100/mbps more expensive, I will still go with Internap.

wlandman

01-07-2003, 11:32 PM

Yes NAC is a backbone. I say this because they have numerous core routers connected to each other with an OC3/OC12. They have public and private peering at the NAP.

Why would you say that NAC is not a backbone?

nacjamesc

01-08-2003, 12:21 AM

Originally posted by wlandman
Yes NAC is a backbone. I say this because they have numerous core routers connected to each other with an OC3/OC12. They have public and private peering at the NAP.

Why would you say that NAC is not a backbone?

NAC is definetly a backbone. It may not be as big as Worldcom, MFN or Sprint, but we do operate an international backbone.

InterNAPs model is, in my opinion, very broken. They charge (or atleast when we were a customer of them a few years ago) insane pricing and are hardly flexible on it. InterNAPs quality and uptime was definetly good. They have little pods in major cities and buy transit from most of the big players. Instead of peering with people to try to save money and offer more competitive pricing, they'd inflate pricing due to their high transit costs (because they have like 3498349832 providers in each pod.) You do definetly get what you pay for though.

Does NAC own the fiber connecting NYC and London? No! NAC only has a router/POP in LINX and nothing more. InterNAP also has datacenter in London, not only a POP.

NAC has a redundant OC12 network from Ashburn Virginia to New York to London, UK and several other DC's in between. We use Juniper M series routers for the IP core.

Having a datacenter in London means nothing; why not let people who already have well established datacenters (IE: Telehouse Docklands) maintain that, and we just provide connectivity. After all, the customers Its obvious NAC would be wasting money if we had 100k SQ Feet data centers in each of our major cities. Undoubtedly we'd be chapter 11 like the rest of the players in our industry. NAC is growing where others are cutting back.

You'd be surprised to find out that most people lease fiber from MFN or Tycho and dont have fiber themselves.

I'll not list any of our larger customers on this forum but if you really take a step back and think about it; would there be this much buzz about NAC if we werent doing something right?

You make it sound that since internap has 923012039 transit providers it makes them better. Infact you are wrong. Not only do they not have direct connectivity to anyone besides the people they purchase from, but they are throwing money out the window. Then again, I guess they would be out of business by now if they werent doing something right aswell.

I have no problem with internap, I just wish youd think more intelligently about your statements about why InterNAP is > *. Every ISP has flaws.

If you ever have a problem with NACs service as a customer, feel free to ask for me by name if you would like to complain or get something resolved.

Thanks

FHDave

01-08-2003, 12:50 AM

Originally posted by nacjamesc
InterNAPs model is, in my opinion, very broken. They charge (or atleast when we were a customer of them a few years ago) insane pricing and are hardly flexible on it.

Just because their pricing is higher does not mean their model is broken. But anyway, I bet my Internap pricing will match NAC's pricing. But even if it's $100/mbps more, I wills till go with Internap.

You make it sound that since internap has 923012039 transit providers it makes them better.

Incorrect. Why is Internap better in my opinion? In addition to their Intelligent Routing Technology (that goes beyond simple BGP4), they also offer 100% network uptime guarantee and they have been able to deliver it to us. Not even a single second of network downtime.
All serious datacenter (Inflow, Internap, Cables and Wireless, UUNet, Level3, etc) will offer 100% network uptime guarantee under SLA and deliver that. Does NAC offer any uptime guarantee (i.e. 100%). If so, does NAC able to deliver it?

Don't get me wrong. When we decided to move into a new datacenter last March/April, NAC was also considered as well as 10 other datacenters. But after doing carefull comparison and constant network monitoring for around 4-6 weeks, we finally decided to go with Internap. And there was for a reson. We did not make any decision out of the blue. But a decision based on a careful comparison/testing for 4-6 weeks!

I don't say NAC is bad. All I said is Internap is better than NAC, and this is from my experience when I had a servers inside NAC and from monitoring NAC network from time to time as well as hearing some downtime/DOS attacks reported by people. And I have not had a single complaint nor I have ever heard any complaint about Internap.

No "NAC" bashing is intedended. Just my objective view. Take it as what it's worth.

Deb

01-08-2003, 02:01 AM

NAC was also considered as well as 10 other datacenters. But after doing carefull comparison and constant network monitoring... We've been monitoring for a few years now. It's almost a game at this point.

Each time we visit a site that is sluggish or down we'll quickly start running tests on the address to pinpoint which network is causing the problem. Sprint, Cogent, and C&W were quickly scratched off of our list of potentials. NAC was good but we found ups and downs with it. Qwest and UUnet both tend to have bottlenecks in certain areas. etc....

Internap though... wow. Always a way to get from A to B and Back Again with Internap. Once the presales calls began, Internap AGAIN went above and beyond for us. We worked with them for a few months, as well as our attorney and theirs, to ensure everything we wanted to accomplish would be accomplished. We were the presale clients from hell when it came to choosing Internap and they hung in there with us through thick and thin.

Now that we've been using them for a while, I continue to 'sing their praises'. EXCELLENT response times, superb support, fabulous communication tools (their notices are almost too chatty they are so quick to let us know every tiny lil thing that is going on), and our clients aren't complaining ;)

It's always a fear that once you sign the dotted line you'll wish you had not done so. This has not been the case with Internap. We are 100% satisfied with their services.

This isn't to say the others are not doing good as well. I can only note that my experience with Internap from the beginning through now has been positive and well worth it.

wlandman

01-08-2003, 02:15 AM

I can't remember having unscheduled downtime with NAC. They usually warn a few days in advance when they are doing something.

Plus they dont hide anything like other providers. We are going to follow NAC's model of bieng open with the customer because obviously it works.

FHDave

01-08-2003, 02:29 AM

Originally posted by wlandman
I can't remember having unscheduled downtime with NAC. They usually warn a few days in advance when they are doing something.

With Internap, I have not seen any scheduled downtime. Internap has always been careful with any scheduled work to ensure that the network (with all of its redundancies) will not go down even during the work.

and ... yes, of course. There is no such thing as unscheduled downtime.

FHDave

01-08-2003, 02:33 AM

Originally posted by Deb
Now that we've been using them for a while, I continue to 'sing their praises'. EXCELLENT response times, superb support, fabulous communication tools (their notices are almost too chatty they are so quick to let us know every tiny lil thing that is going on), and our clients aren't complaining ;)

Yes, their support is top notch. I still remember one day we got a call from Internap's NOC because we pulled out one of the two 100 Mbps ethernet drops to test our switch. Got a surprise call within several minutes asking whether everything was allright with us!

Great to know that FQ is also an Internap host! Way to go Deb!

wlandman

01-08-2003, 04:01 AM

If Internap decided that they need to do som maintenance on the exact border box you are connected to, and their was no way they could hot-swap it, wouldn't they need scheduled maintenance?

So you were basically lucky and Internap did not have to do any work yet on your specific router that you are connected to. Which means either that they dont need to upgrade it, or dont want to in order to save more money (as jamesc said, they pay a lot of money for bandwidth).

Any how, if your on the east coast and dont want to go with us (we dont do dedicated / colocated servers for the public yet), or with NAC :-)

JTY

01-08-2003, 10:09 AM

InterNap provides colo/ethernet customers with Dual Ethernet feeds from different routers/switches so that the customer can stay up.

porcupine

01-08-2003, 10:33 AM

Originally posted by arnaudpignar
MFN (above.net) is a great choose if you haven't yet any carrier.

Very good and redondant backbone and connected to all major peering exchange in the world with a open peering policy.

But they aren't the cheapest but a good choose if i where you.

I had a picture of MFN's (though i think a different one, metromedia fiber net i believe it was) eviction notice from 151 Front, I dont think it was above.net though :).

FHDave

01-08-2003, 10:43 AM

Originally posted by wlandman
If Internap decided that they need to do som maintenance on the exact border box you are connected to, and their was no way they could hot-swap it, wouldn't they need scheduled maintenance?

As John has mentioned, my core switch is connected to both of Internap's border routers. Each of Internap's border routers has its own backup router. I believe the same thing will go with Inflow, UUNet, etc. Your description above sounds like something you will have from a much smaller data center ;)

So you were basically lucky and Internap did not have to do any work yet on your specific router that you are connected to.

Incorrect. From time to time we do receive notifications that Internap is working on one of its border routers. But again, their redundancy does work and we did not get any downtime. It's not about lucky, it's about good network redundancy design.

Any how, if your on the east coast and dont want to go with us (we dont do dedicated / colocated servers for the public yet), or with NAC :-)

I am in the east coast. But I am very content with what I have right now :)

nacjamesc

01-08-2003, 11:14 AM

Originally posted by FHDave

Incorrect. From time to time we do receive notifications that Internap is working on one of its border routers. But again, their redundancy does work and we did not get any downtime. It's not about lucky, it's about good network redundancy design.

You make it sound like they are the only ones to do this type of thing or they invented it or something. It is definetly not InterNAP proprietary.

tazzy

01-08-2003, 11:24 AM

Originally posted by wlandman
NAC compared to Internap? I can compare the two pretty well.
A. NAC has a regional eastern network, Internap does not

InterNAP connects directly to 9 Transit carriers, NAC peers.

B. NAC has direct uncongested connections to London (LINX). Internap does not.

Actually, InterNAP connects to LINX.

C. NAC has always been ultra responsive to us. We have not had to leave a voice mail in years. I am not sure about Internap and this.

Can't answer this.

D. NAC is able to offer many services you want. For example if your their customer, you can say "Dont send my traffic out to Cogent" or announce me only to your peers not your transit customers, etc, etc. I am pretty sure Internap does not offer this.

This is very possible, except they don't use Cogent. Only 9 GOOD carriers.

E. NAC is a regional (now with London you could say international) backbone. Internap is not a backbone, and does not want to be.

NAC peers, remember that. InterNAP buys transit for quality tier-1 carriers.

If I personally would be looking for new bandwidth, and could afford it, NAC would be my choice simply because I would get more personalized service from them.

Go with INTERNAP over NAC anyday of the week.

porcupine

01-08-2003, 11:26 AM

NAC peers and internap buys transit.....

That'd make me want NAC assuming it's free peering :D, that means a cheaper price for the end user, why buy something when you can get it for free? :)

tazzy

01-08-2003, 11:35 AM

Originally posted by porcupine
NAC peers and internap buys transit.....

That'd make me want NAC assuming it's free peering :D, that means a cheaper price for the end user, why buy something when you can get it for free? :)

Hello,

Cogent followed that motto as well ;)

Now look what's happend to them :cartman:

porcupine

01-08-2003, 11:38 AM

Originally posted by tazzy

Hello,

Cogent followed that motto as well ;)

Now look what's happend to them :cartman:

Followed what model? What have you been smoking?? :eek:. Every provider i know of prefers to peer as much transit as possible as it's *FREE*, cogent only peers because they cant afford to buy transit, the true definition of tier1 defines providers who peer and don't need to buy transit. You dont think the "big boys" in the carrier market actually *buy* transit do you? :eek:

nacjamesc

01-08-2003, 11:40 AM

Originally posted by tazzy

InterNAP connects directly to 9 Transit carriers, NAC peers.

NAC has 3 primary transit carriers. Williams, AT&T, and Cogent. We also have backup transit through MFN.

Actually, InterNAP connects to LINX.

You are absolutely wrong on this. The InterNAP model is to NOT do ANY peering, but to purchase transit. They do this so a) they are always a customer of someone so they get the fastest response time and b) they avoid the overcongested exchange points.

Cant argue with point A, but B is a thing of the past. It happens, but its very rare to see these days, so its effectively moot.

If you look at :
http://www.linx.net/members/inap/index.thtml

You'll see that its a company called I-Nap, not InterNAP. They are in no way affilliated with each other.

Without maintaining a network, peering would be useless because only their London based customers would have access to the peers.

NAC peers, remember that. InterNAP buys transit for quality tier-1 carriers.

Yes. NAC has better connectivity to alot companies.

Peering is always better than transit, in my opinion. You get direct connectivity. You cant beat that.

For example, from InterNAP, do a traceroute to www.aol.com or www.cv.net (Cablevision). Since InterNAP doesnt peer with anyone, it has to traverse many networks before getting to its destination.

From the NAC Network, it takes 8ms to get to www.aol.com, and 4ms for www.cv.net. What does this mean? NAC gets your packets to the EYEBALLS faster. Ofcourse the difference is relatively immeasurable, but every little bit counts.

Go with INTERNAP over NAC anyday of the week.

I'm glad you have love for InterNAP; I do not want to turn this into a InterNAP vs NAC war. What i'm saying is that alot of people dont even know what makes a provider better than another one besides what they see as far as uptime. That definetly matters but other things are more important.

IE: (This is hypothetical, no provider names involved) Would you rather have a provider that is up ALL the time but has terrible connectvity or would you rather one that has its fair share of growing pains but still maintains 99.999% uptime, but provides the best service possible?

InterNAP is quite good, but dont rule NAC out, especially if you dont know what makes them good.

FHDave

01-08-2003, 11:52 AM

Originally posted by nacjamesc
IE: (This is hypothetical, no provider names involved) Would you rather have a provider that is up ALL the time but has terrible connectvity or would you rather one that has its fair share of growing pains but still maintains 99.999% uptime, but provides the best service possible?

Internap is up ALL the time and it has great connectivity anywhere. So I am not sure what you are comparing with.

Does NAC really have 99.999% uptime (only around 5 minutes downtime annually), considering all the network outages and DOS attacks NAC has experienced from time to time? Does NAC even provide uptime guarantee to its colo customers? Does NAC even provide coast to coast, intercontinental latency gurantee?

I am sorry, my customers won't even notice a 4-8 ms because of this peering vs. buying transit difference. But when the networks is down, they all will scream. So yes, uptime guarantees is much better for us and it's the primary thing to be considered when it comes to define "best service".

Let's not turn this into NAC vs Internap thread. I did not mean that to happen. I just pointed out (from my direct experience with NAC and Internap) that Internap's network is better. So, please do not try to refute my experience. If my experience is wrong, then please prove it wrong.

FHDave

01-08-2003, 11:55 AM

Originally posted by nacjamesc
You make it sound like they are the only ones to do this type of thing or they invented it or something. It is definetly not InterNAP proprietary

Please read my comments carefully. I even included Inflow, UUNet, etc when talking about good redundancy network design.
I am sure they are not the only one. That's not the question. But But I am pretty sure designing a redundant network is not easy that any datacenter can do, isn't it?

hostpath.com

01-08-2003, 11:56 AM

InterNAP support is fantastic, which is key for a serious business. Can't comment on NAC.

I remember some time back doing an update to a server in a rack at InterNAP and rebooting the machine following the update. Well, about 30 seconds later my phone rings. It's InterNAP, who saw the reboot and wanted to know if everything's okay with my server. And I don't pay them to monitor my server uptime.

Thumbs up, baby.

FHDave

01-08-2003, 11:59 AM

Not sure when this started, but is this what you will get becuase of the free peering? Congested network that happens at random time? This is a traceroute result from Boston to NAC.

traceroute to nac.net (207.99.0.69), 64 hops max, 40 byte packets
2 border7.fe5-20.fluidhosting-1.bsn.pnap.net (66.150.201.251) 0.423 ms 0.338 ms 0.322 ms
3 core3.ge1-0-0-bbnet2.bsn.pnap.net (63.251.128.67) 3.479 ms 0.400 ms 0.392 ms
4 12.125.40.117 (12.125.40.117) 1.084 ms 1.425 ms 1.771 ms
5 gbr1-p52.cb1ma.ip.att.net (12.123.40.90) 1.172 ms 1.191 ms 2.308 ms
6 tbr1-p013401.cb1ma.ip.att.net (12.122.11.193) 1.981 ms 2.056 ms 1.546 ms
7 tbr2-p013701.n54ny.ip.att.net (12.122.10.22) 7.263 ms 7.067 ms 8.647 ms
8 gar1-p340.nwrnj.ip.att.net (12.123.214.185) 7.671 ms 8.242 ms 7.004 ms

9 12.119.140.30 (12.119.140.30) 144.091 ms 147.617 ms 141.256 ms
10 1189.at-0-1-0.gbr1.oct.nac.net (209.123.11.69) 151.752 ms 157.548 ms 154.141 ms
11 0095.gi-1-1.msfc1.oct.nac.net (64.21.102.2) 161.567 ms 159.620 ms 164.579 ms
12 mail.nac.net (207.99.0.69) 157.159 ms 155.581 ms 158.763 ms

Time now: 11:00AM EST

I can see random packet losses on the last two hops

tazzy

01-08-2003, 12:18 PM

Originally posted by nacjamesc

NAC has 3 primary transit carriers. Williams, AT&T, and Cogent. We also have backup transit through MFN.

You are absolutely wrong on this. The InterNAP model is to NOT do ANY peering, but to purchase transit. They do this so a) they are always a customer of someone so they get the fastest response time and b) they avoid the overcongested exchange points.

Cant argue with point A, but B is a thing of the past. It happens, but its very rare to see these days, so its effectively moot.

And

Followed what model? What have you been smoking?? . Every provider i know of prefers to peer as much transit as possible as it's *FREE*, cogent only peers because they cant afford to buy transit, the true definition of tier1 defines providers who peer and don't need to buy transit. You dont think the "big boys" in the carrier market actually *buy* transit do you?

Been smoking my usual

If you look at :
http://www.linx.net/members/inap/index.thtml

You'll see that its a company called I-Nap, not InterNAP. They are in no way affilliated with each other.

Without maintaining a network, peering would be useless because only their London based customers would have access to the peers.

Yes. NAC has better connectivity to alot companies.

Peering is always better than transit, in my opinion. You get direct connectivity. You cant beat that.

For example, from InterNAP, do a traceroute to www.aol.com or www.cv.net (Cablevision). Since InterNAP doesnt peer with anyone, it has to traverse many networks before getting to its destination.

From the NAC Network, it takes 8ms to get to www.aol.com, and 4ms for www.cv.net. What does this mean? NAC gets your packets to the EYEBALLS faster. Ofcourse the difference is relatively immeasurable, but every little bit counts.

I'm glad you have love for InterNAP; I do not want to turn this into a InterNAP vs NAC war. What i'm saying is that alot of people dont even know what makes a provider better than another one besides what they see as far as uptime. That definetly matters but other things are more important.

IE: (This is hypothetical, no provider names involved) Would you rather have a provider that is up ALL the time but has terrible connectvity or would you rather one that has its fair share of growing pains but still maintains 99.999% uptime, but provides the best service possible?

InterNAP is quite good, but dont rule NAC out, especially if you dont know what makes them good.

Hello,

I don't have anything against NAC, infact I get very good pings/traceroute's to NAC. I use AOL so I go directly to NAC. I'm also based in the UK and know lots about LINX.

Followed what model? What have you been smoking?? . Every provider i know of prefers to peer as much transit as possible as it's *FREE*, cogent only peers because they cant afford to buy transit, the true definition of tier1 defines providers who peer and don't need to buy transit. You dont think the "big boys" in the carrier market actually *buy* transit do you?

Been smoking my usual.:pimp: :smokin:
I said MOTTO not model, I guess you been smoking some as well then. ;)
I know all the "big boys" peer and do not buy transit. They don't need to. I'm just pointing out a fact.

Also, Some traceroutes:

>tracert nac.net

Tracing route to nac.net [207.99.0.69]
over a maximum of 30 hops:

3 139 ms 142 ms 131 ms accessl1-loh-P6-3.router.aol.com [195.93.52.117]

4 129 ms 132 ms 134 ms pop4-loh-P3-0.atdn.net [66.185.146.69]
5 141 ms 134 ms 135 ms bb2-loh-P0-2.atdn.net [66.185.146.66]
6 201 ms 199 ms 198 ms bb1-new-P5-0.atdn.net [66.185.152.140]
7 202 ms 232 ms 204 ms bb1-ash-P13-0.atdn.net [66.185.152.48]
8 217 ms 206 ms 198 ms pop2-ash-P0-0.atdn.net [66.185.139.209]
9 213 ms 199 ms 213 ms NetAccess.atdn.net [66.185.151.158]
10 213 ms 215 ms 211 ms 0195.at-0-1-0.gbr1.oct.nac.net [209.123.11.138]

11 225 ms 227 ms 199 ms 0095.gi-1-1.msfc1.oct.nac.net [64.21.102.2]
12 221 ms 213 ms 207 ms mail.nac.net [207.99.0.69]

Trace complete.

>tracert fluidhosting.com

Tracing route to fluidhosting.com [66.150.201.71]
over a maximum of 30 hops:

3 139 ms 135 ms 127 ms accessl1-loh-P4-3.router.aol.com [195.93.52.109]

4 146 ms 143 ms 127 ms pop4-loh-P2-0.atdn.net [66.185.146.73]
5 129 ms 126 ms 135 ms bb2-loh-P0-2.atdn.net [66.185.146.66]
6 195 ms 207 ms 190 ms bb1-new-P5-0.atdn.net [66.185.152.140]
7 211 ms 206 ms 207 ms bb1-ash-P13-0.atdn.net [66.185.152.48]
8 204 ms 199 ms 198 ms pop2-ash-P0-0.atdn.net [66.185.139.209]
9 195 ms 199 ms 199 ms sl-st20-ash-15-2.sprintlink.net [144.223.246.13]

10 211 ms 207 ms 206 ms sl-bb27-rly-15-3.sprintlink.net [144.232.20.6]
11 215 ms 206 ms 215 ms sl-bb20-rly-13-0.sprintlink.net [144.232.25.17]

12 203 ms 207 ms 199 ms sl-bb25-nyc-14-1.sprintlink.net [144.232.18.254]

13 210 ms 206 ms 207 ms sl-gw18-nyc-9-0.sprintlink.net [144.232.13.166]

14 220 ms 214 ms 215 ms sl-internap-77-0.sprintlink.net [144.232.235.178
]
15 219 ms 215 ms * border8.ge1-1-bbnet1.bsn.pnap.net [63.251.128.13
]
16 214 ms 214 ms 214 ms fluidhosting-1.border7-8.bsn.pnap.net [66.150.20
1.250]
17 218 ms 214 ms 206 ms mail.fluidhosting.com [66.150.201.71]

Trace complete.

sjs

01-08-2003, 12:21 PM

Originally posted by FHDave
Not sure when this started, but is this what you will get becuase of the free peering? Congested network that happens at random time? This is a traceroute result from Boston to NAC.

[code]

I can see random packet losses on the last two hops

well i really didn't want to take it there, but:

7 sl-bb21-nyc-10-0.sprintlink.net (144.232.18.250) 71 ms 227 ms 92 ms
8 sl-gw18-nyc-10-0.sprintlink.net (144.232.13.161) 81 ms 155 ms 142 ms
9 sl-internap-77-0.sprintlink.net (144.232.235.178) 134 ms 235 ms 212 ms
10 border8.ge1-1-bbnet1.bsn.pnap.net (63.251.128.13) 210 ms 291 ms 180 ms
11 fluidhosting-1.border7-8.bsn.pnap.net (66.150.201.250) 193 ms 138 ms 140 ms
12 mail.fluidhosting.com (66.150.201.71) 159 ms 155 ms 160 ms

route-views.oregon-ix.net>traceroute www.fluidhosting.com.
Translating "www.fluidhosting.com."...domain server (128.223.32.35) [OK]

Type escape sequence to abort.
Tracing the route to www.fluidhosting.com (66.150.201.71)

1 nero-gw.oregon-ix.net (198.32.162.2) [AS 2914] 0 msec 0 msec 0 msec
2 ptck-core2-gw.nero.net (207.98.64.2) [AS 3701] 4 msec 4 msec 0 msec
3 ptck-core1-gw.nero.net (207.98.64.137) [AS 3701] 4 msec 4 msec 0 msec
4 POS6-1.hsipaccess2.Seattle1.Level3.net (63.211.200.245) [AS 3356] 8 msec 4 msec 8 msec
5 * * *
6 * * *
7 * * *
8 * * *
9 * * *
10 * * *

route-views.oregon-ix.net>show ip bgp 66.150.201.71 | i flap
Dampinfo: penalty 673, flapped 2 times in 00:08:59

route flapping is bad m'kay.

sjs

01-08-2003, 12:30 PM

9 bb2-ash-p15-0.atdn.net (66.185.152.54) 2.412 ms 2.287 ms 1.934 ms
10 pop2-ash-p1-0.atdn.net (66.185.139.211) 1.752 ms 2.292 ms 2.058 ms
11 sl-st20-ash-15-2.sprintlink.net (144.223.246.13) 2.26 ms 2.146 ms 1.866 ms
12 sl-bb27-rly-15-3.sprintlink.net (144.232.20.6) 4.2 ms 3.295 ms 3.641 ms
13 sl-bb20-rly-13-0.sprintlink.net (144.232.25.17) 4.404 ms 4.027 ms 4.676 ms
14 sl-bb25-nyc-14-1.sprintlink.net (144.232.18.254) 30.901 ms 229.4 ms 207.616 ms
15 sl-gw18-nyc-9-0.sprintlink.net (144.232.13.166) 6.423 ms 6.522 ms 6.97 ms
16 sl-internap-77-0.sprintlink.net (144.232.235.178) 15.258 ms 14.884 ms 15.976 ms
17 border8.ge1-2-bbnet2.bsn.pnap.net (63.251.128.77) 18.917 ms 18.914 ms 18.135 ms
18 fluidhosting-1.border7-8.bsn.pnap.net (66.150.201.250) 19.421 ms 18.633 ms 18.263 ms
19 mail.fluidhosting.com (66.150.201.71) 17.572 ms 18.122 ms 17.977 ms

and:

8 bb2-rtc-p0-1.atdn.net (66.185.140.98) 1.978 ms 2.101 ms 1.582 ms
9 bb2-ash-p15-0.atdn.net (66.185.152.54) 2.266 ms 1.681 ms 1.638 ms
10 pop2-ash-p1-0.atdn.net (66.185.139.211) 2.378 ms 2.144 ms 2.535 ms
11 netaccess.atdn.net (66.185.151.158) 2.09 ms 2.054 ms 1.652 ms
12 0195.at-0-1-0.gbr1.oct.nac.net (209.123.11.138) 9.103 ms 8.624 ms 8.506 ms
13 0095.gi-1-1.msfc1.oct.nac.net (64.21.102.2) 8.329 ms 8.919 ms 8.88 ms
14 www.nac.net (207.99.0.101) 8.597 ms 8.715 ms 9.298 ms

See since InternNAP doesn’t peer, traffic from a network such as AOL needs to traverse through one of InterNAP’s transit providers. While I’m not going to say that fewer hops are better I will say that less autonomous network hops are better. The more autonomous network hops you have, the more room for failure.

FHDave

01-08-2003, 12:31 PM

Also, Some traceroutes:

Thanks tazzy. It seems there is a major lag inside ATT network right now and hence it's not shown on your traceroute to NAC, but on my traceroute to from Internap via ATT.

Also from your end, there seems to be no difference between NAC's direct peering with AOL and Internap lack of thereof. Interesting.

Anyway, just wonderring ... since it seems ATT is having a bad day today, can NAC remove ATT from its routing table and route it from other providers to prevent unnecessary lags until ATT is back normal again?

sjs

01-08-2003, 12:35 PM

Originally posted by FHDave

Anyway, just wonderring ... since it seems ATT is having a bad day today, can NAC remove ATT from its routing table and route it from other providers to prevent unnecessary lags until ATT is back normal again?

I do not believe the problem is within AT&T. It’s the return route, which is going via Verio’s network.

% traceroute www.fluidhosting.com
traceroute to www.fluidhosting.com (66.150.201.71), 30 hops max, 40 byte packets
1 ge-2-0.r00.asbnva01.us.bb.verio.net (206.223.115.12) 97.721 ms 91.762 ms 88.921 ms
2 p4-1-0-0.r02.asbnva01.us.bb.verio.net (129.250.2.234) 85.966 ms 84.386 ms 83.172 ms
3 p16-0-1-2.r21.asbnva01.us.bb.verio.net (129.250.2.62) 82.832 ms 83.564 ms 86.191 ms
MPLS Label=118584 CoS=0 TTL=1 S=1
4 p16-0-1-1.r21.nycmny01.us.bb.verio.net (129.250.5.98) 92.615 ms 96.996 ms 104.897 ms
MPLS Label=103827 CoS=0 TTL=1 S=1
5 p16-7-0-0.r04.nycmny01.us.bb.verio.net (129.250.3.49) 103.064 ms 106.517 ms 108.109 ms
6 p4-0-0-0.r00.bstnma01.us.bb.verio.net (129.250.4.214) 113.044 ms 113.813 ms 105.293 ms
7 * * *

tazzy

01-08-2003, 12:35 PM

Originally posted by FHDave

Thanks tazzy. It seems there is a major lag inside ATT network right now and hence it's not shown on your traceroute to NAC, but on my traceroute to from Internap via ATT.

Also from your end, there seems to be no difference between NAC's direct peering with AOL and Internap lack of thereof. Interesting.

Anyway, just wonderring ... since it seems ATT is having a bad day today, can NAC remove ATT from its routing table and route it from other providers to prevent unnecessary lags until ATT is back normal again?

Hello,

Yes. I always get good pings to yourselves and nac.net. NAC peers directly with AOL. However, I would still go with InterNAP over NAC, simply because not everyone uses AOL you know. Also, InterNAP performs very well against NAC in the AOL stakes and slightly better in most other tests I have done for global locations.

tazzy

01-08-2003, 12:41 PM

Traceroutes:

>tracert fluidhosting.com

Tracing route to fluidhosting.com [66.150.201.71]
over a maximum of 30 hops:

3 131 ms 142 ms 135 ms accessl1-loh-P4-3.router.aol.com [195.93.52.109]

4 132 ms 125 ms 135 ms pop4-loh-P2-0.atdn.net [66.185.146.73]
5 140 ms 134 ms 135 ms bb2-loh-P0-2.atdn.net [66.185.146.66]
6 204 ms 198 ms 199 ms bb1-new-P5-0.atdn.net [66.185.152.140]
7 204 ms 199 ms 199 ms bb1-ash-P13-0.atdn.net [66.185.152.48]
8 197 ms 216 ms 204 ms pop2-ash-P0-0.atdn.net [66.185.139.209]
9 209 ms 207 ms 207 ms sl-st20-ash-15-2.sprintlink.net [144.223.246.13]

10 219 ms 215 ms 207 ms sl-bb27-rly-15-3.sprintlink.net [144.232.20.6]
11 212 ms 198 ms 215 ms sl-bb20-rly-13-0.sprintlink.net [144.232.25.17]

12 204 ms 206 ms 206 ms sl-bb25-nyc-14-1.sprintlink.net [144.232.18.254]

13 211 ms 207 ms 207 ms sl-gw18-nyc-9-0.sprintlink.net [144.232.13.166]

14 219 ms 215 ms 206 ms sl-internap-77-0.sprintlink.net [144.232.235.178
]
15 219 ms 214 ms 207 ms border8.ge1-1-bbnet1.bsn.pnap.net [63.251.128.13
]
16 203 ms 223 ms 206 ms fluidhosting-1.border7-8.bsn.pnap.net [66.150.20
1.250]
17 218 ms 214 ms 215 ms mail.fluidhosting.com [66.150.201.71]

Trace complete.

>tracert nac.net

Tracing route to nac.net [207.99.0.69]
over a maximum of 30 hops:

3 131 ms 142 ms 127 ms accessl1-loh-P6-3.router.aol.com [195.93.52.117]

4 123 ms 134 ms 127 ms pop4-loh-P3-0.atdn.net [66.185.146.69]
5 130 ms 134 ms 127 ms bb2-loh-P0-2.atdn.net [66.185.146.66]
6 203 ms 199 ms 198 ms bb1-new-P5-0.atdn.net [66.185.152.140]
7 203 ms 215 ms 199 ms bb1-ash-P13-0.atdn.net [66.185.152.48]
8 195 ms 207 ms 206 ms pop2-ash-P0-0.atdn.net [66.185.139.209]
9 211 ms 206 ms 199 ms NetAccess.atdn.net [66.185.151.158]
10 211 ms 207 ms 206 ms 0195.at-0-1-0.gbr1.oct.nac.net [209.123.11.138]

11 218 ms 206 ms 207 ms 0095.gi-1-1.msfc1.oct.nac.net [64.21.102.2]
12 212 ms 207 ms 214 ms mail.nac.net [207.99.0.69]

Trace complete.

>tracert pwebtech.com

Tracing route to pwebtech.com [64.21.5.34]
over a maximum of 30 hops:

3 138 ms 142 ms 134 ms accessl1-loh-P6-3.router.aol.com [195.93.52.117]

4 131 ms 134 ms 135 ms pop4-loh-P2-0.atdn.net [66.185.146.73]
5 131 ms 134 ms 135 ms bb2-loh-P0-2.atdn.net [66.185.146.66]
6 196 ms 199 ms 208 ms bb1-new-P5-0.atdn.net [66.185.152.140]
7 204 ms 198 ms 199 ms bb1-ash-P13-0.atdn.net [66.185.152.48]
8 203 ms 207 ms 207 ms pop2-ash-P0-0.atdn.net [66.185.139.209]
9 218 ms 199 ms 207 ms NetAccess.atdn.net [66.185.151.158]
10 211 ms 207 ms 206 ms 0195.at-0-1-0.gbr1.oct.nac.net [209.123.11.138]

11 203 ms 215 ms 295 ms 0095.gi-1-1.msfc1.oct.nac.net [64.21.102.2]
12 215 ms 215 ms 206 ms fa0-0.core1.pwebtech.com [207.99.8.198]
13 210 ms 206 ms 207 ms pegasushost.com [64.21.5.34]

Trace complete.

>tracert rackshack.net

Tracing route to rackshack.net [207.218.223.142]
over a maximum of 30 hops:

3 132 ms 126 ms 127 ms accessl2-loh-P6-3.router.aol.com [195.93.52.133]

4 134 ms 132 ms 127 ms pop5-loh-P2-0.atdn.net [66.185.146.89]
5 139 ms 142 ms 137 ms bb2-loh-P0-3.atdn.net [66.185.146.82]
6 197 ms 198 ms 194 ms bb1-new-P5-0.atdn.net [66.185.152.140]
7 204 ms 271 ms 207 ms bb1-ash-P13-0.atdn.net [66.185.152.48]
8 204 ms 207 ms 198 ms pop1-ash-P0-0.atdn.net [66.185.139.193]
9 211 ms 214 ms 223 ms AllegianceTelecom.atdn.net [66.185.151.98]
10 260 ms 254 ms 247 ms dca1-core12-pos2-0.atlas.algx.net [165.117.68.19
3]
11 285 ms 254 ms 246 ms dca1-core13-pos7-0.atlas.algx.net [165.117.48.20
6]
12 243 ms 206 ms 206 ms dca10-core1-so-7-2-0.atlas.algx.net [165.117.56.
22]
13 251 ms 247 ms 247 ms atl10-core2-so-0-2-0-0.atlas.algx.net [165.117.2
00.5]
14 243 ms 247 ms 246 ms atl10-core1-so-0-1-0-0.atlas.algx.net [165.117.1
92.1]
15 243 ms 238 ms 247 ms dfw10-core2-so-0-2-0-0.atlas.algx.net [165.117.2
00.2]
16 242 ms 239 ms 238 ms dfw10-core1-so-0-1-0-0.atlas.algx.net [165.117.1
92.17]
17 250 ms 255 ms 247 ms iah10-core2-so-2-0-0-0.atlas.algx.net [165.117.2
00.54]
18 244 ms 246 ms 254 ms iah10-edge1-pos7-0.atlas.algx.net [165.117.192.1
46]
19 * * ^C

>tracert rackspace.com

Tracing route to rackspace.com [64.39.2.185]
over a maximum of 30 hops:

3 131 ms 134 ms 127 ms accessl1-loh-P4-0.router.aol.com [195.93.52.45]

4 132 ms 134 ms 126 ms pop4-loh-P2-0.atdn.net [66.185.146.73]
5 130 ms 135 ms 127 ms bb2-loh-P0-2.atdn.net [66.185.146.66]
6 195 ms 191 ms 198 ms bb1-new-P5-0.atdn.net [66.185.152.140]
7 202 ms 199 ms 199 ms bb1-ash-P13-0.atdn.net [66.185.152.48]
8 203 ms 207 ms 206 ms pop1-ash-P0-0.atdn.net [66.185.139.193]
9 211 ms 207 ms 207 ms winstar-gw.napny.ip.att.net [192.205.32.5]
10 1204 ms 751 ms 1502 ms tbr1-p014001.wswdc.ip.att.net [12.123.9.82]
11 1724 ms 238 ms 239 ms tbr1-p013701.sl9mo.ip.att.net [12.122.10.66]
12 * * ^C

As you can see, Pwebtech performed the best (if you exclude the one odd 295ms). RackShack the worst and Rackspace didn't at all due to AT&T problems.

Deb

01-08-2003, 02:23 PM

During our presales go around with Internap we did ask about the peering vs. paid transit. We, of course, were in communication with multiple providers during that time and one provider (I'll leave unnamed) had noted, in a round about way that going with Internap was silly since everyone offered the same thing anyway. The [OtherProvider] noted "They aren't doing anything Qwest, UUNET, AT&T, etc. aren't already doing too." We sent his comments to Internap which formed a question that was something like; "You boast 9 of the Tier 1 providers, yet any self respecting provider peers with most of these providers as well. That's just how the Internet works. Is this just a marketing spin or is there really something different than what the others are providing?" The answer included the following:

--
Peering, no matter how ubiquitous it gets, does not address the performance
concerns that plague the public infrastructure - latency, packet loss,
consistency, bandwidth capacity, etc. Peering between providers is void of
economic settlement. As a result, providers circuitously route traffic
based on convenience, not performance. Peering, whether it's public or
private, often is the problem.

[Mr. "OtherProvider"'s] understanding of the Internap model is incorrect and
misrepresented. Internap does NOT peer with our providers for all the above
reasons. Rather, we PAY for upstream connectivity to 8-9 of the top Tier 1
providers. As a paying customer, Internap gets highest priority routing,
across all the major Tier 1 backbones, for on-net traffic. Thus, Internap
can reach the overwhelming majority of the global IP destinations on-net.
In addition, our Assimilator technology, unlike BGP, dynamically measures
latency, packet loss, capacity, consistency, and other relevant performance
metrics, across all the major Tier 1 backbones, to determine the most
consistent, best performing path from source to destination. Internap's
Video and Voice over IP customers will be the first to tell you how the
Assimilator technology enables their applications.

[Mr. "OtherProvider"'s] is correct in uncovering the limitations of BGP4 routing.
That is, BGP will choose best path purely based on the AS hop count. So,
theoretically, the provider who has more off ramps or better peering
agreements may have a better chance of being chosen "best path", as defined
by BGP protocol. BGP chose [Provider] strictly based on the number of AS hops
not because that path consistently performs better or "[Provider] has better
peering arrangements than Internap". Again, Internap does not have peering
arrangements.

That being said...

Please continue to monitor our performance, work with our NOC, and you'll
see the difference first hand.

-----

Since we do not -only- use Internap we're able to see what's working best for our clients etc. As noted earlier in this thread I'm pleased with both Qwest and Internap (I have not personally done business with NAC yet so couldn't say anything about them from this side of the fence). I can just say again that I'm quite pleased with Internap and toss a reminder that I'm also quite pleased with a non-Internap provider ;)

Traceroutes change throughout the day and there are many factors that work to create their end results. The location of PointA vs the competing PointB's is a big deal. More hops isn't necessarily worse, sometimes it can be better. Routing around congestion or outages may mean a 'slower route' than another is seeing but it sure beats taking what is normally the fastest route if that route is currently choking. At any rate the whole traceroute deal is for a different thread LOL The above is just one answer that came straight from Internap to address the "Peering vs. Paid" comments within this thread.

Deb

01-08-2003, 02:35 PM

P.S. When it comes to peering... I think it's worth also noting that Qwest is doing an excellent job in the area. They often beat Internap in our own routing due to the quality of their peering.

hostpath.com

01-08-2003, 02:39 PM

InterNAP's response to Deb is exactly on point, and their technology is, indeed, superior to BGP. On top of that, InterNAP remaps their network every day to ensure low latency. In fact, they have a latency guarantee in their SLA.

Like I said, I can't comment on NAC, they may be absolutely terrific. But I can comment on InterNAP and I say they're truly first rate.

porcupine

01-08-2003, 02:43 PM

Originally posted by Deb
P.S. When it comes to peering... I think it's worth also noting that Qwest is doing an excellent job in the area. They often beat Internap in our own routing due to the quality of their peering.

It's too bad that qwest can do such a good job on some things, then ignore equally essential parts of running a network, like dealing with abuse/answering their abuse emails/staffing a 24x7 department thats actually capable of dealing with abuse. Nothing like cracking down on a DoS, Calling the offending IP's uplink by uplink, getting to Qwest and having their monkey tell you "im sorry, but the department that handles that doesen't get back in until monday at 9:00" :rolleyes:.

mgriffin

01-08-2003, 02:43 PM

Just as an FYI, intelligent routing hardware is now available which can create a Internap-like configuration...

http://www.routescience.com/index2.html
http://www.netvmg.com/index2.html

- Mike

mgriffin

01-08-2003, 03:16 PM

Originally posted by Deb
P.S. When it comes to peering... I think it's worth also noting that Qwest is doing an excellent job in the area. They often beat Internap in our own routing due to the quality of their peering.

Unfortunately around Denver Qwest's peering (policies) have a lot to be desired. While Qwest has direct peering with AT&T, UUNET, Genuity, Cable & Wireless (and possibly others) directly in Denver, a good chunk of their inbound routing passes through Seattle, WA (???).

- Mike

nacjamesc

01-08-2003, 06:09 PM

Lets also chat for a moment about provider redundancy. From what you say FHdave, (Re: Pure InterNAP bandwidth) what if InterNAP was to fall off the face of the earth tomorrow? Wouldnt it be wise to multihome? For all of the SLA and uptime you talk, it would seem foolish to leave provider diversity out of the redundancy equation, wouldn't it?

XN CEO

01-08-2003, 06:11 PM

NAC, great choice for me and my clients, not one problem in the last 12 months. That goes for my company, and for the company i worked for before.

I would go with NAC anyday.

FHDave

01-08-2003, 06:22 PM

Originally posted by nacjamesc
Lets also chat for a moment about provider redundancy. From what you say FHdave, (Re: Pure InterNAP bandwidth) what if InterNAP was to fall off the face of the earth tomorrow? Wouldnt it be wise to multihome? For all of the SLA and uptime you talk, it would seem foolish to leave provider diversity out of the redundancy equation, wouldn't it?

I don't think they will just disappear without notice in 24 hours :)
If they are to go out of business, then there will be plenty of time to reallocate to the nearby datacenters :)

nacjamesc

01-08-2003, 06:30 PM

What I mean is, what if InterNAP closes shop over night, or the rack with InterNAPs routers (yes, all of the ones that you connect to) completely loses power or blows up and takes 492394 hours to replace. Your rack still has power and is just tossing packets to /dev/null. If you are really concerned about 100% uptime, you would have thought about this a while ago.

Its completely off topic, I know this, but this thread is already gone askew.

FHDave

01-08-2003, 06:46 PM

Originally posted by nacjamesc
What I mean is, what if InterNAP closes shop over night
or the rack with InterNAPs routers (yes, all of the ones that you connect to) completely loses power or blows up

Just changing "Internap" -> NAC and rephrase it:

So, "what if NAC closes shop over night or the rack with NAC's routers completely loses power or blows up?"

Those question is a hypotethical one and may happen to any particular datacenter (perhaps due to terorrist attack). What really matters is, aside of the large scale problem as that, what kind of redundancy is built in into the network?

Here is Internap's redundancy according to my limited understanding of it (please do not take this as if I am trying to promote my company, as a particular newbie tends to think so) .

Internap brings in nine providers (UUNet, ATT, Sprint, Verio, CW, BBN/Genuity, Global Crossing, QWest, and Digex) into its datacenter. Nine providers via nine diverse routes. The nine providers terminate in Internap's two border routers, each of these border routers has its own backup router (I believe). Now Internap groups its nine providers into two, each groups will be connected to each of Internap's border routers. So Internap may route, for example, UUNet, Sprint, CW, and BBN/Genuity into its first border router and the rest of the providers into its second border router. Then, for its customers Internap will provide 2x100 Mbps ethernet drops into the customers' cabinet. Each of the ethernet drops is connected to each of Internap's broder routers.

Even if one of Internap's border router and its backup completely fails, there is still another border router to fail over to. Of course I do not need to mention that each router is powered by redundant power supply from a separate power grids, etc. Nor do I need to explain the standar redundancy in the power supply (UPS power backup, Diesel power backup, etc). When visiting their datacenter, I asked this question to their network engineer; "what's the possibility for the network to completely go down". To which he answers "well, if all routers blow up at the same time". But there is negligible chance for Internap's all routers to go down at the same time leaving a total network blackout. Of course I will not argue that there is still possibility, but this possibility can as well happen with any datacenter, including NAC :)

What we, as customers, care about at the end is the SLA guarantee; network uptime guarantee, packet losses guarantee, and network latency guarantee.

XN CEO

01-08-2003, 06:53 PM

Hope you dont mind lil ol me popping in here.

Question, and Im not favoring to either side here but...

At the end of the day, contract agreements dont matter. Even if someone guarantees 99.9% uptime. What if they shut down? Sure, you'll get your money or sue them, but you'll still be offline. So doesnt it matter that you actually stay up; not just what you say you'll stay up?

nacjamesc

01-08-2003, 06:56 PM

Here is Internap's redundancy according to my limited understanding of it (please do not take this as if I am trying to promote my company, as a particular newbie tends to think so) .

Internap brings in nine providers (UUNet, ATT, Sprint, Verio, CW, BBN/Genuity, Global Crossing, QWest, and Digex) into its datacenter.

I've already heard InterNAPs model like 90 times, FEEL FREE to stop redescribing.

Nine providers via nine diverse routes.

If you've ever maintained a real WAN you'd know that this is COMPLETELY incorrect. There are not even 9 diverse entrance/exit points in any of the major carrier hotels.

Just changing "Internap" -> NAC and rephrase it:

So, "what if NAC closes shop over night or the rack with NAC's routers completely loses power or blows up?"

Just change that to any name; this still doesnt answer the question.

I was speaking of Fluid Hosting as a whole,

BUT...

I'll just stop this right here while we're ahead.

FHDave

01-08-2003, 06:59 PM

I've already heard InterNAPs model like 90 times, FEEL FREE to stop redescribing

Hm .. then how come you still ask that hypothetical questions?

Feel free also to consider me as your potential customer that you may want to have business in the future. So, feel free to convince me how NAC will be better than Internap. Then, I may be thinking of moving my cabinet over. Until then, I am not sure what you are trying to do especially with me and my company.

cheers

nacjamesc

01-08-2003, 07:02 PM

Originally posted by FHDave
feel free to convince me how NAC will be better than Internap. Then, I may be thinking of moving my cabinet over.

cheers

Perhaps you are confused. I was having a technical/operational descussion, NOT trying to solicit or attempt you to think NAC was better than anyone. You are entitled to like whomever you like. I would love it if I was in your situation and had complete love for my ISP/NSP cause I know they are few and far between.

FHDave

01-08-2003, 07:06 PM

Originally posted by nacjamesc
Perhaps you are confused

I just re-read your comments again. I think you are confused with us too, especially with the "multihome" part. We are multihome and directly connected, through Internap's border routers, to nine providers. Am I right in saying this?

We do not get a bandwidth from Internap through any local loop, if that's what you mean. We are inside Internap's datacenter, feel free to re-read my description above ...

nacjamesc

01-08-2003, 07:11 PM

Originally posted by FHDave

I just re-read your comments again. I think you are confused with us too, especially with the "multihome" part. We are multihome and directly connected, through Internap's border routers, to nine providers. Am I right in saying this?

No. Your provider is multihomed, You are not.

InterNAP == multihomed
FluidHosting == NOT multihomed

The term multihomed refers to provider diversity, not intraprovider diversity, meaning you physically have connectivity to two or more providers, NOT two physical connections to ONE provider.

I hope this clears things up. :)

FHDave

01-08-2003, 07:16 PM

Originally posted by nacjamesc
No. Your provider is multihomed, You are not.

InterNAP == multihomed
FluidHosting == NOT multihomed

Feel free to elaborate more.

Is NAC multihomed? Will I be multihomed if I have server coloed at NAC?

nacjamesc

01-08-2003, 07:53 PM

Originally posted by FHDave

Feel free to elaborate more.

Is NAC multihomed? Will I be multihomed if I have server coloed at NAC?

NAC is quite multihomed. We have many transit providers such as Williams, AT&T, Cogent, MFN and UUNet. In addition we have connections to several peering exchange points and such.

Unless you were to get connectivity to another ISP/NSP, you (your company) would not be multihomed. Multihoming usually consists of running BGP with both providers, although I have seen some ..ahem..ghetto setups which avoid BGP (Not recommended! :).

I hope you are not thinking i'm being malicious or flaming you, just trying to make the community more aware as I know alot of people are unsure the real difference between multihoming and having two connections to their provider.

I am not a sales person what so ever. I would gladly welcome your business, however I would not be the best person to propose how your company would mesh with ours. If you are serious about purchasing service from us (It is hard to tell if you are being sarcastic or not) then please private message me and i'll get you in touch with the proper person. Again, I'm not trying to sell things here, just trying to be a good member of the hosting/network provider community by spreading useful information.

I think I'll stop posting on this thread from here on out though as its just getting out of hand.

Deb

01-08-2003, 07:54 PM

The key is having connections to more than one, vs. switching completely to a different one. It's not always the most economical choice but it is a wise choice.

Currently we use two separate providers that we are quite pleased with. When we are ready to add another DS3 it will be with a third provider and not with either of the ones we are already using. It's hard not to go with what we already know works but in the name of keeping our connectivity options open, even when catastrophe strikes, it's important.

Either of our current providers would love to sell us a bigger line, and boy do they try lol, but I have to agree with James in that it's important to remain multihomed with as much 'multi' as possible.

When you only have a single provider it's vital that the provider is a good one. You certainly have that! When it's time to expand, the suggestion is to take the 'next best one' you can find so that you can remain diverse.

FHDave

01-08-2003, 07:59 PM

Since when did I start to compare Fluid Hosting and NAC? But thank you for the honor!

Originally posted by nacjamesc
NAC is quite multihomed. We have many transit providers such as Williams, AT&T, Cogent, MFN and UUNet. In addition we have connections to several peering exchange points and such.

We are directly connected to Internap's routers (via ethernet drops) that's multihomed to nine providers. I don't understand how come Internap is multihomed and I am not.

You have not answered my questions. Yes, NAC is multihomed. But are you trying to say that I will not be multihomed if I get bandwidth from NAC inside NAC datacenter, connected directly to NAC's routers via ethernet drops?

nacjamesc

01-08-2003, 08:07 PM

We are directly connected to Internap's routers (via ethernet drops) that's multihomed to nine providers. I don't understand how come Internap is multihomed and I am not.

You have not answered my questions. Yes, NAC is multihomed. But are you trying to say that I will also be not multihomed if I get bandwidth from NAC inside NAC datacenter, connected directly to NAC's routers via ethernet drops?

Until the day comes that you call up Joe ISP and get service from them over a physically different circuit (in addition to Internap), you will not be multihomed. That holds true no matter which provider you are with. If you only connect to them, its a single point of failure EVEN IF you have multiple connections to said provider. Just saying that InterNAP has like 4 routers that you are connected to doesnt mean that they couldnt have a routing problem which would cause your sites not to be accessible from the outside world. Getting connectivity to another provider independent of and in addition to Internap is the only solution to truly be multihomed.

Youd have to run BGP with Internap and Joe ISP, and announce your IP address blocks to both providers. If one fails, the other will take over (Just like Internap having 9 providers; if one has issues, others will take over)

Does this make more sense now?

FHDave

01-08-2003, 08:13 PM

Originally posted by nacjamesc
Until the day comes that you call up Joe ISP and get service from them over a physically different circuit (in addition to Internap), you will not be multihomed. That holds true no matter which provider you are with. If you only connect to them, its a single point of failure EVEN IF you have multiple connections to said provider. Just saying that InterNAP has like 4 routers that you are connected to doesnt mean that they couldnt have a routing problem which would cause your sites not to be accessible from the outside world. Getting connectivity to another provider independent of and in addition to Internap is the only solution to truly be multihomed.

Youd have to run BGP with Internap and Joe ISP, and announce your IP address blocks to both providers. If one fails, the other will take over (Just like Internap having 9 providers; if one has issues, others will take over)

You still have not answered my question. So let me ask you for the 90th times.

"Yes, NAC is multihomed. But are you trying to say that I will not be multihomed if I get bandwidth from NAC inside NAC datacenter, connected directly to NAC's routers via ethernet drops? "

Just a simple Yes or No will suffice. Thanks ...

nacjamesc

01-08-2003, 08:18 PM

Originally posted by FHDave

You still have not answered my question. So let me ask you for the 90th times.

"Yes, NAC is multihomed. But are you trying to say that I will not be multihomed if I get bandwidth from NAC inside NAC datacenter, connected directly to NAC's routers via ethernet drops? "

Just a simple Yes or No will suffice.

NO

As I said previously being multihomed or not has nothing to do with which providers; just the numberof providers. If you want to hear this with NAC's name in it then here:

If you connect to NAC you will not be multihomed. You are still relying on NAC as your only connection to the internet. EVEN if you get multiple connections to NAC, you will still be relying on NAC alone. NAC has multiple connections to the internet and is multhomed, but you as a customer are not multihomed unless you connect to another provider for yourself at the same time as NAC and announce your networks to both providers (or even just have another provider in standby mode, only preferring in and outbound traffic through your "primary" provider.

You are currently not multihomed with Internap as your only provider.

I really dont think this is too hard of a concept to understand; please let me know (I know you will anyway ;)) if you are not understanding and how I could possibly explain it better.

Sprynex

01-08-2003, 08:18 PM

FHDave

01-08-2003, 08:23 PM

No

Ah OK ... it's a fair comparison then :) So not-multihomed with NAC and not-multihomed with Internap. Which to choose? I will follow Deb's suggestion :)

Thanks .

nacjamesc

01-08-2003, 08:24 PM

You will not be multihomed with any provider if they are your only provider.

tazzy

01-08-2003, 08:28 PM

Originally posted by nacjamesc
You will not be multihomed with any provider if they are your only provider.

Hello,

Multi - short for multiple - more than one
Single - one

Multihomed is what InterNAP is. Using Verio only via Single DS3 - you would be singlehomed. Using InterNAP via dual 100Mbps ethernet drops to completely different routers. Then going onto on of 9 different carriers. So thats singlehomed as well is it?

Deb

01-08-2003, 08:33 PM

Ok... I can see it coming... so let's begin (?end?) with: Multihomed describes a computer host that has multiple IP addresses to connected networks. A multihomed host is physically connected to multiple data links that can be on the same or different networks. For example, a computer with a Windows NT 4.0 Server and multiple IP addresses can be referred to as "multihomed" and may serve as an IP router.
Using the Stream Control Transmission Protocol (SCTP), multihoming allows a single SCTP endpoint to support multiple IP addresses, which means that a session is more likely to survive a network failure. In a single-homed session, a network failure can isolate the end system or make transport temporarily unavailable. Multihoming means that redundant local area networks (LANs) can be used to support local access. Various approaches, such as using addresses with different prefixes to force routing through different carriers, or even using redundant core networks, can be taken to reduce the effects of failures.

Multihoming is commonly used in Web management for load balancing, redundancy, and disaster recovery. Did that make it worse? :D

FHDave

01-08-2003, 08:33 PM

Originally posted by tazzy
Using InterNAP via dual 100Mbps ethernet drops to completely different routers. Then going onto on of 9 different carriers. So thats singlehomed as well is it?

I don't know ... /me got confused a bit :)

The funny thing is, there is not single point of failure in our network (well, up to our core switch at this moment). There are two ethernet routes to two of Internap's border routers, each of which is connected to several different providers. Are we really still single homed? Where is the single point of failure?

nacjamesc

01-08-2003, 08:35 PM

Originally posted by tazzy

Hello,

Multi - short for multiple - more than one
Single - one

Multihomed is what InterNAP is. Using Verio only via Single DS3 - you would be singlehomed. Using InterNAP via dual 100Mbps ethernet drops to completely different routers. Then going onto on of 9 different carriers. So thats singlehomed as well is it?

Correct. You can have 90239123 connections to a provider to their 23929321932193 routers and still be down if (by chance) they have a global network issue. Youd still result in being down. Now had you had another provider who was running normally, coexisting with the other provider, you'd be up and running.

This is the whole point of why ISP/NSPs connect to multiple providers, (like NAC's 5, and Internaps 9) so if one fails, they can still be visible on the net through one of their other providers.

nacjamesc

01-08-2003, 08:36 PM

Originally posted by FHDave

The funny thing is, there is not single point of failure in our network (well, up to our core switch at this moment). There are two ethernet routes to two of Internap's border routers, each of which is connected to several different providers. Are we really still single homed? Where is the single point of failure?

The single point of failure is only having one provider. Internap has multiple providers for a reason; so should anyone who is serious about uptime.

tazzy

01-08-2003, 08:37 PM

Originally posted by FHDave

I don't know ... /me got confused a bit :)

The funny thing is, there is not single point of failure in our network (well, up to our core switch at this moment). There are two ethernet routes to two of Internap's border routers, each of which is connected to several different providers. Are we really still single homed? Where is the single point of failure?

Hello,

You are not. Your multihomed. Multiple routers the data can take. When one route may be down, the data can go the other way(s).

Sprynex

01-08-2003, 08:39 PM

That isn't correct..

What if internaps routing gettings whacked?? what happens then?? it will go to all their routers (they will share routes).. and then your DOWN.

if you had another connection to another provider then you would still be announced via them and be up.

so if you have *1* back bone provider, it does not matter how many connections to them you have.. you are SINGLE homed.

Internap has multiple providers, they are multihomed

YOU have internap, therefore all you have is internap, so you are single homed.

you have have 100 connections to internap and that won't change, as james said above.

tazzy

01-08-2003, 08:41 PM

Originally posted by nacjamesc

The single point of failure is only having one provider. Internap has multiple providers for a reason; so should anyone who is serious about uptime.

Hello,

Fluidhosting is located in an InterNAP operated data-center and they are connected directly to InterNAPs' multiple routers via multiple connections. InterNAP doesn't per sa have a network it simply uses 9 tier-1 bandwidth providers and buys transit from them and sells its product of providing that bandwidth using its intelligent gateway system.

Is it too hard to understand that Fluidhosting is multihomed?

tazzy

01-08-2003, 08:43 PM

Originally posted by Sprynex
That isn't correct..

What if internaps routing gettings whacked?? what happens then?? it will go to all their routers (they will share routes).. and then your DOWN.

if you had another connection to another provider then you would still be announced via them and be up.

so if you have *1* back bone provider, it does not matter how many connections to them you have.. you are SINGLE homed.

Internap has multiple providers, they are multihomed

YOU have internap, therefore all you have is internap, so you are single homed.

you have have 100 connections to internap and that won't change, as james said above.

Hello,

Yes. Also in that instance, no matter how many connections InterNAP has... the connection would be down... right?

Bingo.

nacjamesc

01-08-2003, 08:45 PM

Fluid hosting is by no means, absolutely not multihomed.

My brain is hurting trying to re-explain this in every post, so I kindly ask anyone who is unsure to re-read my previous posts, perhaps it will make more sense.

Thanks

tazzy

01-08-2003, 08:47 PM

Originally posted by nacjamesc
Fluid hosting is by no means, absolutely not multihomed.

My brain is hurting trying to re-explain this in every post, so I kindly ask anyone who is unsure to re-read my previous posts, perhaps it will make more sense.

Thanks

Hello,

You can "re-explain" it however much you like. May be, one of these next times coming up, you could perhaps/possibly "re-explain" it correctly?

FHDave

01-08-2003, 08:52 PM

I think I just rename Internap's routers to our name, without changing anything else, then I will be multi homed?

This is getting funny .... I believe I am multi-homed since I don't see any single point of failure (note: I do not get any bandwidth from Internap, I am directly connected to Interna's routers which is directly connected to the nine providers) within our network (again, up to our core switch).

Sprynex

01-08-2003, 08:53 PM

Originally posted by tazzy

Hello,

Yes. Also in that instance, no matter how many connections InterNAP has... the connection would be down... right?

Bingo.

I think you are misunderstanding something

Being multihomed does not mean you will always be up.

If you were TRULY multihomed, like internap.. and you had a routing issue that affected your entire netblock.. you would be done.

Just because you go with internap does not mean you are multihomed.. the point is this

if you had a connect to internap and nac

let's make this pretty.

[YOUR NETWORK]
| |
[Internap] [NAC]

And internap has a routing issue.

YOU ARE STILL UP! thanks nac! .. and vice versa.. that is true multihomed.

If internap goes down, 100%, then you are gone.. that is single homed.

FHDave

01-08-2003, 08:54 PM

[YOUR NETWORK]
| |
[Internap] [NAC]

So who will manage the BGP between Internap and NAC? I assume I will. So what happen if the BGP table on my end got corrupted? All network will down. Am I single homed then?

/me get more confused ;)

tazzy

01-08-2003, 08:57 PM

Originally posted by Sprynex

I think you are misunderstanding something

Being multihomed does not mean you will always be up.

If you were TRULY multihomed, like internap.. and you had a routing issue that affected your entire netblock.. you would be done.

Just because you go with internap does not mean you are multihomed.. the point is this

if you had a connect to internap and nac

let's make this pretty.

[YOUR NETWORK]
| |
[Internap] [NAC]

And internap has a routing issue.

YOU ARE STILL UP! thanks nac! .. and vice versa.. that is true multihomed.

If internap goes down, 100%, then you are gone.. that is single homed.

Hello,

Since when has BGP worked like that?

RackSpace is "multihomed" yet, they had a problem with their AT&T line (AT&T backbone problems) yet.... oh god.. they were down....

Multihomed doesn't mean redundent by any means. Multihomed depending how you set it up to be "multihomed" doesn't mean your data will be re-routed when the one line is down.

When your routers a being ddos', i'm afraid no matter how many connections you do have to how many backbones.... when your routers are down..... they ... can't... well "route" data.

nacjamesc

01-08-2003, 08:59 PM

Alright. My last try.

Say you are a customer of JOEISP. You have two ethernet connects to them. No matter how many routers they have, if there is a routing problem which causes your servers to become unreachable to the internet. Right? are you with me so far. You are down at this point.

The next day you call up DICKISP and get two ethernets to them, in addition to your two ethernets to JOEISP. You set up everything properly with BGP so both providers will route your traffic to you/from you. Still with me? Good.

Later that day JOEISP has another routing problem, but instead of being completely unreachable, your traffic will flow exclusively through DICKISP until JOEISP resolves their problems. Almost there, hang in there.

What have we learned? that if you have connectivity to more than one provider, your servers will remain to get hits and function properly should one incur problems. Did you notice that it didnt matter that you had two connections to both providers?

See, JOEISP having routing issues took you down hard, but the introduction of DICKISP into the scene increased your chances of remaining up exponentially.

How was that?

Sprynex

01-08-2003, 09:01 PM

Originally posted by FHDave

So who will manage the BGP between Internap and NAC? I assume I will. So what happen if the BGP table on my end got corrupted? All network will down. Am I single homed then?

/me get more confused ;)

Don't get more confused, you have it exactly right Dave
If your bgp table gets messed you are down.. that is 100% correct.. but atleast its on YOUR side and you can fix it :)..

there is no always up solution, the goal is to put as many aspects of your network redundancy in YOUR control, not others.

FHDave

01-08-2003, 09:02 PM

Originally posted by Sprynex
the goal is to put as many aspects of your network redundancy in YOUR control, not others

That's a very fair statement :) Thank you ....

tazzy

01-08-2003, 09:05 PM

Originally posted by nacjamesc
Alright. My last try.

Say you are a customer of JOEISP. You have two ethernet connects to them. No matter how many routers they have, if there is a routing problem which causes your servers to become unreachable to the internet. Right? are you with me so far. You are down at this point.

The next day you call up DICKISP and get two ethernets to them, in addition to your two ethernets to JOEISP. You set up everything properly with BGP so both providers will route your traffic to you/from you. Still with me? Good.

Later that day JOEISP has another routing problem, but instead of being completely unreachable, your traffic will flow exclusively through DICKISP until JOEISP resolves their problems. Almost there, hang in there.

What have we learned? that if you have connectivity to more than one provider, your servers will remain to get hits and function properly should one incur problems. Did you notice that it didnt matter that you had two connections to both providers?

See, JOEISP having routing issues took you down hard, but the introduction of DICKISP into the scene increased your chances of remaining up exponentially.

How was that?

Hello,

That isn't "multihomed" :eek:. That could be primary/secondry and you could refresh your BGP preference every 5 mins or whatever.... but.... what happends if the connection below the connection to the router is being ddos'd ? Where is that "multihome" then?

InterNAP just manage routers as you would with 9 providers, they then provide you with a port and run fibers to your switches/routers. Understand "multihomed". Okay, re-read.

Sprynex

01-08-2003, 09:07 PM

Originally posted by tazzy

Hello,

Since when has BGP worked like that?

RackSpace is "multihomed" yet, they had a problem with their AT&T line (AT&T backbone problems) yet.... oh god.. they were down....

Multihomed doesn't mean redundent by any means. Multihomed depending how you set it up to be "multihomed" doesn't mean your data will be re-routed when the one line is down.

When your routers a being ddos', i'm afraid no matter how many connections you do have to how many backbones.... when your routers are down..... they ... can't... well "route" data.

Hey,

Yes, if internap had a problem with 1 provider, they would be fine.. ofcourse.. because they are multihomed :).. but I was going under the fact that internap themselves had a routing issue.. IE they had a route entry that said something that was incorrect, bringing everything down.

Single Point of Failure.. for you.

in that situation, your other provider would be just fine. (if you had one).

IF being multihomed meant that you were connected to somebody who was multihomed, EVERYONE IN THE WORLD WOULD BE MULTIHOMED :).

I hope this helps..

tazzy

01-08-2003, 09:12 PM

Originally posted by Sprynex

Hey,

Yes, if internap had a problem with 1 provider, they would be fine.. ofcourse.. because they are multihomed :).. but I was going under the fact that internap themselves had a routing issue.. IE they had a route entry that said something that was incorrect, bringing everything down.

Single Point of Failure.. for you.

in that situation, your other provider would be just fine. (if you had one).

IF being multihomed meant that you were connected to somebody who was multihomed, EVERYONE IN THE WORLD WOULD BE MULTIHOMED :).

I hope this helps..

Hello,

Yes. I do understand that the terms "multihomed" are used a lot more loosely here. You could call having two internet connections "multihomed" with the right set-up. However, that is all good and well if your internal network is the same "multihomed" and redundent.

I wouldn't say that if the data being sent out via A to D, could only go via B to get to D and not use C aswell.

Maybe i'm using the term "multihomed" too much. :blush:

Good technical conversation which I understand come by very rarely..... so.. I took this one with both "hands" .... :eek: :blush:

Sprynex

01-08-2003, 09:16 PM

Hello,

Yeah, 2 internet connections to 2 different providers is multihomed, pending you actually have bgp and not just 2 different class c's (haha, dont laugh I bet somebody does this).

I was talking some stuff over with james.. and if we want to get uber technical, I am not multihomed either.

I am connected to my datacenter via 100 MB ethernet to my switch.

They are connected to: Verio, Williams, Time Warner, and FNSI.

I'm multihomed as far as it goes.. I have connections to 4 providers..

But TECHNICALLY, I am relying on my datacenter's routers to bring me traffic.. and if they have a routing issue, I have 1 point of failure.

so short of running your own routers, with your own bgp, and your own connectivity, you are not *TRULY* multihomed.

It just depends on how deep you want to go with it :)..

tazzy

01-08-2003, 09:18 PM

Hello,

Yes I was going FAR too deap it seems..... I do believe InterNAP is just as good as being multihomed though (running BGP4 , if not better in some situations).

Atleast this thread didn't get into a flame war. :D

Sprynex

01-08-2003, 09:22 PM

I try not to get involved in flame wars anymore :).. Not on a board where I conduct business anyway...

It seems alot of people around here really love internap.. I'm not sure why honestly.. and I am not going to knock or, and I don't really care to debate it since I know alot of people really really like it.. but I wouldn't even want it on my network... I'd rather go with a mix of providers directly.

Just from my experiences though :).. Most providers like it I believe because its not horrible quality like cogent, and it's affordable.

tazzy

01-08-2003, 09:24 PM

Hello,

I think most people like InterNAP due to the connectivity they offer and the way their system works.

haesu

02-05-2003, 02:37 AM

Originally posted by FHDave
There will be no cogent on Internap, so why bother compares this point. Besides, Internap routing is much more than just simple BGP4. Its intelligient routing goes beyond BGP4. I guess you are clueless about this too

Internap's intelligent routing? It's simply BGP4 with 24/7 NOC monitoring and automatic rerouting based on prefix community-set strings and as-path prepends. The bottomline is: The protocol is BGP4 period.

NAC or whomever else competent enough can duplicate the same 'intelligent routing' process as the internap. Why? Again: The protocol is BGP4 here, not some The Intelligent Routing Protocol v1 or whatever names people make up during their whooping sales conference.

It doesn't matter whatever proprietary Internap uses (if any) to route their network. They peer via BGP4, regardless of what they use for their IGP protocols. as path prepends and community strings are the key in manipulating BGP routes all day and all night long.

I don't mean to bash InterNAP at all. I've heard great things about them.. But looks like your statement here does not address the full fact

-hc

FHDave

02-05-2003, 02:42 AM

Originally posted by haesu
NAC or whomever else competent enough can duplicate the same 'intelligent routing' process as the internap.

Good. Let them do that.

:rolleyes:

haesu

02-05-2003, 02:59 AM

Hello,

Since when has BGP worked like that?

RackSpace is "multihomed" yet, they had a problem with their AT&T line (AT&T backbone problems) yet.... oh god.. they were down....

Multihomed doesn't mean redundent by any means. Multihomed depending how you set it up to be "multihomed" doesn't mean your data will be re-routed when the one line is down.

Since when BGP worked like that? You are not understanding the concept of way BGP works.

So let's see.. AT&T backbone starts taking major **** and oh god no, does rackspace go down for long? NO. If Rackspace went down on you b/c AT&T backbone went down, then you are either a) AT&T customer, or b) AT&T connection didn't actually go down.. it was having intermittent route-flapping problems. In case of b) perhaps Rackspace forgot to enable prefix dampening on their router and have AT&T do it on theirs too. Route flapping is worst case scenario in BGP.

But in any event, let us do say that AT&T backbone totally went down for hours (which I don't think that really happened.. did it?) Will Rackspace still be unavailable from the internet? uhhh.. No.

Why? When AT&T backbone takes ****, from all other paths on the internet, every AS paths traversing over AS7018 is going to be withdrawn. Realize that BGP's best path algorithm is based out of the AS PATH. If a middle piece of the complete AS PATH is down (in this case, AT&T as you say.. before getting into rackspace's AS), then the path is invalid.

What if AT&T starts being intermittent where BGP routes are flapping? By then you should be using prefix dampening to isolate intermittent prefixes.

When your routers a being ddos', i'm afraid no matter how many connections you do have to how many backbones.... when your routers are down..... they ... can't... well "route" data.

I am pretty sure Rackspace has MORE THAN ONE border router that peer with the various backbones. Or do they not? If not, then their 99.99999% or 100% whatever uptime guarantee is clearly bulls***.

-hc

haesu

02-05-2003, 03:08 AM

Hey,

Yes, if internap had a problem with 1 provider, they would be fine.. ofcourse.. because they are multihomed .. but I was going under the fact that internap themselves had a routing issue.. IE they had a route entry that said something that was incorrect, bringing everything down.

Single Point of Failure.. for you.

in that situation, your other provider would be just fine. (if you had one).

IF being multihomed meant that you were connected to somebody who was multihomed, EVERYONE IN THE WORLD WOULD BE MULTIHOMED .

I hope this helps..

Yes, Internap having internal routing issues within their AS would break your internet.

But, that's why we have things like OSPF that can delivery redundancy over internal IGP routes.

Regardless of such facts, high-availability people like Internap probably use MPLS and BGP together to route within their AS internally. You use MPLS to ROUTE the packets, and use BGP to GET the routes. MPLS and regular IGP protocols like OSPF are well-designed to evade local routing failures. UNLESS you put every customers on single core router, in which I am sure that's something Internap would never do.

Seeing the fact that a lot of Internap customers I talked with enjoy the availability and uptime in general sense, I am sure Internap doubles up on almost anything when it comes to routing equipments.

-hc

nacjamesc

02-05-2003, 11:11 AM

I agree with you, but...Thank you Mr. Buzzword. :)

Originally posted by haesu
Yes, Internap having internal routing issues within their AS would break your internet.

But, that's why we have things like OSPF that can delivery redundancy over internal IGP routes.

Regardless of such facts, high-availability people like Internap probably use MPLS and BGP together to route within their AS internally. You use MPLS to ROUTE the packets, and use BGP to GET the routes. MPLS and regular IGP protocols like OSPF are well-designed to evade local routing failures. UNLESS you put every customers on single core router, in which I am sure that's something Internap would never do.

Seeing the fact that a lot of Internap customers I talked with enjoy the availability and uptime in general sense, I am sure Internap doubles up on almost anything when it comes to routing equipments.

-hc

haesu

02-05-2003, 11:19 AM

Originally posted by Deb
quote:Multihomed describes a computer host that has multiple IP addresses to connected networks. A multihomed host is physically connected to multiple data links that can be on the same or different networks. For example, a computer with a Windows NT 4.0 Server and multiple IP addresses can be referred to as "multihomed" and may serve as an IP router.
Using the Stream Control Transmission Protocol (SCTP), multihoming allows a single SCTP endpoint to support multiple IP addresses, which means that a session is more likely to survive a network failure. In a single-homed session, a network failure can isolate the end system or make transport temporarily unavailable. Multihoming means that redundant local area networks (LANs) can be used to support local access. Various approaches, such as using addresses with different prefixes to force routing through different carriers, or even using redundant core networks, can be taken to reduce the effects of failures.

Multihoming is commonly used in Web management for load balancing, redundancy, and disaster recovery. Did that make it worse?

You do not understand the meaning of multihome at the network layer. Your definition of multihome has nothing to do with what others are speaking of.

Multihome means a provider (or rather a "network") has full-time connectivity to at least two different upstreams in the world of internet.

So.. go figure..

-hc

haesu

02-05-2003, 11:27 AM

[Mr. "OtherProvider"'s] understanding of the Internap model is incorrect and
misrepresented. Internap does NOT peer with our providers for all the above
reasons. Rather, we PAY for upstream connectivity to 8-9 of the top Tier 1
providers. As a paying customer, Internap gets highest priority routing,
across all the major Tier 1 backbones, for on-net traffic. Thus, Internap
can reach the overwhelming majority of the global IP destinations on-net.
In addition, our Assimilator technology, unlike BGP, dynamically measures
latency, packet loss, capacity, consistency, and other relevant performance
metrics, across all the major Tier 1 backbones, to determine the most
consistent, best performing path from source to destination. Internap's
Video and Voice over IP customers will be the first to tell you how the
Assimilator technology enables their applications.

Wow boy.. That's one heck of a load of sales crap I've ever heard from professional people like InterNap.

So, let's get this straight. That sales guy from internap claimes their "Assimilator" technology is better than BGP? Let's see...

It appears their ASSIMILATOR or whatever technology runs OVER BGP and manipulates as-path prepends via route-maps to isolate sluggish transit provider ;-) So their technology runs over BGP and they are claiming it is better than BGP. *lol*

BGP routing table entry for 66.150.192.0/20
Paths: (26 available, best #7, table Default-IP-Routing-Table)
Not advertised to any peer
16150 8434 3549 14742 14742 14742 14742 14742
217.75.96.60 from 217.75.96.60 (217.75.96.60)
Origin IGP, localpref 100, valid, external
Community: 3549:350 3549:4719 3549:30840 16150:65305 16150:65317 16150:65321
Last update: Wed Feb 5 15:19:04 2003

3549 14742 14742 14742 14742 14742
208.51.113.254 from 208.51.113.254 (207.136.163.27)
Origin IGP, localpref 100, valid, external
Community: 3549:350 3549:4719 3549:30840
Last update: Tue Feb 4 17:52:57 2003

8121 6461 2914 14742
199.74.221.1 from 199.74.221.1 (199.74.221.1)
Origin IGP, localpref 100, valid, external
Community: 6461:5997
Last update: Wed Feb 5 05:52:17 2003

---
bash-2.05# whois -h whois.arin.net 14742

OrgName: Internap Network Services
OrgID: PNAP

ASNumber: 14742 - 14745
ASName: INTERNAP-BLOCK-4
ASHandle: AS14742
Comment:
RegDate: 2000-02-01
Updated: 2000-02-01

TechHandle: INO3-ARIN
TechName: Network Operations Center, InterNap
TechPhone: +1-206-256-9500
TechEmail: noc@internap.com

OrgTechHandle: INO3-ARIN
OrgTechName: Network Operations Center, InterNap
OrgTechPhone: +1-206-256-9500
OrgTechEmail: noc@internap.com

OrgAbuseHandle: IAC3-ARIN
OrgAbuseName: Internap Abuse Contact
OrgAbusePhone: +1-206-256-9500
OrgAbuseEmail: abuse@internap.com

RyanK

02-05-2003, 11:36 AM

nacjamesc is correct no provider is multihomed unless the provider is connected to multiple ISPs. A quick answer is do you have an AS number? If not or don't know what that is then you are not multihomed. Your provider may be but you are not.

We are just got our new AS number because of the above. We use two different providers but each provider peers with about a half-dozen other providers (both interlinked). Thus if a connection goes down we can route around that problem, it could still be a problem but at least we can put load on the other provider.

7out

02-05-2003, 10:11 PM

I believe the discussion here on hosts being technically multihomed is useless. Some of Internap's competitors on the boards that are so quick to point out that you are not truly multihomed unless you have your own AS number and handle your own routing to multiple backbone providers are missing the point in my opinion.

FHDave's network is pure Internap. If I put a server on his network it will have N+1 redundancy everywhere on the network except for his core switch and my machine. I don't know anything about FHDave's skills, but I can guarantee you that if I had a choice of Fluidhosting maintaining the network or Internap, I would choose Internap.

Internap has the strongest SLA in the business. Why do I want to hire 10 network engineers to set up a nine provider BGP network that is monitored 24/7/365 and rerouted with prepends and all these things that according to some people on the boards can be done by anyone when Internap has proven that they can do it efficiently and they gaurantee it. Not to mention I then have to juggle contracts and installs from nine different providers.

I can guarantee you that Internap can run a more efficient network than I can. Why should I subject my customers to a lesser network just so that I can technically be multihomed. I would much rather plug directly in to an N+1 redundant multihomed connection that guarantees 100% uptime then attempt to reinvent the wheel.

Any company that is reading this that believes that their network performance is as good or better than Internap's, feel free to offer your customers a 100% uptime SLA. Don't forget though that Internap also guarantees latency, uptime, and packet loss across the entire internet.

As far as the "if Internap goes down, you are down argument" I can deal with that. If Internap has a problem with all nine of it's providers at one time, or all of it's redundant equipment goes down at the same time so be it. I'm sure that if I go with pure Internap I will be explaining network downtime a lot less than if I maintain my own network. Until I can afford to build out a network that can outperform what Internap is already doing, I think Internap's a pretty good choice.