Page 2 of 2 FirstFirst 12
Results 26 to 50 of 50
  1. #26
    something as basic as DNS is a red flag that you're not "real".
    In theory everything looks 'basic', in fact it isn't.

  2. #27
    Quote Originally Posted by vburke View Post
    I have no idea why outsourcing your DNS to someone else would be a big draw for your customers. In my experience, outsourcing something as basic as DNS is a red flag that you're not "real".
    There are huge benefits to outsourcing your DNS. Look back to the data center failure at The Planet in late May of this year. Customers had servers down for days. Some servers were up, but The Planet's DNS was down. If your DNS is outsourced, you could restore your backups to a new server and update the DNS zone files with the new IP... without going through the hassle of updating your name server IPs with the domain registry and worrying about any customers who might have "private labeled" your name servers.
    DurableDNS - Managed DNS Hosting

  3. #28
    Quote:
    I host with ServInt and haven't had any downtime in the last 18 months.

    Impossible, do you have any statistics that prove it? If not i bet there were some downtimes, but you didn't noticed.
    Actually, it is quite possible. Maybe not in that specific instance, and maybe never seen by you, but it is possible. And has been, as a personal experience, at 585 days straight. On a production server exposed to the internet.
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  4. #29
    Join Date
    Nov 2004
    Location
    Australia
    Posts
    1,737
    Plum, an uptime of 585 days is impressively spunkily awesome. But it also doesn't mean there were network outages, as they don't affect uptime at all, and do affect users! And of course an uptime of 585 days does mean kernel security patches weren't being applied!! Not to negate the 585 days of awesomeness!

    Quote Originally Posted by HYB-Bryan View Post
    There are huge benefits to outsourcing your DNS. Look back to the data center failure at The Planet in late May of this year. Customers had servers down for days. ... but The Planet's DNS was down.
    That's a great argument certainly, for doing your own DNS, but not necessarily for outsourcing if you run a shared host (if you run high-impact servers, sure, it's a different kettle of fish.)

    Back in the OP's world, I'm not sure that updating IP addresses in 50-60 zones via a potentially clumsy and slow web interface on an outsourced DNS (*) is going to be any easier than updating a single master nameserver to point to the new server, and using a command like "replace" to change the old IPs to the new IPs on the new server in one hit. Additionally, if the old IPs and TTLs are being served up by the outsourced server, unless the TTLs are small you'll have to contend with more of a caching problem when switching over. (failure is only cached for 30 mins, whereas real TTLs are usually higher, like 4 hours)

    I just love it when we get emotionally attached to our favorite solutions rather than thinking clearly about 'em. (of course, I *never* get emotionally attached to My solutions! *never*!

    (*) unless you have global search and replace via the web interface ... or have everything CNAMEd, which is ugly for customers...

    (**) Replace command is part of MySQL (perl can do similar):
    Quote Originally Posted by Manual page
    Invoke replace in one of the following ways:

    shell> replace from to [from to] ... -- file [file] ...
    shell> replace from to [from to] ... < file
    Last edited by brianoz; 09-06-2008 at 03:32 AM.

  5. #30
    Network always can be down - sometimes, even in servint.

    SERVINT: We do have a 99.9% uptime guarantee, which constitutes our network SLA.
    You can only measure uptime using an external service to check your site from several worldwide locations. Servint has very good service, but i bet there were some network downtimes during almost 2 years. 99.9% uptime quarantee / year = 0,365 day of downtime.

  6. #31
    Join Date
    Oct 2007
    Posts
    455
    Quote Originally Posted by vburke View Post
    I have no idea why outsourcing your DNS to someone else would be a big draw for your customers. In my experience, outsourcing something as basic as DNS is a red flag that you're not "real".

    Vern

    Most small companies do not have an extra $1,000,000 to invest in routers, switches, IPS / IDS, servers, UPS, remote reboots, etc.. in 10+ different facilities to do a true DNS IP anycast deployment. Believe it or not many companies that are starting out only have one VPS / server. To invest $1mil+ just to setup DNS is a lot.

    I'm thinking that is why people will choose to outsource DNS. Get it done the right way... Have the ability to change IPs even when your server is down... And never have to worry about the service.

    Worth a few cents per month per domain for sure.

  7. #32
    Plum, an uptime of 585 days is impressively spunkily awesome. But it also doesn't mean there were network outages, as they don't affect uptime at all, and do affect users! And of course an uptime of 585 days does mean kernel security patches weren't being applied!! Not to negate the 585 days of awesomeness!
    Well, three things

    1. no kernel patches required, not Linux.

    2. network outages would be known, as this was a wholly owned type of situation with complete access to everything. anyways, if a few packets were dropped, it still did not take away from the fact that old gramps chugged away for 585 days, before ...

    3. someone very trusted, and a *licensed* electrician, pulled the wrong plug while trying to wedge more stuff into the rack

    3a. this has been addressed by insisting on redundant power supplies, and zip tying all connections.

    The machine went on to do a number of 400+ day runs, but never again reached 585 before it died an honourable death. Dual Pentium Pro 200MHZ, 512MB, 4x2.0GB hardware raid. You should be able to get more than that for $0.25/month these days

    In the meantime, there are a number of BSD machines on the internet that have 900+ day uptimes in the *.jp zone. They seem to make a habit of it. The problem for BSD is that the uptimes as recorded in the network stack are capped at around 900. So, some of those machines have been up for more than that. It just can't be seen.

    Oh, little known factoid. Of the top 50 uptimes observed globally, not 1 machine runs Linux. The OS's that are in the list include mostly *BSD, then Windows, then Solaris. The cutoff to make it into the top 50 is about 800 days.
    Last edited by plumsauce; 09-06-2008 at 05:14 AM.
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  8. #33
    Quote:
    Originally Posted by vburke
    I have no idea why outsourcing your DNS to someone else would be a big draw for your customers. In my experience, outsourcing something as basic as DNS is a red flag that you're not "real".

    Vern
    To be very simplistic, installing dns is easy. Doing dns well is hard. DNS outsourcers are specialists in one thing, DNS. Tech support levels vary, but at least in theory, you are benefiting from dealing with a specialist.

    Most small companies do not have an extra $1,000,000 to invest in routers, switches, IPS / IDS, servers, UPS, remote reboots, etc.. in 10+ different facilities to do a true DNS IP anycast deployment. Believe it or not many companies that are starting out only have one VPS / server. To invest $1mil+ just to setup DNS is a lot.

    I'm thinking that is why people will choose to outsource DNS. Get it done the right way... Have the ability to change IPs even when your server is down... And never have to worry about the service.

    Worth a few cents per month per domain for sure.
    Agreed

    Except for the million dollars. If you look way back in history, most of the outsourcers, including Ultradns/Neustar, started on less. They got their VC money after building up and out.

    The hardest part of the exercise, for at least one person on the team, is to read and *understand* reams and reams of RFC's about what is probably the most important and most badly designed protocol used on the internet. Then, if you do anycast, better have someone who is also very comfortable with BGP. Finally, if you decide that the usual three pieces of software aren't suitable, then you also have to have someone who can code all of the above. Correctly.

    Ummm.., ok, maybe the million is about right
    Last edited by plumsauce; 09-06-2008 at 05:38 AM.
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  9. #34
    Join Date
    Oct 2007
    Posts
    455
    Quote Originally Posted by plumsauce View Post
    Ummm.., ok, maybe the million is about right
    Yeah.. You have to figure (even if you did it the cheapest possible).
    Router / Switch (cheapo cisco that can hadle multiple providers and do BGP) - $8k
    IDS / IPS (the cheapest TopLayer) - $20k
    Servers - $8k
    Remote Reboot - $500
    Let's skip KVM over IP for now...
    ------------------------
    $36,500 per location
    Let's just say you go small with 10 locations... That's $365,000 in just hardware.
    Each location you figure needs setup costs... shipping... installation... rack costs... power costs... bandwidth commit costs....

    Yeah.. a million isn't too far off.

  10. #35
    Yeah.. a million isn't too far off.

    Well, the infrastructure part is not insurmountable. There are a variety of ways to bootstrap that. Sort of like the old joke:

    Q. How do you become a millionaire in the stock market?

    A. Start off as a billionaire

    But, the intellectual property part, that's the kicker. It's not off the shelf stuff that you can just download from somewhere.

    And then, the hardest part ... finding customers
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  11. #36
    Join Date
    Nov 2004
    Location
    Australia
    Posts
    1,737
    Quote Originally Posted by BuffaloBill View Post
    Most small companies do not have an extra $1,000,000 to invest in routers, switches, IPS / IDS, servers, UPS, remote reboots, etc.. in 10+ different facilities to do a true DNS IP anycast deployment. Believe it or not many companies that are starting out only have one VPS / server. To invest $1mil+ just to setup DNS is a lot.
    And most companies don't need that sort of infrastructure. We started out in a one room office, not a 6-story skyscraper! For sure it's worth a few cents a month, if you have a single dedicated host, maybe, but it's not worth it if you run a shared host as tangible business benefits just aren't there.

  12. #37
    And most companies don't need that sort of infrastructure.
    Exactly, but isn't that why third party services exist?

    It's very much like webhosting. The customer rents a piece of the pie from the webhosting company, who has rented a piece of the pie from a server provider, who has rented a piece of the pie from a data center ...

    Australia, and most of the PacRim region is actually a hotbed of interest in innovative dns techniques for traffic distribution. With bandwidth being so expensive, some sites want a fast local presence, but also want to process international traffic where bandwidth is cheaper. The solution for them is to host locally for the local audience and elsewhere for international audiences.

    Indeed, some Asian multimedia companies could not achieve their latency requirements for North America until they placed servers in North America and distributed their traffic. And, some of those "servers" are in fact squid caches running in reverse proxy mode. They condidered the fact that bandwidth was cheaper to be a bonus in the whole exercise.
    Last edited by plumsauce; 09-07-2008 at 05:59 PM.
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  13. #38
    Join Date
    Jun 2008
    Posts
    1,471
    Just wanted to throw this in there for those with cPanel and want some off server/site DNS server. cPanel offers a standalone DNS server software that syncs with your main server and you could run it in a VPS or a dedicated server somewhere and have redundant DNS this way.
    http://www.cpanel.net/products/dnsonly/index.htm

  14. #39
    cPanel offers a standalone DNS server software that syncs with your main server and you could run it in a VPS or a dedicated server somewhere
    That would be a second dns server, yes. The same thing you could do with any old piece of free dns software capable of AXFR transfers.

    But, it does not address the concerns of the OP with respect to the business aspects of running DNS.
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  15. #40
    Join Date
    Oct 2005
    Posts
    517
    Thanks, everyone, for all the valuable info. My head is still spinning from trying to absorb all the technical details and especially the pros and cons of managed DNS.

    I also discussed this back and forth with the dnsmadeeasy support staff (I have an account there, but no services yet) and it didn't seem too difficult until it got to the point where I asked them how to handle the mirroring of the domains, content and the databases to achieve full dns failover etc. They said I would have to get a techie to do this for me.

    To keep a long story short, it does seem worthwhile to add a couple of vanity nameservers through dnsmadeeasy.com (other than my regular nameservers) and explore the full dns failover and load balancing options when the business grows enough to demand it. I also read somewhere that an alternative may be to use CDN to have proxied web sites. Is that the same or cheaper or am I getting confused?

    Frankly, I started exploring this after reading Seth Godin's Purple Cow book which is all about trying to be 'remarkable' to succeed in business. I wrote down a list of things that would differentiate my company from the run-of-the-mill host and enterprise DNS was one of them. This thread has given me a lot to think about.
    Last edited by ramdak5000; 09-09-2008 at 09:55 AM. Reason: typo

  16. #41
    I also discussed this back and forth with the dnsmadeeasy support staff (I have an account there, but no services yet) and it didn't seem too difficult until it got to the point where I asked them how to handle the mirroring of the domains, content and the databases to achieve full dns failover etc. They said I would have to get a techie to do this for me.
    Yes, the devil is in the details. You definitely want a vendor whose support people are familiar with, or have done the exact same thing themselves.

    I also read somewhere that an alternative may be to use CDN to have proxied web sites. Is that the same or cheaper or am I getting confused?
    It is rarely cheaper.

    Frankly, I started exploring this after reading Seth Godin's Purple Cow book which is all about trying to be 'remarkable' to succeed in business. I wrote down a list of things that would differentiate my company from the run-of-the-mill host and enterprise DNS was one of them. This thread has given me a lot to think about.
    What you could do is learn about all of the options available so that you are ready to go when a customer inquires about "can you help me with?" Customers don't mind "I don't know, but I will find out for you" as long as you actually do it.

    For example, when I receive a problem report, the process is as follows:

    response #1

    1. acknowledge problem immediately
    2. describe the proposed remedy
    3. estimate a time to fix
    4. describe any interim workarounds
    5. promise to inform customer of the status
    6. thank the customer for bringing the problem to light

    response #2

    1. inform customer that problem has been fixed
    2. describe what had to be done
    3. thank the customer again for taking the time to report a problem

    This gives the customer the two things he wants to know without having to dig: when will it be fixed, and how. Furthermore, that the vendor accepts responsibility for the problem. Most customers are satisfied with this, even if the quoted time is 3 days, as long as they feel that they are getting a truthful answer and will be kept informed.

    Sometimes, you even have to say, "sorry, but that is not what the product does, and here is why, as well as the alternatives".

    Most customers can be satisfied if you take the time to explain and educate. Even if the answer is no. Knowing why lets them feel that their concerns have been addressed specifically.

    Yes, my emails can get even longer than some of my posts

    PS. Never compete on price alone. Compete on what you deliver at a fair price. If you deliver more, charge more
    Last edited by plumsauce; 09-09-2008 at 05:19 PM. Reason: ps
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  17. #42
    Join Date
    Oct 2005
    Posts
    517
    plumsauce, thank you for all the time you spent on this thread answering questions and educating people about the pros and cons of managed DNS. Much appreciated.

  18. #43
    ramdak, your thanks are much appreciated

    The point was to highlight that the make or buy decision is not always strictly financial.

    And now, a simple analogy to wrap up.

    Take the example of getting a haircut. There are a number of choices.

    -- you can do it at home. most people can handle a little trim around the edges if absolutely necessary. if it ends up in disaster, well, you can always wait until it grows out. in the meantime, you can wear a hat.

    -- you can go to the neighbourhood barber shop. you get a pretty good haircut for a pretty good price. no hat required.

    -- you can go to an upscale hair salon. you get a slightly better haircut, and if you need some hair conditioning, want a special style, or need to cover the grey, they'll be happy to oblige. they might even offer manicures and massages. why would you hide that hair under a hat?

    -- you can fly in your personal grooming consultant from London or Belair. they will bring all the necessary tools and their gorgeous assistants to cater to your every whim. if you need to ask how much, you can't afford it. no hat available, you've just been scalped.

    There you have it, the haircut theory of business management
    edgedirector.com
    managed dns global failover and load balance (gslb)
    exactstate.com
    uptime report for webhostingtalk.com

  19. #44
    Join Date
    Jan 2008
    Location
    Malaysia
    Posts
    95
    we use managed dns for some of our clients. they are a few dns provider that can provide as low 50cent per zone (500 records) and it's reliable. since the dns is geo disperse, timeout would be avoided if the net route to the cpanel dns is congested or failed.

    we can't find way to sync the managed dns with the cpanel dns, it has to be added manually.
    HostWaves.com Managed Drupal e-Commerce Hosting/Web Development Service for SMB/SME/SMI.

  20. #45
    Join Date
    Oct 2005
    Posts
    517
    plumsauce, I had a good laugh reading your haircut analogy. It's very apt.

  21. #46
    Join Date
    Aug 2003
    Location
    /dev/null
    Posts
    2,132
    Self-experience: Outsourcing the DNS allows you to bear more easily with infrastructure changes (especially IP changes). If you do a lot of reseller hosting, then this means you don't have to tell the resellers to change the DNS IPs everytime you move out. You benefit in fact from never changing the end user's DNS servers anymore.
    For most models, if the web server goes down, at least email can keep running (most of my customers outsource email to Google Apps). If you do the "economy-mode" hosting of everything-in-a-single-server, that may not make sense, but even in this situation, the DNS server can simply go down and all other resources stay unavailable.

    Windows control panels (DotNetPanel for example) have taken a big step into integrating with the API of specialized DNS servers like Nettica (which I use) or UltraDNS.

    Since I switched to Nettica my DNS problems were turned to zero occurences. Also we could do very faster migrations, as we set the records' TTL to a low time, and no need of DNS server changes, plus there were easy ways to replace a single IP address in all the zones that we host with them.

    I have a big corporate customer who is now also switching to Nettica as soon as he implements his control panel solution. His 1500 domains are suffering from DNS server hiccups and misroutings and slow manual needs of updating them - he has no control panel involved - as well as Microsoft DNS Server is quite strange to deal with and proved to be slow and unresponsive. And he will gain also in the fact that there will be multiple redundant servers serving his DNS with multiple paths, so he's elliminating a single point-of-failure.

    Another plus is that we didn't need to worry about patching the DNS servers for these recent discovered flaws.

    Another stupid but interesting point is that these services help on the global economy of IP addresses, as there will be no more need of 2-addresses-per-server due to DNS minimum 2 nameservers requirements, what happens with 90% of CPanel webhosting companies.

    I wish CPanel could integrate with Nettica, this way we'd elliminate its point of failure (although we run cPanelDNS on a second VPS in order to share a single secondary nameserver IP address for all our cPanel VPS and dedicated customers and not needing to give them a secondary IP just for that).

  22. #47
    Join Date
    Nov 2004
    Location
    Australia
    Posts
    1,737
    Nearly all of that can be done with cPanel as it is out of the box. We only change IPs every few years and last time we did it, it was totally seamless (we tend to stick with our servers once we have them).

    Should be possible to integrate cPanel with Nettica if they have an IP, it's just a matter of doing the programming.

    I wouldn't use Microsoft DNS server for anything serious! If I had to run one standalone it would probably be easier to run djbdns with some simple scripts on top of it.

  23. #48
    Join Date
    Nov 2002
    Posts
    192
    I'm not familiar with cPanel internals but couldn't you simply use AXFR/IXFR from the cPanel server to whomever you've outsourced to?

  24. #49
    Join Date
    Aug 2003
    Location
    /dev/null
    Posts
    2,132
    Quote Originally Posted by aarong View Post
    I'm not familiar with cPanel internals but couldn't you simply use AXFR/IXFR from the cPanel server to whomever you've outsourced to?
    The idea is NOT to have to do any manual intervention :-)

  25. #50
    I like the idea of managed dns services or self-managed ones at least. dnsmadeeasy is a good one that I have never had issues with.

Page 2 of 2 FirstFirst 12

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •