Do I understand it right that their network could be down about 4.45 hours a day before the SLA becomes valid?

http://serveroutsource.net/SLA/index.htm

Looks that way :eek: :confused:

But they have a maintence window every day...

Bit sucky if they did use that every day.

Indeed. Any host who says that they are allowed 4 hours per day downtime and THEN make a 99.9% uptime guarantee is a bit strange :eek:

Most bigger hosts around only offer 99.9% uptime in a garentee because they know they can't garentee anything more.

It what there actual uptime is as the majority of hosts now adays don't hold up to any uptime garentee, besides what exactly does an SLA garentee all its saying is they may refund your money if you noticed the downtime and send in a ticket about it.

What is the standard uptime guarantee nowadays anyway? Is it 99.9%. I was just cooresponding with a host who didn't display an uptime and when inquired stated that there was a 95% uptime guarantee. I assume this includes maintenance and everything, but that would be around 438 hrs of possible downtime or 10 40 hr weeks?

Like Yipes guarantee 99.9999% uptime but they do lot of maintenance.

ISP we colo with is 95/96% i think. When I get home I will look on the contract to confirm.

That's low tbh, 99.9 is the normal uptime guarantee nowadays.

Originally posted by Paul_9cy
Most bigger hosts around only offer 99.9% uptime in a garentee because they know they can't garentee anything more.



The amount of uptime that is possible depends upon the redundancy built into your network and the quality of the providers. We have gone 25+ consecutivie months w/ no downtime. I know others have gone a year or more. The problem is this level of redundancy is expensive and a lot of people aren't willing to pay for it.

I am not saying we will never have downtime as the odds say we will some day, but we do all we can to prevent it. And, if it does happen, we are prepared to offer a credit to customers.

Excellent uptime is possible, just depends upon what you are willing to spend to achieve it. 100% forever, who knows. We are banking revenue on the fact our engineers know enough to keep it up and running without hickups. Others are doing the same, and some are just guaranteeing 100% while hoping noone notices if they do go down.

The page doesn't say: is that 99.9% a **day**, or a **month**? The "allowable" downtime would be less than one hour in a 30-day period, not in a 24-hour period. Unless I'm really seriously missing something.

24 hours X 30 days = 720 hours

99.9% of 720 hours is 719:28.

Even 99.5% uptime is 716+ hours of uptime in a 30-day period.

most SLA's go by the month, and only then from the time the ticket is acknowledged. So lets say there has been 5 hours total of downtime, but your SLA says only 3 hours per month. If your ticket wasnt acknowledged until two hours into the downtime, you get zilch. :(

thats most places. Thankfully I've never had to deal with SLA's because hurricane has had 100% uptime since I've been with them.

Originally posted by stephenM
That's low tbh, 99.9 is the normal uptime guarantee nowadays.

No, its a realistic figure.

gee I must be the only host that can not compete in the entire market place. I can only offer 99.73 SLA

Just had a looK at the contract with the ISP, its 99.9% over a year.

just foir a note there are different uptime SLA's.

1) network
2) server
3) http
4) pingable

I can offer 99.93 network
I can offer 99.63 server
I offer 99.73 http
i can offer 99.93 pingable

the only reason I can not go any higher is that I can not do a full restore of a windows server with 40 gigs of different data in less than 1 hour or so.

But ask yourself this? Which is the more responsible to the client, HTTP nothing else, if the page don't load, the client can not generate the web page revenue ( in whichever way they measure revenue )

people play the game but spend some time looking at the contract. I took a moment to look at a carriers contract recently that stated 99.999 network uptime to the router. OK that sounds good, 30 seconds in any given month of down time average, if a network can not perform at those levels then they are not really doing thier jobs. interestingly enough that they could not offer a performance guarantee of any sort ( what the use of chicago to new york @ 180ms ) nor a basic network packet drop guarantee.

I now sit looking at the cogent contract. this is real interesting. for those that have it, look at the burstable levels, packet drop rate and calculate the max burstable level.

well anyway, different strokes for different folks

I don't understand why it is so hard to offer a good uptime guarantee? If the server isn't online, you should be more than happy to refund some dough. Afterall, your customers pay you to keep their site online right?

I can offer 99.63 server
I offer 99.73 http

How can you offer a higher http then server? If your server is down .37%, then your http server would be too. Did you flip these numbers by mistake?

After contacting that host regarding the 95%, I found out it was an error and should be 99.5%. This is looking a lot better. As far as 99.9%-100% uptime, personally, I'd prefer some reboot time in there to flush out all of the crud that always tends to accumulate as well as giving the host some time to actually upgrade some software. I'd rather have the new stuff, a fresh start once in a while, and a little lower uptime than 100% uptime rather than the system running slow or old software because it never gets rebooted. For me, 99.5%-99.7% should be fine. This gives the host time to do what they need to do to make sure the servers run efficiently.

Originally posted by GlideTech
I don't understand why it is so hard to offer a good uptime guarantee? If the server isn't online, you should be more than happy to refund some dough. Afterall, your customers pay you to keep their site online right?

Thats because when people want credits for downtime, they're totally rediculous amounts that they expect. I've had the good 'ol "my site had 30 minutes of downtime when (insert problem here, occasionally insert non-existant problem [eg. users dialup was not connected, or traceroute from singapore clearly indicates route drops before it even gets overseas, let alone to our network]) occurred, i want a full 6 months of service credits!".

We had one month back in the summer with our uplink that was pure hell as they were upgrading the routes to BGP4 (mind you we never saw downtime again after that month), but there must have been a combined 5 hours of downtime over the entire month, and we had 1 user on a relatively expensive dedicated (nice scsi setup), and they wanted 6 months of compensation for it.

We offer 99.5% network uptime, and frankly, we rarely have to use it, but if something does happen, its that little bit of flex, because things do break, people make mistakes, and techs do trip over cables sooner or later [with feet or fingers, you decide], a lot of providers offering 99.9% uptime guarantees have total garbage after the guarantee, people see 99.9% and go "oh yay" but forget to read the terms that often say stuff like "if down for more then .1% of a month, we'll give you 2x the credit for the amount of time down" (which if your service is down for say 5% of the month [1.5 days], they'll only give you a 3 day refund).

If you want a shining example, go look at Cogent's terms of service/AUP, they've modified it multiple times since i first read it (originally it stated you'd get like 5ms across their network average guaranteed, now its like 75ms), but they guarantee 99.99% uptime last i checked, and if you read the terms, the refund amount you're entitied to is total garbage, not even enough to cover the phone bill of collecting it in most cases.

Another good one is rackspace (who guarantee 99.999% uptime), if you go down for more then 24 seconds, and you have it logged in their ticket system, your compensation starts, for the first 24 seconds you get 5% refund, then 5% for each additional 1/2 of downtime, meaning with rackspace you need 10 hours or more of consecutive downtime to get a monthly fee refund. Dont get me wrong, thats a decent SLA, and i've never heard of rackspace ever being down before, but you gotta admit, for paying the amount people do for rackspace, personally, if i was shelling out $800/mo or so for a mid level server, if i had say 3 hours of downtime, with the rate being charged, i'd expect more then $280 back.... thats just me though :).

How many places are offering 99.xx% uptime guaruntee for services (http, smtp, ftp, etc)?

I may be mistaken, and apologize if I am, but most DC's offer a network uptime guarantee, not a server uptime guarantee. With the 99.9% uptime, the approximate downtime would be 42 minutes per month, and most go on a month-month basis, offering 5% monthly fees for each additional (instert time here) that they are down. Excluded in that norm is scheduled maintenance with notification of 24 hours or more. Problem with a server uptime guarantee, is and would be IMHO, properly diagnosing whom was at fault. For example, if a client logged in, shut down http, didn't restart, and services/SLA's cover server downtime, then the provider would be responsible for EU error. To compensate for server failure on the part of the provider, most offer a separate SLA with hardware replacement guarantees, and response times. The only providers I can see who would offer uptime guarantees for services, are those providers offering managed services/solutions that the client pays for. Dedicated or unmanaged servers ultimately fall into the responsibility of the customer of the provider to monitor and repair/contact provider to assist with getting back online. Again, I could be wrong, that's just always how I've perceived these types of things.

Originally posted by UmBillyCord


>I can offer 99.63 server
>I offer 99.73 http

How can you offer a higher http then server? If your server is down .37%, then your http server would be too. Did you flip these numbers by mistake?

you would think so but nope, weird but this is how we break it down:
platform is h-sphere
control panel is one server, e-mail has it's own server, http/ftp/website has it's own server and databases have thier own servers.

so there is a 1 in 3 chance of a failure that will take out the http. so even if the control panel/email crap out I still have http working.

most people can live with e-mail failures that last 30 seconds to 30 minutes. most people can not live with http failure that last any sort of time.

Mike

Originally posted by clockwork
How many places are offering 99.xx% uptime guartunee for services (http, smtp, ftp, etc)?

I've seen http as high as 99.9999 ( high avalibility priced in the millions ), mail ( stmp ) with different rules and reg's. all depending on what the inbound and outbound mail requirements were. and if web mail and pop ( or apop ) would be applied. Web mail means that you have an outside access and require http to work. Never seen ftp, but it would most likely be considered under the http. network uptime you pass along the average from your contract, router same thing ( since you should have failover )


Now whom offers 99.95%+ : Gateway and IBM and a few others offer http and server, but you got to pay for that 99.95%+ that they have.

Mike

Originally posted by clockwork
How many places are offering 99.xx% uptime guartunee for services (http, smtp, ftp, etc)?

We have found that unless they are using global load balancing, this will be next to impossible **unless** they have in their SLA crap about maintenance, hacks, etc... Sure these are legit, but what many small host (we saw one offering this 99.999% on a RS server!) is play the odds. They will accept a 45 minute outage. Why? Because maybe only 1 in 100 will ask for a refund. So lets say 10 request it. They can simply say "They were DOSed", "It was maintenance". Another words, this uptime guarantee is just as much BS as unlimited bandwidth.

As Mike said, the bigs boys offer it with a detailed SLA to protect both sides. But you also pay a premium for this SLA.

Oh, I think you had my questioned mistaken, I know anyone, other than properly load balanced setups, offering that sort of services guaruntee is laughable, but I was just curious as to who was offering it.