server uptime

[paul0130]

I aim for 100%, and we pretty much keep it that way.

But anything in the 99.9% mark is ideal, a little downtime for a reboot and so on is not to bad.

[reedhurst1026]

anything above 99 % can be considered as good.

[wswd]

anything above 99 % can be considered as good.

So 7 hours of downtime per month is acceptable? Not with my company it ain't! How can 99% possibly be good?

[b2netsolutions]

I think 99,98% is a trustworthy uptime.

[WebDevNow]

Thanks all, Got Info

[JJW]

99.5% is about the line that I can draw in the sand. As you have to account for network issues that are out of your control, and other things that are out of your control

[wswd]

If your network is having almost 4 hours of downtime per month, it's time for a new network, to be honest.

Even if you reboot the server once a week (which is ridiculous), you're only talking about 20 minutes of downtime per month, and that's on the high end, with a crazy 5 minute reboot. That equates to a 99.95% uptime. Hell, our datacenters change entire raid arrays and swap out RAM in less time than that. Anything much beyond that and clients should absolutely be compensated for the downtime.

I honestly don't get how hosts are saying that hours and hours of downtime every month is somehow acceptable. There is obviously an acceptable amount of downtime, as no service (including cloud) can really guarantee 100% uptime. That amount of acceptable downtime, however, shouldn't be anywhere close to hours every month.

[JJW]

Well ours is above that but, we only gaurentee 99.5% uptime. Our average uptime is displayed proudly on our page, Even though right now it is not at 100% we still display it. 99.86% over something like 8 months, since we started keeping track.

[wswd]

99.86% over something like 8 months, since we started keeping track.

That's about an hour of downtime per month.

[JJW]

April was the month that put us there. As there were ddos attacks on our dc and then and issue with software.

[Seraphim]

I honestly don't get how hosts are saying that hours and hours of downtime every month is somehow acceptable. There is obviously an acceptable amount of downtime, as no service (including cloud) can really guarantee 100% uptime. That amount of acceptable downtime, however, shouldn't be anywhere close to hours every month.

That's the difference between real numbers and advertised figures. The goal is always 100% of course, but in practice 99% usually happens because things do go wrong from time to time.

So by calculation a 99.95% uptime might mean 20 minutes a month, while the 99.5% figure allows 2.5 hours a month. Okay nobody is perfect, for sanity sake you almost have to allow one good screwup a month in your advertised values and then give yourself a pat on the back if you managed to not use that allowed downtime.

It's like snow days in the schools. Every year they allow X number of snowdays just in case bad weather makes it too unsafe to get staff or students to the school. Some years like this one, they don't use any of them at all- and in many cases will waste them by adding the unused outage time to other scheduled events such as vacations. Other years they end up using more than they allowed for, and end up taking the remainder out of vacation time in order to maintain state-mandated attendance hours.

With servers, give it your best shot but anyone who doesnt' want to risk a false advertising lawsuit has to claim a realistic figure like 99.5%, which works out to one big screwup or a handful of little blips a month.

[wswd]

With servers, give it your best shot but anyone who doesnt' want to risk a false advertising lawsuit has to claim a realistic figure like 99.5%, which works out to one big screwup or a handful of little blips a month.

That's why you shouldn't advertise any uptime. It could very well be false advertising if you do. The better method is simply utilizing an SLA instead.

I guess I just don't understand how hosts can be okay with 4 or 5 hours of downtime a month...in fact I was shocked by some of the responses in this thread. I'm not really sure why any client would accept such an SLA. Imagine a brick and mortar storefront or such, that was forced to shut down 5 hours (almost a complete day) every month. Then take an online store, and that 5 hours somehow becomes acceptable?

Sure, things happen from time to time with servers. There is going to be some minor downtime. It's pretty much guaranteed. Parts need to be replaced. Routers break, etc. But 5 hours a month, every single month? What kind of SLA or overall goal is that?

[Seraphim]

Indeed. Advertising any uptime guarantee at all is a risky tactic.

But internally tracking uptime is a good idea, and in that regard if you are doing better than 99.5% you probably are in good shape and just need to keep track of what causes the outages in the first place so you can avoid them.

Also, most stores in these parts at least are only open from 9AM-5PM, for an utterly astonishing 33% uptime. And that is completely normal.

I think it all comes down to knowing how much downtime you can realistically expect, and then planning accordingly. Right now my highest measured availability is 99.96%, with the lowest at 99.94%. But even with those figures, one of those two servers had a visible outage in the past 2 weeks.