Uptime Average

Just wonder, what is your server average uptime ? Is 3.04 thru 4.5 a good value ?

How exactly do we measure this ? I’ve done man uptime and still quite confused about it. :stuck_out_tongue:

$97 off using [color=#CC0000]LuckyYou[/color] promo code :cool: . Sign Up NOW

I assume you are talking about the load average, not the uptime as your post might suggest (3.04 days to 4.5 days average would be kinda crappy :wink:

loadaverages generally come in a 3-column format. They give you the load average for 1,5, and 15 minutes. Load average is defined as the number of processes ready to consume a timeslice on the CPU at the time, plus the number of processes blocking on uninterruptible I/O.

This has several implications; one of which is that load average itself is not a perfect metric; the CPU may be idle and the load may be in the hundreds if a NFS volume dies or a disk is thrashing like there is no tomorrow. Conversely, the CPU may be pegged at 100% usage, but almost no I/O happening. (to figure out which is the case, vmstat and iostat can be used).
It also means that a “good” value can well depend on the number of CPUs or cores in the computer; on a fully loaded dual xeon with hyperthreading (=4 cores), a load of below 4 would indicate that the machine is idle at least some of the time, even with CPU-heavy processes. Most dreamhost machines have two cores. A CPU-heavy load average of 2 would indicate that the CPU is completely busy doing something (and anything below 2 is an underutilized machine). 4, if CPU-heavy, would mean that your server is rather busy, but it should still be responsive. If it’s an I/O-heavy 4, the server might not be busy at all, just some people doing heavy I/O (which might not even be on your NFS server, so you’d be completely unaffected by that in any case). Once your load climbs to 6-10, I’d go ahead and contact dreamhost to have them investigate it, especially if you experience a sluggish server.

Load average alone is just a very rough indicator. If it’s really high you know SOMEthing is wrong, but you can’t really say what. Executing vmstat 1 (or vmstat 10 for better averages) will yield a bunch more information; happy manpage-reading :wink:

It’s usually called “load average” and yes, those are good numbers. Another number that you should pay attention to is the cpu utilization. I use the “top” command for short periods of time.

Free unique IP and $67 off with promo code [color=#CC0000]FLENSFREEIP67[/color] or use [color=#CC0000]FLENS97[/color] for $97 off. Click here for more options

Wow… great explanation but add two confusing commands, vmstat and iostat :slight_smile:

Just kidding ! Thanks for your info, will check on the manpages. Seems like that vmstat makes a clearer info for me.

$97 off using [color=#CC0000]LuckyYou[/color] promo code :cool: . Sign Up NOW

I’m looking at the server pages requested, and simply divided the 500s status page responses by the total responses. On a static-only site I have, it was .001, meaning 0.999 pages came back with a response per request (while the web server itself was up, natch). On another site, that is database-driven, it was 0.958, which seems incredibly poor.

Of course, if the web server itself was down, it wouldn’t even log the request, and it doesn’t even make it into the count.

Are there sites out there that track uptime for various web hosts?