Disk I/O rate?

based on graph in daskboard, mine hovering around 200-300

Average of 1.1k with a high of 3.9k

Webserver/Jabber/Teamspeak

avg in the high 60s, peak of about 1000

webserver, svn, a few other apps, nothing IO bound

max: ~780

avg: ~100

Average of 162, peak of 1441.

I'm not real sure what it should be under though. I know the default alerts are 300.

Over the last week, my max is 7k blocks/sec and my average is 71 blocks/sec.

I run a Drupal-powered website and discussion forums, and the max is when my crontabs run in the early hours of the morning. Otherwise, the regular disk load is very low.

(note that a large MySQL disk cache helps to drastically cut down the amount of disk accesses. You should ideally have enough RAM to keep frequently accessed tables entirely in memory)

– Doug

Disk IO:

Max: 1426

Avg: 21

Last: 7

(5 small php sites, 3 mysql databases, ubuntu 8.10)

max: 880

avg: 110

lighttpd + small-to-mid-sized php forum + mysql with lots of memory

Having a PHP cache, such as APC, eAccelerator, or XCache, helps cut your disk I/O by a significant margin. Especially if your blog or forum software is made up of hundreds of little PHP files. And of course, less disk I/O = faster page load.

Linode 360: 118 max, 30 avg.

How much more sites can I put in this box?

I'd say you could put purple more sites on your Linode.

Seriously though… disk i/o is only one variable (and not a terribly important one). It's impossible to tell you how many more sites you can add without knowing how much of all your other resources you're consuming, how well your system is tuned, what the other sites would be like… etc…

~JW

Average of 13.6 handling about a million pageviews per month (driven by Lighttpd/PHP/MySQL)

@JshWright: thanks, that's what I think too. My CPU is constantly idle at 99% by 'top', RAM constantly 60MB free by 'free -m' and SWAP constantly 0MB by 'free -m'. So does this mean I can put more sites into my box?

@Guspaz: Wow, I suppose it's all content pages with little logics in PHP? Didn't know Lighttpd can make that much a difference. I may use it instead of Apache but what kind of compromises it comes with in favor of performance? And how about Litespeed? Nginx?

@Guspaz:

Average of 13.6 handling about a million pageviews per month (driven by Lighttpd/PHP/MySQL)

Damn. It makes me feel bad. 3 million page views/month and I am already doing average 250 IO.

(Need to tune MySQL access with memcached… Or maybe I will just spend a bit more on a bigger plan.)

@scotty:

@Guspaz:

Average of 13.6 handling about a million pageviews per month (driven by Lighttpd/PHP/MySQL)

Damn. It makes me feel bad. 3 million page views/month and I am already doing average 250 IO.

(Need to tune MySQL access with memcached… Or maybe I will just spend a bit more on a bigger plan.)

3m monthly page views on what plan? 360?

Isn't memcached for distributed applications?

Keep in mind that turning on caching in applications will cause them to use more ram, which means that Linux does less caching on its side. So, more cache in MySQL and web apps is not always the best way to decrease I/O. I think it's a good idea to keep about 30% of your ram free so that Linux can do its own caching.

In-application caching normally caches processed results; it's usually not intended to reduce disk I/O (the OS already handles that), but to reduce CPU load or response times.

For example, if a MySQL query normally takes one second to compute a result, and you've got ten hits per second calling that same query, you might be in some pretty big trouble. But if you cache the results, suddenly you're golden. Misconfiguring MySQL's query cache can result in far higher CPU utilization and page load times than you could get.