determining bottle necks in a linux server

Question

I have a quad core cpu with 4gb ram. I am running a VMware image of Ubuntu which I use as a web server for development/testing. I assigned this image 1gb ram and 2 cpus.

While it hosted my web portfolio just fine, I installed a number of CMSs (joomla, drupal, wordpress) which take up to 30s for it to fully load.

What options are there in determining why it is running so slow?

I can post the .vmx file or detailed computer specs if needed.

Answer 1

There are many reasons for this. Your first place to look would be anything I/O (disk) related as this is the slowest component of any system. If you run top when accessing one of the sites, and you observe that the value for %wa is consistently high (50%+ for a 2 CPU system), it's likely you're saturating your disks. You can run vmstat 1 as well which will give you a better breakdown of what your system is doing. The wa value (usually the second last column under the cpu heading) is the same as the one in top. You could be out of RAM and therefore swapping, or your database(s) could be hitting the disks too often. In this case, I recommend increasing the RAM on the VM.

Second place to look is CPU contention. If yous CPU is constantly high between the user and system values (us and sy values in vmstat), you likely need more CPUs on the system.

I/O and CPU contention are typically the causes of poor performance. As you can tell from this answer, there's no simple solution. Between top, vmstat and even iostat, you should be able to pinpoint what your system is doing.