It seems that Squid cache is living in the past. With support for only a single core and support for SMP only loosely on the horizon, we were looking for ways to improve the performance of our Squid proxy / filtering server which we use for filtering. To overcome the limitation, users can set up multiple instances of Squid, however on a single machine this isn’t supported with the plugin that we need to use. As a result, we’d have to resort to using multiple servers. Still, this seemed like a waste of resources to dedicate many powerful boxes to the task.
Thanks to the free release of VMWare ESXi this past Monday, we decided to load ESXi and create 4 seperate VM’s for Squid. This will allow us to have 4 seperate instances with a functional filtering plugin. Additionally, we will use the old server (which has quite a bit of umph in it left) as well as the old reporting server (whose function will now be handled by a VM on our other ESXi box we purchased in the months previous) will also become Squid caches. We had all sorts of witty comments about a school of squid and the like, but we settled on the name “kraken” for the DNS round robin result. It will also give us the ability to take the server down for any necessary maintenance. As a result, I left on a high note, a smile on my face, compiling a newer Linux kernel with VMI support enabled.