Date: Tue, 16 May 2000 10:20:49 -0700 (PDT) From: Scott Hess <scott@avantgo.com> To: Nathan Vidican <webmaster@wmptl.com> Cc: Brennan W Stehling <brennan@offwhite.net>, freebsd-questions@FreeBSD.ORG Subject: Re: Clustering FreeBSD Message-ID: <Pine.LNX.4.21.0005161013490.3824-100000@river.avantgo.com> In-Reply-To: <391B0758.5F0D758C@wmptl.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 11 May 2000, Nathan Vidican wrote: > There are many different ways you can approach this. I'll give you my > two suggestions, but I'd be curious as to what you do end up using? Another option would be to get a TCP-level load balancer. You can get very expensive ones, or you can get a Linux box to do it for you. It can be a very powerful option, because you just throw a bunch of identical machines behind it and then you can rotate machines in and out with very little impact (other than performance of the cluster, of course). > Secondly, you could use some sort of caching system (ie squid), to > accelerate your httpd server(s) through the user of cache-pools. > Although descriptively shorter, this method is much more difficult to > implement. Not at all! You should be able to implement a single Squid box in http accelerator mode in a day or so. If your content is reasonably cachable, that right there can leverage your performance - the Squid box quickly serves static content without involving the Apache box, and the Apache box is only called in for the heavy lifting. A Squid http accelerator with an Apache engine can easily exceed the performance of two Apache engines, if your content is more than marginally cachable. If your content is extremely cachable, you could even put multiple Squid boxes in front of a single Apache engine. In my experience, very few website would benefit from a Squid http accelerator using cache-pools (I assume this refers to sharing cached items with sibling Squid servers). This works very well when you're caching the Internet, but usually websites only have a couple tens of Meg of static content, so that type of scalability isn't an issue. > Anyhow, I think you under-estimate the capabilities of a one-machine > webserver; The nice thing about a TCP-level load balancer is that it increases the availability characteristics of the machines it balances to. Taking one machine out doesn't take the service down. Later, scott To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.4.21.0005161013490.3824-100000>