Hi, I've found that caching does both improve performance and reduce bandwidth usage. Tuning a cache is a not a 20 minute job, it takes some time to setup right, and after reading several howtos/suggestions/tuning guides all over the show I'm pretty happy with how our caches perform now. Generally we only sit around 14%-16% byte hits (without IMS hacking). IMHO, IMS-hacking isn't worth the extra 6% byte hits it gives, compared to the headaches it causes for customers/helpdesk. Its more so the request hits that make the web feel more responsive, and I find it is quite noticeable, especially for europe and asia sourced web content. I would suggest that the 7%-8% of overall bandwidth the caches save (assuming 50% of traffic is web, which might be a bit generous) is worth it, especially if you are buying >10Mbit.. Even at 5% total savings, you are saving reasonable sums of money. I would suggest hand-rolling your own squid box as apposed to using a cacheflow, and take some time reading up how to get the most out of it (whether it be speed, savings or both). Regards, Relihan. David Fox wrote:
I agree wholeheartedly with Steve comments below, the cost saving is negligible while the required resource to administer was high. I redeployed my cache to a new position in our network: it is working very well as a footrest under my desk.
I found removing the cache actually improved overall performance, my customers - and even managers - noticed the improvement ... some of them even rang to say Thanks (that makes a change!)
David
-----Original Message----- From: Steve [mailto:steve(a)focb.co.nz] Sent: Wednesday, 9 June 2004 12:33 PM To: Cameron Kerr Cc: NZ Network Operators Group Subject: Re: [nznog] Google and MaxNet
On Wed, 9 Jun 2004, Cameron Kerr wrote:
Is this a trick question? The answer, naturally, is money, more specifically, it is far cheaper (essentially free) to serve something from an ISPs local cache than to request it all over again. This way, the ISP `earns' (saves) money, plus it has the very real potential to be faster when hit from cache (though I doubt that is the primary motivation for using transparent proxying in an ISP environment.
Many times however, when you check the bandwidth in vs. the bandwidth out the saving is minimal if you are only doing light cacheing.
While it is possible to override "dont cache" headers and expiry times on web sites this will more than likely break dynamic content and requires a lot more work to maintain lists of dynamic pages that break when the cache headers are overridden hence driving up the costs such that the "savings" in bandwidth are artificial.
-- Steve.
_______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog
_______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog