On Wed, 7 Apr 1999, Neil Levine wrote:
On Wed, Apr 07, 1999 at 05:21:23PM +0100, James R Grinter said:
indeed. I'd not heard of this product until I saw the bakeoff PR - but clearly their product must be worth at least a look.
Are you referring to the backoff PR or Infolibria's PR resulting from the backoff?
I've found the PR generated by two of the cache vendors to be highly misleading. Infolibria mentioned a peak of 1680 requests/s to me (I didn't tell them that I'd been at the WCW where the results were released and explained) and failed to point out that the hit rate had dropped to about 14% at that point. Closer inspection of the graphs showed that the hit started dropping off at just 400 requests/s.
If the benchmark showed any product off particularly well, it was the Novell/Dell-L system coping fairly easily with 1500 requests/s. The results also look better now that the overwriting bug has been fixed. The Novell developers, who also presented several strong papers at the caching workshop, also came out of it all very well. The only problem is that I gave up running Netware 5 years ago and didn't really want to go back.
We just plugged in a test box this morning. Our Netcache reboots almost every other day now due to memory leaks in the latest code they have given us. If the InfoLibria does what they say it does then...
Guess I must be lucky. Most common uptime is 3 weeks (when we last installed new software) and the boxes run at 80%+ load for 5-12 hours/day.
Steve Clarke Admitting an interest - NetApp unlike Infolibria did buy me a couple of drinks