FYI performance of Squid 2.2 on Solaris 2.7 using Polygraph "datacomm-1"

From: Peter Jenny <pjenny@dont-contact.us>
Date: Wed, 22 Dec 1999 16:39:36 -0500

I've been benchmarking Squid in the lab trying to find it's maximum
performance if not limited by disk I/O. So with Squid 2.2stable5 running
under Solaris 7 with a 256 MB cache_dir mounted on a tmpfs file system in swap
space (in RAM), here's what it could do with 2 different speed Sun Ultras
(Ultra 5 or Ultra 10) equipped with a total of 1 GB RAM (Polygraph using
roughly the "datacomm-1" benchmark, just didn't run it as long i.e. lower
goal):

270 MHz, peak SPECint95 rating 9.17, 13 KB object size: 100 responses/sec
440 MHz, peak SPECint95 rating 18.1, 13 KB object size: 190 responses/sec

So, with this OS, and if disk I/O is not the limiting factor (either because you have enough spindles or you're using a RAM-based file system), the most Squid 2.2stable5 can sustain is about 10 times the machines SPECint95 rating. I set cache_mem to 64 MB for these tests.

I also tried a few different size objects with Polygraph.
270 MHz, peak SPECint95 rating 9.17, 6 KB object size: 140 responses/sec
270 MHz, peak SPECint95 rating 9.17, 20 KB object size: 95 responses/sec

                   Peter Jenny
            Internet Engineering Services
         GTE Internetworking - powered by BBN
3 Van de Graaff Drive, PO Box 3073, Burlington MA, 01803
       pjenny@bbn.com http://www.bbn.com
       Tel: 781-262-6041 Fax: 978-428-7020
Received on Wed Dec 22 1999 - 14:47:20 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:50:05 MST