Tony McDonald wrote:
Hi all, did some more Pystone benchmarks. As usual, with all benchmarks treat these as total fabrications...
PowerMac 5400 (180 MHz 603e) running linuxppc (this is my dev box at home) = 890 iMac (266 MHz G3) MacOS 8.6 = 4800 Sun (4 x 250 MHz Ultrasparc II) running Solaris 5.6 = 2200 (lots of stuff going on) (a) Sun (1 x 350 MHz Ultrasparc IIi ) running Solaris 5.7 = 2600 (lightly loaded) (b) DELL (Pentium III 450MHz) running SUSE 6 = 5800
using Apache benchmark with a static HTML page through Apache, not Zope (page may have a couple of SSIs in it); (a) concurrency = 250, number of times = 10,000 gives 457 requests/s (b) c= 250, n = 10,000 gives 141 requests/s ie these boxes aren't slow...
Q: How comes the Sun boxes are just so bad at Python? is the Sparc architecture *that* bad?
It must have something to do with some loops fitting in the cache (code tends to be a lot larger on RISC architecture) An IBM NetFinity M20 (2 x PIII Xeon 500) has the following specs ./zeus_bench localhost / -c 250 -n 10000 = 1047 req/sec pystone.py = 6135 My desktop Celeron300A ./zeus_bench localhost / -c 25 -n 1000 = 75 req/sec # apache ./zeus_bench localhost / -p 8080 -c 25 -n 1000 # zope 2.0.1 pystone.py = 3831 and apache throughput does not go much higher even when I overclock it to 450MHz, even thoug pystone jumps to about 5800 too. Both are running linux 2.2.1X and apache 1.3.9 Conclusion: servers just serve faster than desktop computers ;) --------------------- Hannu