Your numbers are very interesting. I've been telling people Sun hardware is just as fast but cheaper and they look at me like I've got 3 eyes or something.
v240 2 processor 2 gig ram 2 x 73 gig drives Solaris 9 $6895 1747 c/s Compaq DL 360 2 processor 2 gig ram 2 x 73 gig drives Redhat AS 2.1 $7405 1835 c/s
On the low end I run Sun X1's off ebay for $400 or the V100 for $995. Even if they only manage 1% of the v240 numbers that's 17 c/s. More than enough for development and most sites.
The Sun is a 64 bit machine and has LOM so you can manage it completely from the serial port. I've never used Compaq. Can you power it up and install it without being there?
Not that is matters but did you test the 1 gig or 1.28 gig Sun box. the 1.28 just came out. Also in that config the v210 would be $4945 with the 1 gig processors. Basically $2500 cheaper than the Compaq and within 10% of the performance.
Given the variability of benchmarks I'd call it a dead heat in performance and price/performance. It's really more which you'd rather use. I personally think you need a 64bit box to run a database and I'd rather have all Sun than mix Solaris and Linux.
I would like to see Xserve numbers. I have run the Xraid and it's very impressive
Barry
On Jun 23, 2004, at 6:58 PM, Adam Leff wrote:
Sure. I think the Ops guys there might still have an Xserve laying around.
He's right... we do. I'm still not sure if it's ever been turned on. ;)
The bigger question is what the bench marks would be exactly.
Bingo.
It all depends on the application. Most of the testing that's done on webservers is *usually* how it deals under duress serving static pages... yes, there are dynamic testing benchmarks out there.
There are so many flippin' ways we use AOLserver just at AOL that it's completely unfair to say that AOLserver X.X is better on Solaris than it is on Linux. You have to take so many things into consideration (external dependencies, databases, compilation options, OS tuning parameters). Comparing AOLserver as used by AOL.com to AOLserver as used by Moviefone.com or even AOLserver as a backend application layer isn't fair. So that's why we do our best to test each application and its dependencies. Sadly, usually the testing is done after the hardware is purchased. Yay for compressed timeframes.
But then again, the prices of x86 hardware (and the associated support contracts) make executives happy. :)
That being said, I did a test a little bit ago slamming the begeezus out of an .adp page with a bunch of ns_adp_puts in it (so I was exercising the Tcl interpreters, not just the fastpath stuff) on a few platforms:
Sun Fire V240, Solaris 9, AOLserver 3.5.10: 1871 conns/sec Sun Fire V240, Solaris 9, AOLserver 4.0.1: 1747 conns/sec Compaq DL 360, RH AS 2.1, AOLserver 3.5.10: 1880 conns/sec Compaq DL 360, RH AS 2.1, AOLserver 4.0.1: 1835 conns/sec Compaq Proliant, RH AS 3.0, AOLserver 3.5.10: 2220 conns/sec Compaq Proliant, RH AS 3.0, AOLserver 4.0.1: 2256 conns/sec
As predicted, Red Hat Advanced Server 3.0 came out on top, most likely due to NPTL. The boxes were all hovering between 60-80% CPU utilization... network saturated.
~Adam
-------------------- Adam Leff AOL Web Operations
-- AOLserver - http://www.aolserver.com/
To Remove yourself from this list, simply send an email to <[EMAIL PROTECTED]> with the body of "SIGNOFF AOLSERVER" in the email message. You can leave the Subject: field of your email blank.
-- AOLserver - http://www.aolserver.com/
To Remove yourself from this list, simply send an email to <[EMAIL PROTECTED]> with the body of "SIGNOFF AOLSERVER" in the email message. You can leave the Subject: field of your email blank.
