eWeek Retest Shows 2.5-fold Apache Speedup

A reader writes "There's a retest story on ZD-Net about an upgraded Apache system. Apache 1.3.19 running on a Red Hat Linux 7.1 system upgraded to the 2.4.5 kernel was able to process at peak throughput 4,602 Web requests per second. The last time eWEEK Labs did a big round of Web server benchmarking on Linux was two years ago this month, when we did a retest of Linux and Windows performance numbers as an audit of the Mindcraft Inc. tests. "

Re:fp for 2 in a row!

[Snowfox]

2nd tsop tsrif in a row!

I believe that the reversal of characters would constitute a grammatical error.

Keeping /. free of grammatical errors for 3 years.

As such, your sig is forfeit, grammar nazi. I hope it was worth the pleasure - it's been a long run.

No comparison to NT

[mgkimsal2]

But they've not done a recomparison of NT/Linux. It's great to see Linux/Apache together making enormous strides, but NT hasn't just remained static. NT5 and IIS5 have surely made some improvements in speed, etc. Any current NT numbers on similar hardware?

part of larger eWEEK story on Tux 2.0 / Linux 2.4

[Timothy Dyck]

Hi, this story is just one part of a larger package published in this week's eWEEK. In that story, we tested Tux, Tux with Apache, Apache and IIS. The link is http://www.zdnet.com/eweek/stories/general/0,11011 ,2774242,00.html [zdnet.com].

Regards,
Tim Dyck
eWEEK Labs

Dying for Data of Interest

[4of12]

I would really be interested to see a full chart, showing static and dynamic web page performance of different OS+revision/httpserver+revision just so we can see what kind of progress has been made in this area over the past several years.

Yes, even entries with asterisks indicating heavy customization and tuning, like Ingo Molnar's tux server, would be interesting.

Probably, though, this is the kind of information that only gets into white papers that sell for $2500 apiece on tightly-controlled distribution.

now make HTTP caching proxies faster!

[poleshifter]

Since this is generic technology that is now part of the kernel, there is no reason why Squid (a full-featured, but rather slow caching proxy application) can't take advantage of this too! I would love to see some free software that outperforms all the proprietary crap that companies like Network Appliance, Inktomi, Infolibria, and others charge tens of thousands of dollars for. Perhaps it would trickle down to becoming cost-effective enough to embed a caching proxy into a $200 DHCP/NAT box to speed up all users. Just wishful thinking out loud from someone who can't get a high-speed connection at home.

Re:Dying for Data of Interest

[4of12]

Yes, my death wish is the command of the IT news community, as this report [zdnet.com] about the blazing speed of Tux 2.0 appeared recently.

Re:No comparison to NT

[Anonymous Coward]

Another story on ZDNet [zdnet.com] says 5,137 requests per second for IIS 5.0 on Windows 2000.

Re:Dying for Data of Interest

[rgmoore]

Actually, one of the interesting points that they make in the article (taken straight from Ingo Molnar) is that much of the speedup in Apache is a result of inherited advantages from Tux. Many of the speedups that made Tux so fast have now been rolled into the mainstream kernel, so every application can take advantage of them. Molnar is quoted as saying that in the latest version of Tux, only 2% of the total time on the benchmark was spent in Tux-specific code, as opposed to over 50% in the original version.

You can say that this is an example of the real strength of Free Software. Tux served as a testing ground for a whole bunch of cool new technologies. When they showed how useful they are, Linus was able to roll them into the mainstream kernel. None of that would have been possible if Linus and Ingo hadn't been able to share their code.

Re:So after more than a year of tuning...

[khuber]

Not sure I get the intention of your post, but
IIS has been improved also I'm sure. And IIS
probably has more development resources.

-Kevin

Lies, damned lies and benchmarks

[return 42]

I don't doubt that Apache and the kernel have gotten faster, but still...we all know how reliable and meaningful benchmark results are.

useless comparisons

[m08593]

If you want speed for static pages, you can use the Linux kernel server and/or keep the data in RAM. But what's the point? You can more than saturate any reasonable internet connection with a fraction of that performance. This has less relevance to real life that formula 1 racing has to driving.

Most performance problems on real web sites come from dynamic page generation. And that performance is limited by choice of database, implementation language, and (most importantly) data model and distribution model among multiple servers. You can't make general comparisons there. At best, you can test your own application at a particular size on a variety of platforms and see which runs best. But even that comparison may be invalid tomorrow, after you made some changes to your system or your site has grown.

Re:No comparison to NT

[mgkimsal2]

But it didn't say what hardware...

Re:now make HTTP caching proxies faster!

[c0d3z3r0]

As a squid developer the only thing I'm lacking is time. You'll find that the strides being made in apache are partially funded by people being paid to work on this. I can tell you that this isn't happening with squid. :-)

Tux 2.0

[NeuroMorphus]

Tux 2.0....nuff said... ~=NeuroMorphus=~