Stress test for Seaside …

When playing around with Apache and LoadBalancing I found a little tool named ab in the binary directory of Apache. It offers you a way to put very, heavy traffic on your Seaside server.

The syntax of this tool is:

ab [options] url

I tested against the counter example of Seaside on various Smalltalk implementations:

I used

“ab -n10000 -c4”

to query the counter example 10000 times with a concurrency of 4.

In parallel to this tool I used a browser and looked for a correct behaviour of the counter application.

What now happens (I tested it under Windows 7 )?

* VASmalltalk 8.5 introduces the problem with white HTML pages again. You may find this problem described in earlier postings when Seaside was introduced with VASmalltalk. I thought, that this problem went away with 8.5 – but under heavy stress it’s there.

* Pharo 1.3.13315 simply crashes – this is due to the option “-c”

* VisualWorks 7.8 had no problems

All Smalltalks were running at 100% during these tests.

One additional remark: the problems may be due to the Smalltalk implementation on that Windows operating systems. I was not able to reproduce the white page problem under Linux and VASmalltalk … but actually I only had a virtual pc to test it. That would be very interesting to see ….

PS.: GemStone also did its job …

PS.II: I added a Video showing the empty page HTMl error within VASmalltalk

This entry was posted in Smalltalk and tagged , , . Bookmark the permalink.

5 Responses to Stress test for Seaside …

  1. paul says:

    which seaside adaptor?

  2. Marten says:

    The Squeak community found pretty fast a solution for this: I had to use the newest cog virtual machine and the problems went away

  3. Philippe Marschall says:

    Hmm, do you have a PharoDebug.log or something? It should not crash and certainly not due to “-c”. Please file a bug. is actually a very bad test. It basically end up stressing the session creation code. This code doesn’t scale very well. We haven’t optimized this code yet because outside of benchmarks it never was an issue. If you want to stress the sever adaptors there’s the Seaside-Benchmark package.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s