[tbnl-devel] SBCL: Can't cope with too many threads

Bob Hutchison hutch at recursive.ca
Sat Aug 13 13:25:25 UTC 2005


Hi,

On Aug 12, 2005, at 2:49 AM, Stefan Scholl wrote:

> Hi!
>
> Just ran a small test with sbcl and standalone TBNL. "-c 50"
> (ApacheBench, 50 simultaneous requests) was too much for the poor
> litte server. It just froze.

I've done tests on a system I'm working on (Sienna) 5000 hits with 50 
concurrent requests and it has run very very well, even when Sienna is 
running in the development environment. I've tried this apache/mod_lisp 
-> TBNL, Araneida -> TBNL, apache/Araneida -> TBNL, and TBNL 
stand-alone. I'm using LispWorks on a mac. I am quite sure I tried all 
of these combinations (completely sure of Araneida->TBNL and TBNL 
standalone) on linux as well. I let it run overnight once with 50 
concurrent hits, no problem at all.

I've only tried higher concurrency experiments on OS X, but never 
experienced a complete failure, it degraded well. I have never tried to 
find a breaking point, 50 concurrent requests would be way more than 
sufficient for any conceivable use of Sienna (of course, now that I've 
said that I'll be proven wrong immediately :-)

Assuming my results are reproducible in other webapps, I think I'd be 
having a look at SBCL...

>
> Don't really know who's to blame. But maybe it's possible to
> limit the requests on the TBNL (KMRCL?) side that are served
> simultaneously -- according to some magic numbers some SBCL wizards
> could supply.

What happens with a smaller number of concurrent users? Is there a 
point where performance abruptly declines or is it a slow descent into 
failure?

>
>
> How do the Araneida guys deal with it? Don't use threads, just
> serve event?
>
>
> Regards,
> Stefan
>
> _______________________________________________
> tbnl-devel site list
> tbnl-devel at common-lisp.net
> http://common-lisp.net/mailman/listinfo/tbnl-devel
>
>
----
Bob Hutchison          -- blogs at <http://www.recursive.ca/hutch/>
Recursive Design Inc.  -- <http://www.recursive.ca/>
Raconteur              -- <http://www.raconteur.info/>




More information about the Tbnl-devel mailing list