Swazoo vs. Apache benchmarks
Locked
 
|
Dear Web Smalltalkers,
In light of our last debate about Swazoo in comparison to Apache and out of curiosity I did some benchmarks of them to clarify our position on the web serving scene. Here are the preliminary results in a table: http://www.swazoo.org/benchmarks/swazoo-vs-apache.html As you see Swazoo is not bad at all, only 4x slower from Apache on small files on VW! On Squeak it is 18x slower. For a 100KB file VW it is 12x slower while Squeak is 40x. I did benchmarking on Gemstone too but those results are currently quite disappointing. Probably because Hyper is not so optimized as the new Swazoo 2 HTTP server, so that porting a new Swazoo there would help too. I think we can conclude that Swazoo (at least on VW for now) is actually able to serve quite demanding web sites, but as we already said, Swazoo is not there to compete with Apache but to be a simple starting step, which can later evolve if needed to the most demanding serving on Apache. Best regards Janko -- Janko Mivšek AIDA/Web Smalltalk Web Application Server http://www.aidaweb.si _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Re: Swazoo vs. Apache benchmarks
|
|
> In light of our last debate about Swazoo in comparison to Apache and out
> of curiosity I did some benchmarks of them to clarify our position on > the web serving scene. Here are the preliminary results in a table: > > http://www.swazoo.org/benchmarks/swazoo-vs-apache.html Just out of curiosity. How does Apache compare to Lighty (http://www.lighttpd.net/) when serving these files? Lukas -- Lukas Renggli http://www.lukas-renggli.ch _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Re: Swazoo vs. Apache benchmarks
|
|
Lukas Renggli wrote:
>> In light of our last debate about Swazoo in comparison to Apache and out >> of curiosity I did some benchmarks of them to clarify our position on >> the web serving scene. Here are the preliminary results in a table: >> >> http://www.swazoo.org/benchmarks/swazoo-vs-apache.html > > Just out of curiosity. How does Apache compare to Lighty > (http://www.lighttpd.net/) when serving these files? Interesting that it works worse that Apache on httperf tool, That is, Lighty start generating errors at 1100 req/s while Apache at 2000, for 3KB file. For 100KB file errors start at 1500 on both. While Swazoo never raise an error even on very big load. Best regards Janko -- Janko Mivšek AIDA/Web Smalltalk Web Application Server http://www.aidaweb.si _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Re: Swazoo vs. Apache benchmarks
|
|
In reply to this post by Janko Mivšek
2008/2/10, Janko Mivšek <[hidden email]>:
> Dear Web Smalltalkers, > > In light of our last debate about Swazoo in comparison to Apache and out > of curiosity I did some benchmarks of them to clarify our position on > the web serving scene. Here are the preliminary results in a table: > > http://www.swazoo.org/benchmarks/swazoo-vs-apache.html > > As you see Swazoo is not bad at all, only 4x slower from Apache on small > files on VW! On Squeak it is 18x slower. For a 100KB file VW it is 12x > slower while Squeak is 40x. > > I did benchmarking on Gemstone too but those results are currently quite > disappointing. Probably because Hyper is not so optimized as the new > Swazoo 2 HTTP server, so that porting a new Swazoo there would help too. > > I think we can conclude that Swazoo (at least on VW for now) is actually > able to serve quite demanding web sites, but as we already said, Swazoo > is not there to compete with Apache but to be a simple starting step, > which can later evolve if needed to the most demanding serving on Apache. even breaks the 1 MB/s barrier (B = bit?). Without any handbreaks on and without any tuning Apache on my local machine serves 60 to 120 Mbyte/s. That's what a fast ethernet connection can theoretically deliver. What's worse is that serving files with Squeak is not only slow but also hogs the CPU. Since Squeak can use only one core serving files with Squeak takes away desperatly needed CPU cycles from Seaside (Seaside is a very inefficient framework). On a modern CPU Squeak can only use a fourth of the available processing power. I do not want to give any of that away because I don't use the right tool for the job. But the real killer is that Squeak can not handle a sustained load. Do something like: ab -c 10 -n 1000000 http://127.0.0.1:9090/seaside/files/WADispatcherFiles/logo.png This will toast your Squeak image. It will take a very long time because Squeak is so slow and after that your Squeak image will still hog the CPU and stop reacting. To sum up: there a problems for with Smalltalk is not the right tool. Serving static files is such job. We should stop pretending that all our problems are nails. Cheers Philippe _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Re: Swazoo vs. Apache benchmarks
|
|
>>>>> "Philippe" == Philippe Marschall <[hidden email]> writes:
Philippe> To sum up: there a problems for with Smalltalk is not the right tool. Philippe> Serving static files is such job. We should stop pretending that all Philippe> our problems are nails. Serving *lots* of static files, sure. However, I know many websites that get a dozen hits an hour or less. Smalltalk would do just fine for that. -- Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095 <[hidden email]> <URL:http://www.stonehenge.com/merlyn/> Perl/Unix/security consulting, Technical writing, Comedy, etc. etc. See PerlTraining.Stonehenge.com for onsite and open-enrollment Perl training! _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
|
|
On Feb 10, 2008, at 2:52 PM, [hidden email] (Randal L.
Schwartz) wrote: >>>>>> "Philippe" == Philippe Marschall <[hidden email]> >>>>>> writes: > > Philippe> To sum up: there a problems for with Smalltalk is not the > right tool. > Philippe> Serving static files is such job. We should stop > pretending that all > Philippe> our problems are nails. > > Serving *lots* of static files, sure. > > However, I know many websites that get a dozen hits an hour or less. > Smalltalk would do just fine for that. In fact I have not seen any issues worth reporting on the Run BASIC site serving several users per minute. A lot depends on the nature of the application RB runs on Seaside 2.6 and VW 7.4.1 +Web Toolkit. Newer versions are even faster, and I'm eager to port forward. Carl Gundel http://www.runbasic.com _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Re: Swazoo vs. Apache benchmarks
|
|
In reply to this post by Philippe Marschall
Hi Philippe,
Philippe Marschall wrote: >> I think we can conclude that Swazoo (at least on VW for now) is actually >> able to serve quite demanding web sites, but as we already said, Swazoo >> is not there to compete with Apache but to be a simple starting step, >> which can later evolve if needed to the most demanding serving on Apache. > > I can not agree at all. If we look closer at your numbers Squeak realy > even breaks the 1 MB/s barrier (B = bit?). Without any handbreaks on > and without any tuning Apache on my local machine serves 60 to 120 > Mbyte/s. That's what a fast ethernet connection can theoretically > deliver. B=byte, b= bit. My faster Apache benchmark is 156MBytes/s which is therefore near yours results. > > What's worse is that serving files with Squeak is not only slow but > also hogs the CPU. Since Squeak can use only one core serving files > with Squeak takes away desperatly needed CPU cycles from Seaside > (Seaside is a very inefficient framework). On a modern CPU Squeak can > only use a fourth of the available processing power. I do not want to > give any of that away because I don't use the right tool for the job. > > But the real killer is that Squeak can not handle a sustained load. Do > something like: > > ab -c 10 -n 1000000 > http://127.0.0.1:9090/seaside/files/WADispatcherFiles/logo.png > > This will toast your Squeak image. It will take a very long time > because Squeak is so slow and after that your Squeak image will still > hog the CPU and stop reacting. True, when you go over the max requests/s threshold, performance starts to deteriorate very fast. Apache seems to start intentionally breaking connections in that case. > To sum up: there a problems for with Smalltalk is not the right tool. > Serving static files is such job. We should stop pretending that all > our problems are nails. As others already said, Swazoo is primary meant for our web frameworks while static file serving is just for some small stuff you usually have around. I'm doing benchmarks on static serving only to measure Swazoo strengths, nothing more. From memory serving results are therefore most interesting for us. Janko -- Janko Mivšek AIDA/Web Smalltalk Web Application Server http://www.aidaweb.si _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Re: Swazoo vs. Apache benchmarks
|
|
2008/2/10, Janko Mivšek <[hidden email]>:
> Hi Philippe, > > Philippe Marschall wrote: > > >> I think we can conclude that Swazoo (at least on VW for now) is actually > >> able to serve quite demanding web sites, but as we already said, Swazoo > >> is not there to compete with Apache but to be a simple starting step, > >> which can later evolve if needed to the most demanding serving on Apache. > > > > I can not agree at all. If we look closer at your numbers Squeak realy > > even breaks the 1 MB/s barrier (B = bit?). Without any handbreaks on > > and without any tuning Apache on my local machine serves 60 to 120 > > Mbyte/s. That's what a fast ethernet connection can theoretically > > deliver. > > B=byte, b= bit. My faster Apache benchmark is 156MBytes/s which is > therefore near yours results. > > > > What's worse is that serving files with Squeak is not only slow but > > also hogs the CPU. Since Squeak can use only one core serving files > > with Squeak takes away desperatly needed CPU cycles from Seaside > > (Seaside is a very inefficient framework). On a modern CPU Squeak can > > only use a fourth of the available processing power. I do not want to > > give any of that away because I don't use the right tool for the job. > > > > But the real killer is that Squeak can not handle a sustained load. Do > > something like: > > > > ab -c 10 -n 1000000 > > http://127.0.0.1:9090/seaside/files/WADispatcherFiles/logo.png > > > > This will toast your Squeak image. It will take a very long time > > because Squeak is so slow and after that your Squeak image will still > > hog the CPU and stop reacting. > > True, when you go over the max requests/s threshold, performance starts > to deteriorate very fast. Apache seems to start intentionally breaking > connections in that case. even for 10 million requests with 10 requests at a time. But honestly some graceful performance degeneration under high load (eg. rejecting new sessions) is certainly not the worst feature to have. > > To sum up: there a problems for with Smalltalk is not the right tool. > > Serving static files is such job. We should stop pretending that all > > our problems are nails. > > As others already said, Swazoo is primary meant for our web frameworks > while static file serving is just for some small stuff you usually have > around. I'm doing benchmarks on static serving only to measure Swazoo > strengths, nothing more. From memory serving results are therefore most > interesting for us. Yes, that is why I took http://127.0.0.1:8080/seaside/files/WADispatcherFiles/logo.png, which is stored in a literal. But even in this case I can not achieve a throughput of more than 1.5 Mbyte/s. I would at least expect ten times this performance. Cheers Philippe > Janko > > > -- > Janko Mivšek > AIDA/Web > Smalltalk Web Application Server > http://www.aidaweb.si > _______________________________________________ > seaside mailing list > [hidden email] > http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside > _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
«
Return to Seaside General
|
1 view|%1 views
| Free forum by Nabble | Edit this page |