Some benchmarks comparing Seaside on different dialects

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Some benchmarks comparing Seaside on different dialects

Esteban A. Maringolo
Hi,

I've just published the results of a sort of "brute force" attack,
against Seaside on Dolphin, Squeak an VW.
All the setups where with "out of the box" images (no special VM
tuning) + the installing instructions for each Seaside port.

You can access its results and some observations here:
<http://dolphinseaside.blogspot.com/2007/11/some-benchmarks.html>

I would like to share comments about this, and I'm open to
recommendations about better benchmarking procedures.

Regards,

--
Esteban A. Maringolo
[hidden email]
_______________________________________________
seaside mailing list
[hidden email]
http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
Reply | Threaded
Open this post in threaded view
|

RE: Some benchmarks comparing Seaside on different dialects

Ron Teitelbaum
Hi Esteban,

Your benchmarks are very dependent on what hardware configuration you are
using.  You may have mentioned that in your links which I didn't follow, but
I'd suggest that you put that information upfront.  I'd also suggest that
you flush out the details of VW like you did with Dolphin and Squeak.  

You should also consider specifying that the limitations of a single VM do
not take into account load sharing techniques that have been developed by
the Seaside community which allow Seaside to massively scale up from your
limited configuration.

Happy Coding!

Ron Teitelbaum

> From: Esteban A. Maringolo
> Hi,
>
> I've just published the results of a sort of "brute force" attack,
> against Seaside on Dolphin, Squeak an VW.
> All the setups where with "out of the box" images (no special VM
> tuning) + the installing instructions for each Seaside port.
>
> You can access its results and some observations here:
> <http://dolphinseaside.blogspot.com/2007/11/some-benchmarks.html>
>
> I would like to share comments about this, and I'm open to
> recommendations about better benchmarking procedures.
>
> Regards,
>
> --
> Esteban A. Maringolo
> [hidden email]
> _______________________________________________
> 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
Reply | Threaded
Open this post in threaded view
|

Re: Some benchmarks comparing Seaside on different dialects

Esteban A. Maringolo
Ron,

On Nov 15, 2007 10:27 AM, Ron Teitelbaum <[hidden email]> wrote:
> Your benchmarks are very dependent on what hardware configuration you are
> using.

The test were ran in the same computer, on the same OS. Using loopback
interface (i.e. localhost).
I don't know if squeak or VW runs better in Linux and/or Sparc, I
wanted to measure the three in the same "environment".

> You may have mentioned that in your links which I didn't follow, but
> I'd suggest that you put that information upfront.  I'd also suggest that
> you flush out the details of VW like you did with Dolphin and Squeak.

I published the results for VW, Squeak and Dolphin, no hidden details.
I have no bias here, in fact I wanted to know how well (or bad)
performed each.

And believe me, everything was as expected. VW was fast but died fast,
Squeak slower but solid, and Dolphin in a point between both. However
this doesn't mean I prefer one platform over the other just by
considering the test, which is a brute force test, that showed me (and
I wanted to share with the community), how does a VM perform when
requested brutally.

> You should also consider specifying that the limitations of a single VM do
> not take into account load sharing techniques that have been developed by
> the Seaside community which allow Seaside to massively scale up from your
> limited configuration.

Don't be so harsh, I'm from the "seaside side" of the community.
I know the limited configuration, but it was also the intention,
testing it "out of the box". I was trying to measure how much a single
VM can scale under _heavy_ load. To load balance one should know much
load can handle a single VM.

However I don't know what is the concept of "massively scale up" in
Seaside, because we don't have available numbers public.
The two massive sites that come to my mind are Dabble and the Seaside
website itself.

I'm not a raw performance zealot, but I want to know what's under the
hood of the car I'm driving. After that, I know what tune, what not,
and what to avoid.

Best regards,


--
Esteban A. Maringolo
[hidden email]
_______________________________________________
seaside mailing list
[hidden email]
http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
Reply | Threaded
Open this post in threaded view
|

Re: Some benchmarks comparing Seaside on different dialects

jgfoster
In reply to this post by Esteban A. Maringolo
Esteban A. Maringolo wrote:

> Hi,
>
> I've just published the results of a sort of "brute force" attack,
> against Seaside on Dolphin, Squeak an VW.
> All the setups where with "out of the box" images (no special VM
> tuning) + the installing instructions for each Seaside port.
>
> You can access its results and some observations here:
> <http://dolphinseaside.blogspot.com/2007/11/some-benchmarks.html>
>
> I would like to share comments about this, and I'm open to
> recommendations about better benchmarking procedures.
>
> Regards,
It would be interesting to throw GemStone into the mix. Is your hardware
64-bit? Does it support virtualization? Dale has done some benchmarking
and shared his results here:
http://gemstonesoup.wordpress.com/2007/10/19/scaling-seaside-with-gemstones/.

James
_______________________________________________
seaside mailing list
[hidden email]
http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside