This is a "test post" to find out something about the Scratch Server performance. I'm in eMail-Communication with Andres because of that issue. Some time ago it was nearly no possible to write to this forum for me (after "Submit" it lasted 5 minutes and sometimes there was nor reaction. Andres is testing the Scratch Server performance from different locations. To give others an introduction I quote our Mail exchange:
Martin wrote:
Hi Andres,
thank you very much for your fast an proactive answer!
Excuse me that I didn’t describe the problems more detailed:
You are right: The performance of the Scratch Server has improved a lot when receiving data.
Especially this morning when I read your mail and tested (9:30 CET), it seems to be much faster than yesterday evening (23:00 CET).
Perhaps it depends much on the daytime traffic?
However, biggest problems we have is not with receiving but with sending data:
This morning, when receiving performance at the Scratch Server was very good, I again tested to change the description of http://scratch.mit.edu/galleries/view/1638 " DACH German SCRATCHERS Meeting point" and it didn't (it did some month ago).
I now tested the forum again at: http://scratch.mit.edu/forums/viewtopic.php?pid=12318#p12318 and you are right:
It is now really very fast, also with writing.
So (if my test was representative) there only stays the above mentioned problem with the gallery comment: Perhaps that isn’t caused by performance issues but by something else? It’s also bothering to have such a small edit-box for long gallery comments and the international character-problem seems not to be solved either.
Thank you again for your great work at ScratchR!
[...]
Best regards from Europe / Mit freundlichen Grüßen
Martin
Andres wrote:
I found Gomez.com, a company that lets you test the response time of a website from different geographical locations: http://www.gomez.com/info_center/instant-test.php
Apparently Gomez is one the leaders in that kind of performance testing.
I tried different sites to test their response time. I chose London trying to get as close as possible to Germany.
The http://scratch.mit.edu/forums/ page took 3.413 secs while http://www.scratchindeutsch.de took 3.841 secs. This is response time on the first visit, I assume once a browser caches the images is a bit faster for both.
I am now wondering if the problem is on the network connection. You had mentioned you have a high-speed connection, so I am a bit puzzled. Another theory is that I was doing the testing during a low load time. So if you're experiencing slowness on the site, could you try the performance monitoring tool at that moment? It would be great if when you try it on Scratch you also try it on http://www.scratchindeutsch.de to keep as a comparison point.
Here is a list of sites I tested and their corresponding response time. When testing from NYC the results were faster for US-based sites as expected.
Selected URL: http://www.scratchindeutsch.de/
Selected Node: London, UK - Global Crossing
Test Time: 2007-12-01 04:59:54.640
Total Response Time: 3.841
Total Bytes: 445,457
Selected URL: http://scratchindeutsch.de/index.php?option=com_fireboard&Itemid=26&func=vi
Selected Node: London, UK - Global Crossing
Test Time: 2007-12-01 05:01:10.783
Total Response Time: 2.63
Total Bytes: 446,152
Selected URL: http://de.yahoo.com/
Selected Node: London, UK - Global Crossing
Test Time: 2007-12-01 05:02:44.347
Total Response Time: 0.117
Total Bytes: 256,928
Selected URL: http://www.ww-it-consulting.de/
Selected Node: London, UK - Global Crossing
Test Time: 2007-12-01 05:03: 15.623
Total Response Time: 0.315
Total Bytes: 61,750
Selected URL: http://scratch.mit.edu/forums/
Selected Node: London, UK - Global Crossing
Test Time: 2007-12-01 05:05:08.993
Total Response Time: 3.413
Total Bytes: 289,179
Selected URL: http://scratch.mit.edu/forums/viewtopic.php?id=2045
Selected Node: London, UK - Global Crossing
Test Time: 2007-12-01 05:08:05.567
Total Response Time: 2.65
Total Bytes: 324,720
Selected URL: http://scratch.mit.edu/
Selected Node: London, UK - Global Crossing
Test Time: 2007-12-01 05:04: 05.980
Total Response Time: 4.849
Total Bytes: 452,312
Selected URL: http://scratch.mit.edu/forums
Selected Node: New York, NY - Sprint
Test Time: 2007-12-01 05:09: 29.010
Total Response Time: 2.534
Total Bytes: 289416
I also tried this less well known site for testing: http://www.websitepulse.com/help/tools.php They do offer a location in Munich to fire the request. I found similar results:
http://scratch.mit.edu/forums
Munich: 1.7909 secs
Seattle, USA: 2.0699 secs
Brisbane, Australia: 7.7081 secs
See results: http://www.websitepulse.com/tests/cbd5b71b82015a09.html
http://www.scratchindeutsch.de
Munich: 0.3285
Seattle, USA: 18.4481 secs
Brisbane, Australia: 44.1577 secs
http://www.websitepulse.com/tests/a08f2aa3dfd8b580.html
It's almost shocking to find such big discrepancies based on geographical location. It makes me wonder how Australians experience the web!!![]()
Best.
On 11/30/07, Andres Monroy-Hernandez wrote:
Hi Martin ,
Thanks for sharing your thoughts.
I am very surprised that you're still having performance problems with the site because on my side the response time has improved a lot. I am going to see if we could come up a with a way to test response time from different geographical locations. Maybe there is a firefox plugin that does that. I'll email you if I find something good.
I was about to add a German forum in our forums website http://scratch.mit.edu/forums as an initial test to see how popular other languages are. Ideally we will have something like http://scratch.mit.edu/de/forums and http://scratch.mit.edu/es/forums and so on.
However, seems like the response time for you guys is still unbearable. It's weird because if there is something that I experience faster performance is the forums more than the main site. The forums are using a very lighweight and popular backend called PunBB used in forums with much much more traffic than ours.
I am now wondering if the response time is a combination of network and server side issues. I'll follow up with you on this more.
Kevin Caluser has been doing great work in helping with the multi language support of ScratchR, I hope we can release those changes to the production site soon.
Thanks again for all your interest in Scratch.
Cheers.
On 11/30/07, Martin wrote:
Hi Andres, John and all,
[...]
1. The server-performance issue
The performance problems have not really vanished: e.g. in the last days I tried several times to change the text of " DACH German SCRATCHERS Meeting point" http://scratch.mit.edu/galleries/view/1638 and it didn't work (it did some month ago): Wait, wait, no reaction and later the text stayed the same. Other Scratchers told me similar experiences.
Yes, I think Scratch will get so important, that an environment like Amazon's EC2 (http://www.amazon.com/b?ie=UTF8&node=201591011 ) will be necessary in the middle range sight.
[...]
Andres : I would like to help, but I'm sorry that I don't have experience with that server issues. The international Scratch-Community is growing so fast, that there surely will be some helping hands somewhere else. There are many people who have an interested look at your project, but don't use the forum (http://scratch.mit.edu/forums/ ) at the Scratch site. Today, you have to be brave to use it, because of the server performance: To me it seems that your forums technology is made for a much smaller community and can't cope with the upcoming worldwide boom. Perhaps it could help to make a official "Meta-Scratch" Website and Forum with other Technologies, that deals more with the high level meta-stuff, like technologies, marketing and so on? I'm not sure if it is really useful to mix that up with the end-user-stuff?
Perhaps the problem is: If you can't use the scratch-server because of performance-reasons (and limited discussion structure of the forum) to discuss how to solve these issues, you'll not find the international help you could find?
2. The multiple language issue
[...]
Thank you for your feedback in advance!
Best regards from Europe / Mit freundlichen Grüßen
Martin
Last edited by MartinWollenweber (2007-12-02 08:30:55)
Offline