Server upgrade and the last stage of alpha-testing
log in

Advanced search

Message boards : News : Server upgrade and the last stage of alpha-testing

Author Message
Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 62
Credit: 9,675
RAC: 0
Message 78 - Posted: 24 Nov 2016, 20:38:45 UTC
Last modified: 26 Nov 2016, 17:18:07 UTC

Server will be upgraded to Amazon c4.large instance this Saturday. The upgrade should be easy but who knows what can go wrong.
When the server will be upgraded, the number of new WUs generated per hour will be increased 5-10 times. The final stage of alpha-testing will last for 2-3 days. Then the project will be suspended for long-term maintenance which will take a few months.

During the maintenance:


  • The periodic boundary conditions will be implemented in XaNSoNS to get rid of the divergence of the diffraction pattern at low scattering angles.
  • The native BOINC apps will be created (the current apps use the BOINC wrapper).
  • The data calculated so far will be analyzed for possible errors and inconsistencies.
  • The infrastructure of the database of the diffraction patterns will be created.


Update: Server upgrade was successful. The number of new WUs generated per hour is increased tenfold. However, the parameters of size_regulator are to be optimized...

[VENETO] boboviz
Send message
Joined: 16 Nov 16
Posts: 16
Credit: 20,756
RAC: 0
Message 79 - Posted: 25 Nov 2016, 9:45:17 UTC - in response to Message 78.

Great news.

long-term maintenance which will take a few months

How many? March 2017? May?

This site/forum will remain on-line during stop of wus? This can help volunteers to see the progress of maintenance...and restart crunching!! :-)

[VENETO] boboviz
Send message
Joined: 16 Nov 16
Posts: 16
Credit: 20,756
RAC: 0
Message 80 - Posted: 25 Nov 2016, 9:54:31 UTC - in response to Message 78.

*The native BOINC apps will be created (the current apps use the BOINC wrapper).


Are you using wrapper with Cuda and Opencl? :-O

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 62
Credit: 9,675
RAC: 0
Message 81 - Posted: 25 Nov 2016, 11:36:30 UTC - in response to Message 79.


long-term maintenance which will take a few months

How many? March 2017? May?

Timing depends on whether I have to do all the work alone or I'll find someone to develop the database. Let's say May - July. Now I see that the original date (end of 2017) was too pessimistic.

This site/forum will remain on-line during stop of wus? This can help volunteers to see the progress of maintenance...and restart crunching!! :-)

The site and the forum will remain online. I'll post the progress of work from time to time. However, the registration of new users may be suspended.

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 62
Credit: 9,675
RAC: 0
Message 82 - Posted: 25 Nov 2016, 11:49:54 UTC - in response to Message 80.

Are you using wrapper with Cuda and Opencl? :-O

Yes. Actually, I am surprised that CUDA and OpenCL apps work so well with the wrapper.

Currently, the wrapper does not support OpenCL but I found a workaround. The original app uses the built-in XML parser (TinyXML) to parse the initial data, so I wrote a function that directly (without BOINC API calls) reads GPU type and GPU index from init_data.xml.

[VENETO] boboviz
Send message
Joined: 16 Nov 16
Posts: 16
Credit: 20,756
RAC: 0
Message 83 - Posted: 26 Nov 2016, 8:51:24 UTC - in response to Message 82.

Yes. Actually, I am surprised that CUDA and OpenCL apps work so well with the wrapper.


If you have these performances with wrapper, i think Opencl's and Cuda's native app will be great!!! :-)

Steve Hawker*
Send message
Joined: 11 Nov 16
Posts: 11
Credit: 18,493
RAC: 0
Message 86 - Posted: 26 Nov 2016, 22:15:24 UTC - in response to Message 78.

Update: Server upgrade was successful. The number of new WUs generated per hour is increased tenfold. However, the parameters of size_regulator are to be optimized...

Just ran out of work on my fastest machine. Perhaps you need to increase by a hundredfold?

Yavanius
Avatar
Send message
Joined: 18 Nov 16
Posts: 11
Credit: 419
RAC: 0
Message 88 - Posted: 27 Nov 2016, 6:32:32 UTC - in response to Message 78.

For those who want to track their work especially if the site goes down/resets/etc...

Free-DC is carrying the stats so even if Vlad gets hit by a laster beam and disappears into an alternate world. ;)

http://stats6.free-dc.org/proj/xan

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 62
Credit: 9,675
RAC: 0
Message 90 - Posted: 27 Nov 2016, 9:03:14 UTC - in response to Message 86.


Just ran out of work on my fastest machine. Perhaps you need to increase by a hundredfold?

Sorry, it's possible to double the current number of WUs, but not to increase it tenfold again. However, there is no point to do it anyway, because, we just can't process more COD entries than it have). After the restart in 2017, I will add the larger crystallites. This will significantly increase the average execution time of the WU.

Message boards : News : Server upgrade and the last stage of alpha-testing


Main page · Your account · Message boards