Supporting SARS-COV2 related research
Message boards :
News :
Supporting SARS-COV2 related research
Message board moderation
Previous · 1 · 2 · 3 · 4 · Next
Author | Message |
---|---|
capri Send message Joined: 17 Mar 20 Posts: 3 Credit: 2,266 RAC: 0 |
Hello, Just wondering if that limitation is working as expected. My oldish laptop downloaded about 700 tasks today and over 600 are still waiting to start. Thats enough work for the next days while other computers may be idle. |
rituarora Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 119 Credit: 0 RAC: 0 |
Greetings - thanks for letting us know. We will revisit the settings again today. Best Regards, The BOINC@TACC Team Hello, |
rituarora Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 119 Credit: 0 RAC: 0 |
Thanks Greg for all the suggestions thus far and for your kind support! Best Regards, The BOINC@TACC Team At the moment my queue for my machine is full. But I will bump my resource share up to 150% to get your project at the top of the list for new tasks. |
rituarora Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 119 Credit: 0 RAC: 0 |
Dear All, Here is a news article about the research on COVID-19 vaccine for which BOINC@TACC and other TACC systems are being used: https://elpasoheraldpost.com/utep-school-of-pharmacy-developing-covid-19-vaccine-drug-treatments-using-supercomputing/ Thanks, The BOINC@TACC Team |
Henk Haneveld Send message Joined: 17 Feb 19 Posts: 8 Credit: 8,682 RAC: 0 |
Admin, when you do that you may want to reconsider the settings to use. In an other post you gave a fixed number of max 256 results per host. It would be beter to use a fixed number per core. Example: 2 per core on a 128 core will also give 256 results but for a host with only 4 cores this wil be a max of just 8 results. This will give a more even distribution based on the processing capacity of the host. Henk Greetings - thanks for letting us know. We will revisit the settings again today. |
xRXT2toIGA Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 61 Credit: 4,266 RAC: 0 |
Hello, Thank you to you all for bringing it to our attention. We have reduced the number of jobs run in a client as 5 per CPU core. i.e. : a processor with 4 cores can get maximum 20 jobs at the same time. There will be validation run for the same number of jobs. Hence the host would be running 40 jobs in this example. Sincerely, the BOINC@TACC development team |
Henk Haneveld Send message Joined: 17 Feb 19 Posts: 8 Credit: 8,682 RAC: 0 |
Hello, The change to 5 per core looks to work nicely. What I don't understand is how you get the number of 40 jobs What I do notice is that sometimes I get both the _0 en _1 version of the same job. If you want solid validation they should be send to separate hosts and you should check your settings to fix that.. If validation is not that important then there is no point for sending out 2 versions of the same job |
rituarora Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 119 Credit: 0 RAC: 0 |
Hello Henk, It seems like with the current project settings, the validation is indeed running on the same hosts. We will look into adjusting these settings. Thanks, The BOINC@TACC Team Hello, |
fsu95OTBmn Send message Joined: 7 Mar 19 Posts: 15 Credit: 376,286 RAC: 0 |
What is the current settings for maximum work per host? I'm not asking to have anything changed, only trying to understand the limits imposed. If it is 5 per core (thread), then my 128 thread system should get 640 as a maximum. It is only getting 320. Is there also a per host limit set? |
xRXT2toIGA Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 61 Credit: 4,266 RAC: 0 |
Greetings, A 128 core system should receive 640 maximum task at a time. Are all the 128 threads hardware cores or is this the hyperthread number? As of now, we are setting 5 per core, so you should be receiving 640 total. However, if your computer uses hyperthreads or virtual cores, it is possible that you may be receiving lesser tasks. Sincerely, the BOINC@TACC development team |
fsu95OTBmn Send message Joined: 7 Mar 19 Posts: 15 Credit: 376,286 RAC: 0 |
Thanks for the response, This system has 64 hardware cores and 64 virtual cores (hyperthreading) that make up the 128 threads. If your system has detected only the 64 hardware cores, then the 320 is the correct number. However, I have 10 other systems that are a mix of hardware and virtual cores (hyperthreaded) and your system sends work units to those systems based on both combined as expected. Why doesn't the scheduler recognize the virtual cores on the EPYC system but does on the others? |
xRXT2toIGA Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 61 Credit: 4,266 RAC: 0 |
Hello, We have not tested the BOINC clients with hyper-threading enabled. We will post an update once we have tested the mapping of tasks to the hyper-threads. We will also check with other BOINC projects about this through the BOINC mailing list. Sincerely, the BOINC@TACC development team |
rituarora Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 119 Credit: 0 RAC: 0 |
Greetings, As of April 12, 2020, the researchers had completed screening more than 1124781 compounds through BOINC@TACC. There are many more results from the screening jobs completed on April 13 - April 14, 2020, and we will know their exact count after the post-processing scripts have run. The researchers are in the process of preparing new datasets for the next round of screening compounds. Till this dataset is ready (hopefully in the next 2-3 days), the number of jobs running through BOINC@TACC will reduce. Additionally, below is the summary of what we have found so far regarding the faulty work-units that were run on some hosts in the previous few batches of jobs: 1) In a particular batch of jobs, a file in the dataset got corrupted and the application failed to parse it, and hence produced error messages. 2) There were some input files (in the library of millions of files) that did not contain the correct parameters. Such input is hard to detect manually. The researchers have been informed about this and they would be writing scripts to check the quality of the input before submitting the next batch of jobs. 3) Virtual Box versions 6.1.4 and 6.1.2 are not compatible with the BOINC clients on some operating systems, and hence, the work-units may not get processed smoothly on these combinations of Virtual Box and operating systems. Thanks for your continued support for BOINC@TACC! The BOINC@TACC Team |
Elevator Startups Send message Joined: 16 Apr 20 Posts: 1 Credit: 0 RAC: 0 |
Hi, could you tell us. Should we expect new wok units any time soon to fight pandemic? Hope to donate…. |
rituarora Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 119 Credit: 0 RAC: 0 |
Greetings, We may not be able to test VB 6.1.2 and 6.1.4 with the different operating systems before the new work units are submitted by the researchers. We are under "shelter at home" order right now and do not have access to the additional hardware for installing and testing different operating systems and VB versions. However, if we are able to get access to bare metal nodes in the cloud and are able to succeed with the testing, we will post an update. Thanks, BOINC@TACC Team Question: |
rituarora Project administrator Project developer Project scientist Send message Joined: 4 Feb 19 Posts: 119 Credit: 0 RAC: 0 |
Greetings, Yes, new work units should be available soon. There are several million comparisons of compounds that still need to be done. Thanks for your kind support! Thanks, BOINC@TACC Team Hi, could you tell us. Should we expect new wok units any time soon to fight pandemic? Hope to donate…. |
fsu95OTBmn Send message Joined: 7 Mar 19 Posts: 15 Credit: 376,286 RAC: 0 |
Greetings, A 128 core system should receive 640 maximum task at a time. Are all the 128 threads hardware cores or is this the hyperthread number? As of now, we are setting 5 per core, so you should be receiving 640 total. However, if your computer uses hyperthreads or virtual cores, it is possible that you may be receiving lesser tasks. Is this parameter set on the on the server: <max_ncpus>N</max_ncpus> An upper bound on NCPUS (default: 64) If so, could it be set to 128 or higher? There are servers shipping right now that have 256 thread capability. Since this is an upper bound parameter and the server is going to use whatever is reported from the client, why not set it to 512 and not have to re-visit in the near term. Thank you in advance for your consideration.[/b] |
Henk Haneveld Send message Joined: 17 Feb 19 Posts: 8 Credit: 8,682 RAC: 0 |
I run version 5.2.38 |