| | 1 | {{{ |
| | 2 | Toni |
| | 3 | - Extensions - requirements are training |
| | 4 | - perspective - web interface is not general enough |
| | 5 | - GPU Grid's Soap interface is general enough |
| | 6 | David |
| | 7 | - User Defined in this discussion is the Scientist |
| | 8 | - Different categories of users |
| | 9 | - Envisage a context where random volunteers can submit |
| | 10 | - Need to control the scripts where users can inject viruses |
| | 11 | Toni |
| | 12 | - Users should not have ability to upload applications - this is a security risk |
| | 13 | - Applications that accept scripts should be tightly restricted |
| | 14 | Christian |
| | 15 | - Extend template files w/ additional text. |
| | 16 | - Soap can be used for a web page where scientist selects for |
| | 17 | application to submit |
| | 18 | - Scientist must provide a template |
| | 19 | Toni |
| | 20 | - Webpage for submiting: |
| | 21 | 1) Does not lend itself to batches |
| | 22 | 2) Output would not be able to retrieve on webpage |
| | 23 | Christian |
| | 24 | - Suggested a structure where input is packed in an archive, uploaded |
| | 25 | to the server, uncompressed and used by job creation script |
| | 26 | Zoltan |
| | 27 | -Change Execution soap website |
| | 28 | -Client executes application |
| | 29 | -Master feeds the work |
| | 30 | Toni |
| | 31 | -Desscribed how GPUGrid accounts for the priority of workunits. High |
| | 32 | priority starves the lower priority |
| | 33 | -Will release the perl code |
| | 34 | -Does not deal w/ scheduler code |
| | 35 | -One should have a scientist's table that tracks various project |
| | 36 | scientist information relevant to submissions |
| | 37 | David |
| | 38 | - Many options [for scientist's table], eg accounting, controling |
| | 39 | latency, throughput |
| | 40 | Toni |
| | 41 | -Not many scientists on GPU grid currently |
| | 42 | Micheal Weber |
| | 43 | -Leiden Classical has a drop-down menu |
| | 44 | -submissions interface w/ limited characters |
| | 45 | -Registers the heavy users |
| | 46 | David |
| | 47 | -Is Lieden a generic solution? |
| | 48 | Christian |
| | 49 | -Lieden uses a seperate configuration file |
| | 50 | Toni |
| | 51 | -Computes the has of the file, avoid copying files, conserve space, |
| | 52 | use hard links in the download directory |
| | 53 | Kevin |
| | 54 | -There should be a common mechanism to submit |
| | 55 | Zoltan |
| | 56 | -How to submit work remotely |
| | 57 | David |
| | 58 | -It would be good to have a standardized way |
| | 59 | Toni |
| | 60 | -Use MD5 checksum on the input files, store the hashes on the server |
| | 61 | Jack |
| | 62 | -What if there were a script to generate the templates, where user is |
| | 63 | prompted for various parameters? |
| | 64 | David |
| | 65 | -Not the bottleneck for scientists |
| | 66 | -Need to move the WU Generator and Assimilator off the server |
| | 67 | -Keep the things we trust stay on the server |
| | 68 | |
| | 69 | -- Jack http://drugdiscoveryathome.com http://hydrogenathome.org |
| | 70 | |
| | 71 | }}} |
