| | 17 | * Scooter and Elaine move to N453 |
| | 18 | - Department asked Scooter and Elaine to move to N453. |
| | 19 | - Will happen in coming weeks. |
| | 20 | |
| | 21 | * Phenix R24 |
| | 22 | - Eric gave overview of Phenix and R24 proposal ChimeraX tasks |
| | 23 | - Showed demo of ChimeraX phenix douse command. |
| | 24 | - Discussion of whether Phenix REST server of subprocess call should be used to run Phenix methods. |
| | 25 | - Many disadvantages of REST, starting server, determining when it is ready, polling it, shutting it down. |
| | 26 | - Does not seem like Phenix has persistent state that would make REST attractive. |
| | 27 | - One advantage of REST would be having Phenix running on a different machine (cluster). Rare case. |
| | 28 | - Scooter discussed DASK for parallelizing Python numpy/pandas jobs on a cluster like wynton, or on local cores. |
| | 29 | |
| | 30 | * Slow downloads of ChimeraX |
| | 31 | - Elaine and TomG have 15-30 minute ChimeraX download times this week from home. |
| | 32 | - Both using sonic.net, but have fast connection to other sites. |
| | 33 | - Tom Ferrin and Scooter (comcast) had fast download times this week. |
| | 34 | - Greg (sonic) and Zach (comcast) have not tried yet. |
| | 35 | - Scooter notes entire UCSF Mission Bay campus has only 10 Gbit/sec internet connection. |
| | 36 | |
| | 37 | * Wynton BeeGFS problem |
| | 38 | - Last week multiple day downtime was caused by BeeGFS file system bug, metadata server deadlock. |
| | 39 | - Caused by a single job from Schoichet lab using BeeGFS for scratch. |
| | 40 | - This job has been changed to use local disk scratch like all other jobs. |
| | 41 | - Giving BeeGFS vendor test case so they can patch it. |
| | 42 | |
| | 43 | * New Wynton GPUs |
| | 44 | - 13 new Wynton GPU nodes just received, 4 A40 GPUs each (48 GB each). |
| | 45 | - Brings Wynton GPUs from about 150 to 200. |
| | 46 | - Nvidia announced new H100 GPU with 80 GB. Also RTX 4000 series expected in September 2022. |
| | 47 | |