1 / 7

Summary of Processes

Summary of Processes. Rutt offline.recon – memory bound COMSOL – memory bound mcDESPOT and quantitative mapping – CPU bound, voxelwise parallel Zeineh offline.recon ImageJ – memory bound ASHS – CPU bound McNab FSL, bedpost – CPU bound, voxelwise parallel FreeSurfer – serial

helki
Download Presentation

Summary of Processes

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Summary of Processes • Rutt • offline.recon– memory bound • COMSOL – memory bound • mcDESPOT and quantitative mapping – CPU bound, voxelwise parallel • Zeineh • offline.recon • ImageJ– memory bound • ASHS – CPU bound • McNab • FSL, bedpost – CPU bound, voxelwise parallel • FreeSurfer – serial • Probabilistic ICA for resting state fMRI • Matlab diffusion simulations – memory intensive

  2. Summary of Load • Most describe their load as “peaky” • Development time and coming up with new ideas dominates • However, Michael would like to start up larger studies within 5 yrs. • 30-50 subjects • Planning on routine thalamic segmentation?

  3. Summary: Hardware • Needs • High memory – at least 64GB, Michael wants 196GB • Unified data storage and backup • Wants • Prioritize CPU speed over cores as many jobs are not parallelized • In general, having enough memory to make the problem feasible at all is most important • Faster CPU is just an incremental improvement

  4. Summary: Software • Needs • DICOM connectivity • How do we get this without OSX/OsiriX? • Scheduling and job submission • Matlab, FSL, FreeSurfer, etc. • Wants • Remote desktop • Esp. for viewing larger datasets (ImageJ or Trackvis) • Not sure if this is an appropriate use of the system • Would change the hardware of the terminal node

  5. Concerns • Jennifer looking for a solution ASAP • I stated end of Jan purchase date, which she was okay with • Administration • This could turn out to be a fairly demanding job esp. for many users • Sharing and resource management • Need to develop a policy for sharing compute time • At MGH/Martinos, they have one machine that’s a free for all and others must be reserved/scheduled

  6. Hardware Proposal • If there is some urgency, we should focus on expandability: • Get something that meets our core needs now • But that can be expanded as users begin to use it more or better understand what their jobs need • Terminal node • Simple barebones system with minimal CPU power and RAM • Functions as the liaison between user and cluster, manages job scheduling & load distribution • Used for testing programs and scripts for compatibility • Could possibly have remote desktop, but would then need a lot more RAM • Data server • Minimal CPU power and RAM • RAID – 6 (req. 4 drives), 60 (req. 8 drives, more fault recovery) • Compute nodes • 64GB RAM with as many slots as possible • CPU? • Ideally diskless • Need fast (10Gbps) connection to data server

  7. Stanford Services and Rates • IT • Consulting for planning– free? • Basic administration – $3.6k/yr • Secured housing for rack, 10Gbps connection to campus backbone – $3k/yr • Backup storage – $4-5k/yr per 10TB • Med School, IRT • Basic administration – $2.4k/yr • Secured housing for rack – $4.8k/yr • $840/yr per 500GB

More Related