-
Notifications
You must be signed in to change notification settings - Fork 1
September 1, 2017, Friday
Liya Wang edited this page Sep 4, 2017
·
6 revisions
- Create a branch for archiving back to Data Store
- Mainly for supporting RNAseq and Chipseq workflows
- Use Agave for archiving (no pipelining)
- Tacc will resolve the issue for not looking for local copy?
- Is there a way to tell Agave to not to move the data to TACC?
- We need to set a cron job if using pipelining't
- Use Agave for authentication?
- So user can run with their own user name
- Limit to public app and users own private app
- How to authenticate to both Agave and CyVerse?
- When loading workflow, need to check permission on jobs and results? Need to re-design database for storing workflow? Maybe its a bad idea to store them?
- Can always give permission to maizecode for all jobs so user can load workflow on SciApps?
- Remove 'public' and data (top menu) to reduce dependency on local servers
- For CSHL, could we put data on wildcat, then mount to halcott, brie, brie7?
- Would data page browsing be slow? or will it break hd5ai?
- Would computing be slow?
- This will allow us to process Sorghum data
- For CSHL, could we put data on wildcat, then mount to halcott, brie, brie7?
- May put it on production if the paper gets out
- Need to change output url with Agave links (folder needs to be public through Agave)
- Need to sort out how to power Shinny app and BioDalliance
- Very hard to do since they are both reading the data on server directly
- Might be possible to copy the data to server (wget? inside shinny script?)
- Background
- Converting to user centric is hard (user submitting job instead of maizecode)
- Need to figure out how to authenticate with Agave
- Need to rely on public apps and systems, need to figure out the permission on jobs
- Easy to run Stampede and access users data
- Easy to put on Atmosphere/Jetstream
- Supports user private apps easily
- Federation version is good for workshops
- No queue issue
- Data load very fast
- Easy share
- Converting to user centric is hard (user submitting job instead of maizecode)