Questions and Answers :
Wish list :
Lower redundancy with a smarter transitioner
Message board moderation
Author | Message |
---|---|
Send message Joined: 17 Jul 05 Posts: 102 Credit: 542,016 RAC: 0 |
I think expecting faster results from a high initial replication is a "milk girl calculation", there would be a better method : Let's say quorum = initial replication = 3 A lot more different results will be returned within the first stack of WUs, leaving back (roughly estimated) about 10% with overdue results (deadline expired). But there would already be more finished results in the first stack as the current redundancy goes into more WUs instead. Now make the transitioner a bit smarter so it increases the initial replication value of the WU by one for each overdue result plus increase the priority for the feeder and reduce the deadline for the new sent results. This will cause the feeder to send out two results for one expired result with a high priority and those will trigger the EDF mode very soon. Imo. this would increase the experiment speed by about 1/4th with a reduced overall redundancy. It requires a bit work on the transitioner to teach it what to do when a result times out and it needs the "-priority_order" argument for the feeder. |
©2024 CERN