HPSS Tests performed 3/21/23 - 3/23/23

Sent data from all 24 EVBs using rcfBlast(1 sender from each of 4 disks from every node), 9.4GB files each copies of the other.    Event rates vs hours since beginning of MDC, averaged over 10 minute periods:

The lines correspond to periods where 3,6, 12, 18, & 24 EVBs were included, and the behavior scales very well as ~180 MB/sec per EVB, containing 2x1GB bonded links.   The overhead is approximately 10% and scales well.

The reason that the bulk of the time the DAQ was reading out ~4khz was that two EVBs were operating at poor efficiency.   EVB03 was off until the last hour.   EVB07 was operating, but one of its slave ethernet links had been autonegotiated to 100MB/s speed cutting the speed in half.   The later was fixed simply by removing the slave from the link and restoring it using "ifenslave".   The rate autonegotiation worked properly the second time.