HPSS Tests -- Krzysztof Genser presented a follow up report on the HPSS tests which have been done. He discussed what they have learned about how the system works and how it is parameterized. Given their latest knowledge of the bandwidths of the various components, and the numbers of times each byte gets moved in the system in order to end up on tape (approximately 30 - 20 to get to the disk cache and out again and another 10 from there to tape), they are doing some reconfiguring, rather than simply straight replacement of the 580s by the F50s. One F50 is now in the system, in a special capacity, serving disks only.
Krzysztof will give another interim status on the tests next week. The plan is to complete the tests and give a full report in 2 weeks time.
HPSS Requirements -- Don Petravick presented a highly summarized version of the 12-page Requirements document which he and IBM have been working on jointly, mainly through phone conferences.
There is a need to really boil this down to the essential things which are missing from HPSS and which we cannot live without. The current document is an attempt to specify a broader range of needs (given some base assumptions about what HPSS already provides) - many of those requirements may not be 'allocated' to HPSS as the system component to satisfy and implement them. Given the IBM involvement the original goal needs to co-exist, with the goal of distilling out the essential missing HPSS features. Don requests experiment input to do this. A group of readers/reviewers will be convenved, ASAP. Dane to find a time. Lee, Mike and Vicky + Wyatt all willing to participate from D0.