Minutes SLAC-ALTAS TDAQ Meeting 02-Nov-2006
-------------------------------------------

Present: Stephen, Matthias, Steffen, Su Dong, Sarah (phone), Amedeo,
         Rainer

Sarah:

IPC server crash.  No realistic mix of MC.  Not enough minimum-bias
available to run L1.

Partition Maker creates proxies at node level ok.  Rack level still
not quite there.  Haven't found the best way of doing that.

Right now supporting just three levels.  In that setup we are
considering to put more than one node to the top in order to balance
the tree.

Will ask about the freedom to add more layers if possible to leave
open the possibility of optimizing bandwidth.

Flushing the cache: right now plan to do that on every unconfigure.
The proxy will abort if the cache is full.  To keep tests predictable,
and not depend on the history of the testing, want to flush regularly.
Considering to implement a mechanism that will reuse the cache by
throwing out old items.


Stephen:

DetCommon pretty stable now.  Low rate of changes.  Working on adding
more regression tests for HLT builds.  Tests for the nightly builds
that people can add.  People defining slices expected to contribute.

Question about how to test steering.  Is using ATN ("at night"
system), runs as part of the nightly release build.  Only feasible for
small CPU times. RTT (run time tests) is more suitable for heavy CPU
tests.  Can use a data-file.  Uses SQLite files with "nominal"
conditions.

System fills a web-page with results.  Also sends e-mails.
AtlasHLT -> TDAQ-common