Level 3 Online Status Reports

Issue 2
Monday, April 5, 1999

Event Flow

The output node -> consumer server interface was tested with multiple l3_converter programs.
A "Level 3 Control Kludge," analogous to one written for the event builder, is being developed. This will start up l3_converter's and l3_rc_server's on PC's. l3_farm_user (or something else) must then be used to configure and run the processor nodes.
Extended tests with the event builder and Level 3 reading from the 2nd floor TDC crate ran for a day before being manually stopped, accumulating millions of events.
Extended tests with the test wedge and ADMEM crate fail after roughly 6.5 hours, afterwards sending empty events.
The test wedge readout rate was cranked up, but the failure still occurred after roughly 6 hours.
Christoph has also ported the reformatter to VxWorks to run as a stand-alone application without need of the event builder. The same failure mode is seen, so it appears to be a problem with the readout below the event builder.

Jeff will test the consumer server interface with Tony and the real consumer server framework.
Jeff will test the Level 3 control kludge.
Christoph and Ilya will continue robustness tests and wedge readout tests, notifying appropriate people as underlying systems are debugged.

The provided executable was originally linked with an obsolete version of the Level 3 control code.
Christoph has manually verified the data format of the TRYBOS record which the Executable I/O module is presenting to the AC++ modules.
Andreas continues to refine his fake data generator.

The executable needs to be recompiled with updated control code.
Andreas will continue to refine his fake data generator, including fragment size variations.

ORB status from Steve, Sasha, and Ilya:
- ROBIN's IDL compiler is delicate and doesn't implement arrays or sequences of aggregate types.
- ORBacus has no C binding, but the IDL compiler could be extended. However, the ORB and BOA are implemented in C++ with heavy reliance on exceptions; Tornado (VxWorks) doesn't support exceptions yet, so porting the ORB/BOA would be onerous.
- QCorba is no longer supported.
- ILU from Xerox is almost CORBA-compliant and can understand IDL.
Other ideas were thrown around:
- TAO on VxWorks with another IIOP-compliant CORBA implementation on the client computer
- use ORBacus IDL compiler to generate ROBIN stubs
- DIM
- compiled Java RMI (kaffe, Toba, or JIKES)
Level 3 IDL will wait for some of the above dust to settle.
Christoph and Ilya will track database-related issues.

Michael has been working on trying to read and write ROOT Tree's simultaneously. He has a program which reads and writes fake data using the default ROOT buffering.
Andreas and Ivan have begun to investigate DIM v4.2:
- Installed under Linux.
- Graphics utilities had problems (but we don't use these anyway)
- Basic tests indicate that "dead node" notification works, whether the dying node is the server, client, or nameserver.
- Client and server connections to the nameserver work, but client requests to the server have problems.

Jeff Tseng / MIT / jtseng@fnal.gov