Level 3 Online Status Reports

Issue 3
Monday, April 12, 1999

Event Flow

Tony and Jeff modified l3_cs_socket to keep message headers and handshaking messages in network byte order, thus facilitating the transfer of messages between little-endian PC's and bit-endian SGI's. The modified library has been tested for PC-PC communication, but not PC-SGI yet.
The "Level 3 Control Kludge" works and is included in the Control Kludge tar.gz file. The Web documentation has not been updated yet.
The TDC test was repeated in lieu of having the test wedge, and was run over two days.
Frank reports that he has run stand-alone readout on the test wedge which lasted for days, sending data to the software event builder. The evidence thus points to the communication between the ADMEM crate and the hardware event builder.

Send data from the PC output node to SGI consumer server.
Ilya will remove the unneeded l3_request_taker from the control framework.
Jeff will add 6-word command functionality and another fake VRB object (to read fragments from an internal buffer) to the SCPU.
Ilya will run as large a system as possible, with maximum number of SCPU's, converters, processors, etc., and take rate measurements.

Christoph and Tony got the "Hello World" executable to work within the Level 3 control framework.

Andreas will add to the fake data generator: fragment size variations; other features for use in EVB/L3 tests.
Christoph will add CPU profiling to reformatter code.

ILU has been installed on Linux and Solaris.
Ilya has tested ILU invocations on a single node (gobi), including sequences of aggregates.
Steve has tried to port ILU to VxWorks. Host table lookups need modification. There is also concern over the 3MB ILU libraries, which are on the large side for the VME CPU's.
Sasha continues to investigate ORBacus and its possible use with ILU.

Ilya will confirm ILU operation between different computers and platforms.
Steve will try to selectively link or strip ILU down for a C server to see how large it turns out to be.
Sasha will investigate ORBacus/ILU interoperability.
Ilya, Jeff, and Steve will write IDL's for Scanner Control and Level 3 Control.

Ivan and Andreas have found, when working with DIM, that all connections are successfully made between the server, client, and nameserver, but then the system stops. At this point the nameserver needs to be killed (!) in order to let the client and server actually talk with one another.
DIM works fine when all the processes run on one host "connecting" to itself.

Jeff Tseng / MIT / jtseng@fnal.gov