Michigan: M. Abolins
Argonne: R. Blair
UCL: R.Cranfield
Cracow: K. Korcyl
NIKHEF : J. Vermeulen
CERN: S.Wheeler
1. Paper modelling: An overview of the current knowledge on the parameter values is available from the modelling web page. The algorithm execution times are not yet complete: m_95 values (for 95 % of the events the execution time is shorter than the m_95 value) for the Pixel and TRT scans and for analysis of SCT data for the B-physics triggers are still missing. S. Tapprogge will be reminded (by J.Vermeulen) that these values are missing. The overview contains some results from the latest version of the spreadsheet, which is also available from the web page. The next step to be taken is to update the document with the results.
2. ROS modelling: There were no new developments to report. J. Vermeulen mentioned the measurement results for a PC-based ROS using network I/O, which are not understood, and suggested that the model possibly could provide a clue. R. Cranfield is planning to have a look at it
3. Data Collection modelling and 4. HLT modelling: S.Wheeler reporterted that S.Gadomski has obtained preliminary measurement results on the DFM (for a non-optimized version of the software). Further measurement to be done will be discussed next week. The document from S.Wheeler on the test plan will be made available from the Data Collection web pages.
There is no information available from P. Golonka on the status of the L2PU model.
K. Korcyl reported that Asgeir Bjorlykke is working on multi-layer models of switches in conjunction with the measurements done in bdg. 513. He also reported that the strawman architecture model now works. The long startup time (of 6 hours) observed earlier was caused by the way in which the routing tables were set up. It has been reduced to a few minutes. K.Korcyl first wants to understand the queueing behaviour observed before making the model available.
5. Ptolemy:Ê The implementation of sequential processing has to be reviewed. The DFM model has to be changed slightly, other models may also have to be changed. R.Cranfield, K. Korcyl and S.Wheeler plan to discuss this in the next T/DAQ week.
6. Simdaq: Nothing to report. The program will have to be used for checking the paper model results.
7. "Chiba City": The Data Collection software has to run under kernel version 2.4, this could make it possible to run it also on the Chiba City system. R.Blair will have a look at this.
8. Workplan and status in relation to workplan: Earlier on the day of the meeting G.Lehmann had reported the following: "this mail is to inform you that the Phase 2A integration between DataCollection, Online Software and ROS has been succesfully completed. A testbed has been setup in bdg 40 with 2 ROSes, 2 L2PUs, 1L2SV, 1 DFM and 2 SFIs running all on different hosts. The system runs at about 1 kHz RoI request rate for ROS fragments of 1 kB (100% RoI request rate and a 5% accept rate). There are still DEBUG printouts and the code has not been optimized, so that these performance numbers are just very preliminary." It is not clear whether modelling of this system would be of interest, J. Vermeulen suggested to discuss this in the next meeting on the basis of information to be obtained beforehand.
9. Next meeting: The next meeting is planned for Wednesday 24 April at CERN, with possibility for phone connection. J.Vermeulen will ask F.Wickens for a 1 1/2 - 2 h time slot during that day.
10. AOB: R.Cranfield: the implications of the work of the Connect forum for modelling has to be thought about. We will have to come back to this in the next meeting.
J. Vermeulen, 28 March 2002