Michigan: M. Abolins
Argonne: R.Blair
UCL: R.Cranfield
CERN: D. Francia, K.Korcyl, P.Golonka
NIKHEF : J.Vermeulen
1. Paper modelling: F. Wickens has provided new information on the mapping of the calorimeter on the ROBs. M. Abolins will look at the effect of combining ROLs in view of possible deferral scenarios. The results generated by M. Abolins shortly after the meeting of 7 January are available via links in the notes on the meeting. J. Vermeulen still has to implement the "mini-scan" in the paper model program. A telephone meeting with C.Bee, S.Tapprogge, V.Vercesi and F. Wickens, planned for Friday 31 January, was expected to provide answers to questions with respect to the parameters (conclusions of that meeting: menus presented by S.Tapprogge in December to be used without the 5 kHz "other" item. Three scenarios are to be studied: low luminosity (2.10^33), low luminosity with B-physics trigger, design luminosity (10^34). The event building rate scales with the LVL1 accept rate, 1 kHz per 20 kHz accept rate. MU20 trigger at low luminosity: TRT scan not meaningful, a mini-scan in eta-phi region of about 1 by 1 makes sense. 2 MU6 at low luminosity: both need to be confirmed, no scan. The LVL2 algorithm execution times are increased with 75%, 100%, 50% and 100% for feature extraction from the calorimeter, muon, TRT and SCT/Pixels data, respectively. The use of dual 4 GHz CPU boxes is assumed for the LVL2 farm, a 70 % utilization is to be assumed for computing the number required.). Paper model parameters and results should be included in chapter 2 of the TDR, J. Vermeulen intended to distribute a draft via the modelling list before Friday February 7, the deadline for submission.
2. ROS modelling: R. Cranfield has had contact with P. Golonka on the similarities between ROS and L2PU models. The document on the ROS model parameters has to be updated and then passed to P. Golonka for inclusion in the document of K. Korcyl and P. Golonka on the Data Collection model parameters.
3. Data Collection modelling, HLT modelling, at2sim: P. Golonka
and K. Korcyl have updated the draft
document on the DC model parameters. A specification of simple testbed
setups to be used for model validation is now also included. Measurement
results of individual components and as far as available are also included.
K. Korcyl mentioned problems with understanding the scaling of results
from the measurements on the DFM done on 1 GHz and on 2.4 GHz machines.
J. Vermeulen mentioned that different processors (Pentium III, resp. Xeon)
processors are used in these machines, so that results probably will not
scale with the processor speed. It was concluded that model parameters
should be based on results obtained with the 2.4 GHz machines. Results
for the L2PU and LVL2 Supervisor were not yet available, D. Francis is
going to look after this. at2sim has been checked into the general
TDAQ CVS repository with tags at2sim_1_05, and at2sim_1_10 for the latest
version (not yet completely stable at the time of the meeting). The code
can also be browsed
from the web. P. Golonka can be contacted if write access to the repository
is needed.
The work on modelling the testbeds and the full system with at2sim
is not progressing completely in line with earlier plans and doubt was
raised whether at the time of the March T/DAQ week the required results
would be available. See for a further discussion below under "Workplan".
4. Simdaq: J.Vermeulen is trying to set up with simdaq a full model the candidate architecture with the generic models available in the program. So far the complex switch system could be set up succesfully, the next step consists of implementing SFIs (event building is done in the current model in the processors of the Event Filter farm). Next the models described in the document on modelling parameters could be set up.
5. Chiba City: No news could be reported, but a short discussion took place on whether getting the Data Flow software to run on the system (256 engines connected via fast Ethernet) would be worthwhile. Conclusion: only when it is easy to do, D. Francis suggested to wait for the next release of the Data Flow software for finding out and to ask R.Hauser for assistance.
6. Workplan: It was planned that by the end of January the final parameters should be available and the results of a small scale testbed model should have been verified. This is delayed by about one week. Next the full model, including LVL2 sequential processing and correct mapping of the ROBs should be running by the end of February. It seemed doubtful that this will be possible, in the discussion it was mentioned that delay until the March T/DAQ workshop is a more likely. This is undesirable, as at this time results should have been produced and analysed. It was decided that P.Golonka would produce a list of work items, to be discussed in the DataFlow meeting of 3 February and that R.Cranfield would provide assistance starting at February 10.
7. Next meeting: Probably Tuesday 18 February 15.30 h CERN time.
8. AOB: An abstract on computer modelling for RT2003 has been submitted, K.Korcyl will check the guidelines for author names. K.Korcyl has had a first look at a note on possible bottlenecks and will review comments from J. Vermeulen. P. Golonka probably takes care of a presentation on modelling in the ATLAS week (from Cracow). J.Vermeulen has to contribute or organise contributions on modelling to the TDR, there is a deadline for Friday 7 February (10 February at the latest), J.Vermeulen intends to circulate text for chapters 2 and 15.
J. Vermeulen, 7 February 2002