MONARC plenary meeting, 5th February 1999

Participants

around the table
H. Newman, L. Perini, C. Grandi, I. Legrand, L. Luminari, U. Gasparini, L. Servoli, J. Bunn, E. Arderiu, M. Dameri, V. Karimaki, M. Schuller, T. Sasaki, K. Amako, I. Williers, K. Holtman, D. Williams, P. Galvez, S. O'Neale, J.Shiers, T.Smith
in videoconference
L. Barone, W. Jank, I. Gaines, M. Campanella, S. Resconi, V. O'Dell, P. Capiluppi, C. Vistoli, J. Gordon, S. Bernardi, R. Purdes, K. Sliwa, M. Michelotto, A. Bagliesi

Rough Sizing Estimates for a Minimal Regional, T. Smith
Tim's slides summarized the main points of discussion following Les presentation at the previous Architecture WG. The need of other potential RC producing their own estimations was stressed. The Objectivity overhead and the tape role require specific investigation.

Reports from the WG Chairs

Architecture WG, V. O'Dell
The survey of the existing architectures is reported to be almost completed: it shall be available for next WG meeting.
Feedback to Luciano Barone's document is asked for.

Analysis WG, P. Capiluppi
A short summary of the last WG meeting is given, see his minutes. A major open point is the Data Model: how the data are structured in Objectivity, where they are, how to access them.
Work is going on in Perugia in exploring the effects of varying the parameters within their ranges in the kind of model presented in Paolo's table: first results are expected for the next Analysis WG meeting on the 24th.
After Paolo's report Harvey suggested a comparison also between two different scenarios for the final phase of the analysis: centralized at RC against distributed on desktop resources ( 1TB disk and >100 Specint95 is felt normal for a 2005 desktop ).

Simulation WG, K. Sliwa
Krysztof spoke from US and the audio was very disturbed. He was asked to send via e-mail the points of his reports he wants circulated and recorded, besides what will be already taken up in the Iosif's presentation, forthcoming in this meeting.

Testbed WG, L. Luminari
A contact person for every testbed site is needed: the local setting up is proceeding and the list of contact person is almost completed.
CERN is acquiring a SUN quadriprocessor in the same kind of configuration in use for CMS and ATLAS work-group servers: it is going to start with 100GB disk and will be used also for simulation developments.
Harvey reported that from his discussions with Les a configuration of the CERN testbed for 100MB/s I/O should be feasible.
The options for populating the DB were reviewed: the tag DB from Julian Bunn is already available at CERN, with a simple reading code; CMSIM outputs will follow from Caltech. ATLFAST++ could be available for first simple access around end-february.
Contacts are going on ( Mauro Campanella and Cristina Vistoli ) with the TF-TANT/QUANTUM Working Group for networking tests, where MONARC has been proposed as an application to stress and verify the functionality of the QoS techniques. The reactions have been very positive, and the next step from Monarc side should be to define the QoS features on which to focuse, as well as our possible participation in the setting up of the experimental network required for the tests.
Harvey stressed the importance of the common work Monarc/Tf-Tant and encouraged Mauro and Cristina to continue and arrive to a definite proposition for the network tests. Everybody agreed.

Report on GIOD, J. Bunn

Work on Simulation Models (PS PDF), I. Legrand
In the discussion which follows the presentation various issues are raised:

The need for a DB server model
The importance to be given to the tape layer ( feelings are mixed, but everyone agrees in keeping assumptions very simple, also to avoid too many parameters)
The need of validitating each step of the simulation separately is stressed ( David Williams ), and the of monitoring the number of jobs accumulating in the queues in the model and the growing response time is seen as a useful tool at this effect ( Irwin Gaines ).
The importance of the background on the WAN ( J. Bunn )
The importance of the scheduling policy ( Harvey )

Discussion of I2-DSI (and paper), H. Newman

Upcoming Events and Schedule, L. Perini

The next milestones ( March 99 ) were reviewed: the starting of the measurements on most of the testbed seem a realistic possibility; the completing of a first technical run of simulation seem also within reach, but the LHC model involved has to be kept extremely simple.

For the next LCB of March 10th no written material is due. The Monarc referees will report on the status of the project. Laura contacted Lothar Bauerdick and some information exchange will be arranged before the LCB meeting.

The next plenary is foreseen in the 10 days after LCB. Harvey will send around a proposition taking into account the constraint of CMS week. Harvey also proposed to have around end-february a first estended steering group meeting ( with the representatives of possible RC ).

Around mid-May a 1.5-2. day Monarc workshop is foreseen. The goal is to prepare the progress report of 23rd June with the detailed workplan of Phase-2, and first ideas for a continuation of Monarc in an implementation phase ( phase-3). In the discussion there was a general agreement on the opportunity for Monarc to continue, for implementing and focusing the tools developped in the first year and provide a common ground for shaping the computing infrastructure. The degree of commonality has to be widely discussed in the experiments, and their feedback is essential for a useful phase-3. The proposal for the dates will soon be circulated by Harvey.

5th February 1999