Minutes of January 7, 1999 meeting of the MONARC Architecture Working Group ---------------------------------------------------------------------------- (N.B. The next meeting will be on Tuesday, January 19, 1999 at 16:00 MET.) We continued the discussion of regional centers. There are two (at least) goals at this time: first to get a 'framework' for characterizing regional centers which would include, but also would differentiate, the various 'tiers' that are possible; and second to begin to identify the scale of the largest, or so-called 'tier1', centers. The point of departure for the first goal is the 'framework' of questions developed at the Dec. 1 meeting and which is posted on the web (and was circulated by EMAIL just before the meeting). The answers to the questions are intended to provide a good description of the capabilities and capacities of a center. It is generally agreed that we should continue to develop this framework which is based on services and facilities. It was also agreed that we should not make specific recommendations but rather should take the framework to various sites and ask them how they think their plans for their facilities will map onto the LHC analysis problem. Luciano Barone agreed to consolidate the existing writings on regional centers and to collect comments on the current 'framework' from the collaboration. Luciano will send out a request for comments early next week and will prepare a summary for the next meeting. On the second point, Harvey strongly suggested that we start with one example of a 'tier1' center. There have been several attempts to characterize such a center. Our working definition is that it provides a wide range of services and has an aggregate capacity that is of order of 10% or more of the main center at CERN. This will allow us to establish the scale of a regional center in terms of specific services and hardware facilities. We all agreed that it was useful to approach a few specific candidate tier1 sites and that it was NOT useful to make any hard recommendations or set any specific requirements at this point but to simply see what we could learn from the discussions with the sites. Our goal is to have, by early February, a short document that contains 1) some introductory remarks about regional centers, their role in the analysis, and their relation to the central site at CERN; 2) The framework for characterizing regional centers in general; and 3) an agreed on 'scale' for the 'tier1' centers. The group has been interested in hearing a report on how CERN is planning to approach the LHC computing problem. There are many uncertainties and severe funding problems. Les Robertson will make an initial presentation of CERN's assessment of the magnitude of the resources needed to satisfy the requirements of the experiments at the meeting after the next one (which would be at the end of January or very early in Feruary). At an earlier meeting, the Fermilab members of this group were asked to begin to discuss the scale of a possible center at Fermilab. This discussion has in fact been taking place with the FNAL Computing Division and has been 'facilitated' by the Directorate. The Computing Division would like to present an outline of its thoughts on what a 'tier1' regional center at FNAL might look like. Although this does not constitute a committment and funding issues are not yet resolved, it will be an indication of what is being considered and it will provide the basis for a modelling exercise of an analysis approach involving CERN and a large regional center. Since Matthias Kasemann, the head of the FNAL Computing Division, will be at CERN during the next meeting of this working group on January 19, the FNAL members proposed that he present this outline to the group. This proposal was accepted. The next topic was a brief reminder that we have a report due in two weeks. Because of the winter break, there has been little time to make any progress since the Dec 15 meeting which covered the status of the information gathering. I will get in touch with all the information gathering team early next week. I will also try to discuss the report with the leaders of the other working groups as promised at the Dec 15 meeting. To summarize, the work plan for the next two months is to: 1) Complete the report on the survey of architectures of current experiments (end of January); 2) To finish the development of the 'framework for characterizing regional centers; and 3) To develop at least one example of a tier1 regional to set the scale and for input to the modelling exercise. Finally, we discussed the problems with scheduling the biweekly meetings. Despite trying to check with people and believing that I had responses from both ATLAS and CMS people, I managed to schedule today's meeting at a poor time for our ATLAS members. To avoid this in the future, we agreed to stick as closely as possible to the Tuesday 16:00 MET time that we agreed on. Some of us will not be able to make some of the meetings but we will have to find ways to have colleagues cover for us. The next meeting will be on Tuesday, January 19, 1999 at 16:00 MET. Respectfully submitted, Joel Butler