Minutes of the November 17, 1998 Meeting of MONARC Site and Network Architecture Working Group Attending: Luciano Barone, Joel Butler, Paulo Capiluppi, Mauro Dameri, Michael Diesburg, Irwin Gaines, Phillipe Galvez,Jukka Klem, Lamberto Luminari, John Marraffino, Ian McArthur, Harvey Newman, Vivian Odell, Steve O'Neale, Monica Pepe, Laura Perini, Les Robertson, Krzysztof Sliwa, Tim Smith, David Williams,and John Womersley We basically covered the announced agenda. Part I: Review the task description for completeness; Discussion of subtask 1: Survey of existing computing architectures The schedule calls for a first report by January 1. We agreed that this task was was not that relevant because most of the older experiments were on such a small scale compared to the LHC that one could not extrapolate from them. It was felt that it was more productive to spend time on subtask 2. The conclusion was to carry out this exercise by providing a brief table describing the data analysis experience from LEP experiments and the FNAL RUN I experiments. Discussion of subtask 2: Survey of planned computing architectures It was agreed to concentrate on experiments that will start soon, will have very large volumes of data, high rates, and especially those using new technologies and approaches to analysis and data access. We will include in the list of experiments those that are running or will soon run at CERN, such as NA48 and NA45, which have large amounts of data, at least by current standards, and are being used to try out new models of computing that are hoped to be relevant to future experiments with even larger datasets. We will attempt to collect as much information as possible and then to summarize the information in tables. It was emphasized that we we should look for meta-information that is mainly concerned with access methods since these are elements that are relevant to modelling. The specific hardware implementations, while interesting, will change a great deal between now and the LHC startup. We will review the list of experiments to include in the survey, which now is BABAR, BELLE, CDF, D0, HERA-B, RHIC experiments and the CERN fixed target experiments. There was a brief discussion about the relevance of HERA-B, which we will leave on the list for now. The PEP also mentions summarizing the plans of the LHC experiments. We discussed what this means in view of the fact that these plans are just being developed and no final architecture is yet defined. We agreed to focus on a few parameters such as data volumes and data rates and just try to reformulate the existing information in a more concise form. Discussion of subtask 3: Survey of potential regional centers and proposed architectures There was a lot of discussion on this point. It will present a challenging mix of technical and non-technical issues for us. One question is whether we should identify potential major sites and begin to ask them what they could provide. Or should we just develop a list of likely requirements independent of such considerations? It was accepted that there might be several kinds of regional (or national centers) each providing different kinds of services and facilities and serving a different constituency. It is also understood that political considerations will play a role in the number and location of these various centers. We finally agreed that it would be a good idea to begin to sound out potential major centers so that we can get real information on the scale of resources that they might provide. This will help make sure that models which involve regional centers stay in contact with reality. It was stressed that we need excellent communication and mutual feedback with the Analysis working group and the Modelling working group, especially in connection with this subtask. We need a suggestion on how to insure this. We discussed the role of the CERN central facility to help understand what level of resources CERN would be expecting from outside sources. While CERN clearly expects to provide basic computing for production reconstruction and data storage, they clearly do not expect to have the resources to support all of the analysis work out of their budget. It was also noted that some outside groups will want to contribute resources to be located at CERN - a model that has worked in the past. This is a kind of externally funded "regional center" at CERN. We acknowledged that we would probably have to spend several meetings discussing all these issues. Discussion of subtask 4: Technology evaluation and cost tracking We heard that the CERN IT Division was reviving its own technology tracking efforts. It was not clear how quickly this would happen. We decided to prepare a list of important issues and to present it to them. They will presumably react by agreeing to some of the investigations and if there are any left that we feel strongly about we will try to do them ourselves with help from local experts. Discussion of subtask 5: Network performance and cost tracking This is obviously a difficult topic since it is very hard to predict how networking, especially international and transoceanic networking, will evolve. CERN has a network technology tracking team, which is now in abeyance. This will obviously be a major topic for future meetings. We agreed to start with the work of Harvey's study for the ICFA Networking Task Force. Part II: Review the milestones. Do the milestones as currently stated reflect our current understanding of priorities? These should be considered in the light of recent LCB comments; The schedule calls for a report on subtask I by January 1, 1999; reports and diagrams of planned computing architectures by April 1, 1999; and reports and diagrams of architectures and potential hardware and manpower deployment for candidate regional centers, also by April 1, 1999 The conclusion here is that we will try to meet all three of these. However, we will 'reduce' the scope of the first one in the manner discussed above. Part III: Determine who is available/willing to work on the various subtasks. Determine who is willing to take responsibility for seeing that each deliverable is produced; and We will make a worklist based on the discussion of the meeting and ask people to volunteer by EMAIL for the various tasks. Then, we will turn the result into a workplan. We did get some people to volunteer at the meeting: Steve O'Neale -- summary of BABAR computing at SLAC (st2) UK National Resource (st3) Luiciano Barone -- regional centers (st3) Michael Diesburg -- CDF and D0 Run I experience (st1) CDF and D0 Run II plans (st2) Vivian Odell -- Experience of FNAL fixed target experiments with large datasets (st1) Harvey Newman -- Technology evaluation (st4) Tim Smith/Les Robertson -- CERN experiments (st2) CERN central architecture (st3) John Womersley, Vivian Odell -- FNAL plans for LHC (st3) Jukka Klem/Veikko Karimaki -- regional centers (st3) esp possibility of a Finnish regional center Part IV: Make various organizational decisions -- how often we should meet and when; set up the various mailing lists and web pages we need, etc. We decided to meet via videoconference on Dec 1 and Dec 15, at 16:00 MET. We will try to have an announced agenda. We will set up a web page for this task of MONARC and get it linked into the main page. At least for now, the web page is at http://www.fnal.gov/projects/monarc/task2/ We will also try to set up a mailing list at CERN for this working group.