Requirements for SA2 and JRA4 20/04/2004 JRA 4 Meeting: ========================= What does SA2 expect from JRA4? Dcncl lesson brd Service level agreement. ---- * Advise on the definition of the service clases TSA2.2.4 and the SL5 TSA2.3.3 in order to end up in a network requirements model * To provide some tools for the measurement of SLA parameters in an E2E ???? GN2 presentation (Nicolas Simar) ---- Nicolas is using the request and response interfaces from the NMWG. Geant Premium IP service: Path finder is based open IS-IS cost metric. The main difference between the path finder for Premium IP: that finds the reals path while the monitoring path finder finds monitoring boxes associated with the path. Nicolas: are there analogous schema to the network ones for CPU and so on? I would guess so but it would be good to find out. stats.geant.net ... but you have to log in Premium IP reservation service ---- At the moment you can only reserve for a day at minimum. Not the few hours that might be required in EGEE. Mark's presentation of GGF stuff ---- Are the GGF covering looking at collaborative applications? Answer: difficult isn't it. ---- So after lunch there will be an SA2 brainstorming session but it will be just a few people sitting in a room. SA2/JRA4 provisioning group (informal meeting) ============================================== Actors = clients 1) Workload scheduler 2) Data management (perhaps just servants of 1 and 3) `- i.e. replica management 3) Applications/ |- in terms of BW | `- Dave Foster |- Biomedical | `- should provide good use cases `- High performance computing? |- DEISA `- visualisation 4) Operations tools including diagnostic tools So all the people volunteered to look for requirements will work together. JRA4 planning ============= OK so we should use a particular modelling system: for example UML. www.cs.ucl.ac.uk/research/grs Grid App | Grid Middleware | MD-BB ---- MD-BB ---- MD-BB | .--' | `--. | SD-RT SD-RT SD-RT | | | Network Network Network Element Element Element Face to Face meeting in end of June. Nicolas: one day per activity. 21/04/2004 JRA1/JRA4 integration ================================ Anand, Nicolas, Sophie Nicoud, Alberto, Pete, Me JRA1 integration cluster has local people but also representatives from other software groups and other JRAs. This is just an interface with integration. Skills for integration representative: familiarity with intregration techniques. But also specific technical knowledge for the components. Languages: Perl: ALIEN is in Perl but it is a proof of concept. Perl is very complicated with respect to dependancies. Javier: for network monitoring Dante is quite advanced. Nicolas mostly working on interfaces (web ones). Did not do much investigation of the best platforms. The target is to support any platforms. However, for the two years, JRA1 want to support Redhat Enterprise v3 and Windows XP. For the latter stuff, may be compiled against Cygwin byt with MS VC++. For Java: version 1.4.2. C++: gcc3.2.3. Perl: 5.8.0. This stops us having to ship the compilers with the software. For libraries any shared libraries must be the same for all software groups. From the automatically generated page you will get a list of all the dependancies. From Savannah, you get project information. Need to get a CERN AFS account to access CVS. Bug tracking: EDG bugs are to difficult to transfer to Savannah. LCG and EGEE bugs will also not be merged. For making modules there is a list of top level directories that must exist. Dante will be developing two separate parts: private stuff for Dante and other stuff with EGEE. Configuration and error handling must be common throughout the project. Configuration in particular. Naming conventions for modules is already there. Modules must be as small as possible: if there is a change in the module, then the whole thing must be rereleased. Remember: we must communicate. If there is something to be discussed there are plenty people in integration to talk to. Also, the bug tracking system will be used for this. 21/04/2004 JRA4 Requirements meeting ==================================== Why do the grid people want to monitor data? Monitoring for middleware is easy to understand. Middleware requirements gathering will start later. Applications: HEP |- CERN (David Foster) `- NA4 Biomedical `- NA4 (Vincent ? IN2P3 Physics) Generic `- NA4 Todo: Find out the basic mechanisms of file transfer and so on. Return to RGMA. It's Web Services in the current implementations.