Topics can be seen in the Slides. Highlights of the meeting:
Big discussion about performance of workstations, most of the participants were experiencing slowdowns. Apparently nothing bad can be explicitly seen on the servers; need to track down where the performance issue is. This calls for proper monitoring. This might lead to changes in the system: shall we keep having shared folders? Discussion went on possibilities not to use NFS for homes on workstations.
ACTION: Florido has plans for monitoring system by the end of the year. In the meantime everybody is requested to contact him in case of slowdowns so that the issue can be seen live.
ACTION: Florido investigates alternatives to NFS
ACTION: Warn users not to use their home folders for computation on Vespa, but to use scratch areas instead.
Computing Model
ACTION: Division should evaluate if the current computing model suits its needs, or if it's better to have some powerful computing cluster instead of scattered workstations.
Software status
Intel Fortran compiler license should be renewed.
ACTION: Florido and Jon should look into Intel licenses.
ACTION: Division should discuss on the usage of Intel compilers, evaluate how many parallel compilations to buy, evaluate what product to buy (only Fortran?)
Software updates should be scheduled
ACTION: Florido will take care of updates.
It is not clear if some software is still used.
ACTION: Florido should send around a survey to see who is using what.
Old machines and OS update. Which one to dismiss? Which one to renew? How to rollout new operating systems?
ACTION: Florido will try to provide a list of the current machines status.
ACTION: Division/Groups should discuss which machines to keep
ACTION: Division/Groups should understand what library versions they need.
ACTION: Florido will propose a scheduled update, maybe during summer.
Vespa cluster
node update: update should be gradual and keep some node with older libraries.
ACTION: Florido will propose a scheduled update.
batch system: is there a need for it? It seems to me that there is not.
ACTION: Working groups might evaluate the need for that.
Accessing IT services from outside
Too many SSH machines opened to the outside world. Florido proposes one or two machines as gateways. This is not accepted for practical and performance reasons.
ACTION: Florido will setup a single gateway that is suggested for everybody to ssh login from outside matfys. Performance tests will be run on it. If performance is not bad, maybe can plan to buy a big gateway machine.
Proposal: a Department VPN
UPDATE: Fysikum IT meeting in the afternoon reported that it is forbidden to create own VPN. Therefore we should push LDC to use a more linux-friendly VPN client. This effort has already been started by LTH.
Inventory/Monitoring:
There is no way of monitoring performance of machines. There is no record of who-bought-what
ACTION: Florido is investigating a monitoring system.
ACTION: Florido and Katarina will do inventory and discuss how to propagate this information. For the moment being even an excel file would do. This can be produced from some of the documents in the drift page.
Website: Would be nice to have a clear list of responsibilities.