VDASC meeting 2013/11/18 : Unusual time: 18.00 CET

Please not the unusual time!

Face to face discussion in cascina after the DA session. Remote participation via TeamSpeak.

Agenda

1. storage/network/computing plans from EGO
2. status of the virgoDev->virgoStaging (PackageManagement)

Participants

Pia, Alberto, Cristiano, Antonella, Livo, Stefano, Benoît, Loïc, Alain, Frédérique, Damir, Franco (please add your name)

Minutes

0.) John's presentation on the DA seesion on porting Ligo pipelines to Virgo clusters.

Questions:

a.) Pia: How the LDR server should be set up. John: Easy to deploy either in Bologna or other places, will return the path of the files in Bologna. Necessary for the jobs to find data locally instead of transferring them.

1.) Antonella presented the new network architecture design: LAN architecture overview at Cascina site (  http://www.ego-gw.it/reserved/documents/codifier/download.aspx?Code=EGO-REP-COM-56)

No immediate discussion, participants need time to study in more detail. Implementation will happen in two phase, in two tenders. a.) recabling of all the network, b.) LAN upgrade to new technology. Time scale is end of january 2014.

2.) Livio's presentaiton about storage strategies, requirements and possible choices. Different data access and usage requirements require differenc class of storage solution. The requirements and the amount should come from the analysis groups. Virtualisation for load balancing, solid state disk cache for performance tuning, etc is planned. A testbed has to be set up. Timeline: First step in 2014, second phase in 2015.

Document:  AdVirgo Storage Guidelines ( http://www.ego-gw.it/reserved/documents/codifier/download.aspx?Code=EGO-REP-COM-57 )

Some number: 2 TB/day for the circular buffer.

Questions:

a.) Benoit was asking for a document describing the implementation plan and the exact amounf of storage going to be bought.

b.) Gergo was asking about how different virtualisation techniques can cope with direct hardware access and performance. Will discuss later, the colleague responsible for virtualisation was not present.

c.) Gergo: planned lifetime of hardware ? Usually 3 year

3.) Stefano's presentation about computing upgrade. Proposal to have the same machines and farm for Ctrl  and olservers (with remote display/ketboad/mouse in the control room and Virgo labs).  Would like to have also the same machine for olnodes, but need some input to know if this is possible. In particular, will GPU be used extensively, in which case the olnodes would be different from the olservers.  Upgrade planned in two steps: upgrade of olservers/ctrl in 2014, upgrade of olnodes in 2015 ?

Questions, comments:

a.) Using remote ctrl seems a good idea.

b.) General agreement to have olnodes similar to olservers, but need to check about the possible needs of GPUs.

4.) Short updates about /virgoStaging area (to replace /virgoDev). Will discuss in detail on one of the following meetings.

Decisions to take in short terms:

a.) VSR4 raw data should be deleted ans resources could be used for other purposes. Does anyboy using still using it ? Has to be answered by 18/11/2013.

c.) For various reasons a homogen olserver, olnode configuration would have lot of advantages. There is nothing againts to this plan but has to be declatred for the next VDASC call, 25/11/2013.

Things to think on for the longer term:

a.) Exact hardware needs for the online clusters. For example how many group and how will use GPUs, etc...