HBP/EGI Management Meeting

Tuesday 2 Feb 2016, 11:00 → 12:00 Europe/Amsterdam

Pilot status, next steps and service delivery planning

Go to Webex (passwd: 2016) https://egi.webex.com/egi/j.php?MTID=md37c00de78a474001f83333f7df3910a

Slides 20160201_hbp_status_update.pptx

Present: Sean Hill (SH), Tiziana Ferarri (TF), Sy Holsinger (SY), Matthew Viljoen (MV)

MV presented his overview of testing status.  SH said he'd tried out the load-balanced ATLASViewer instances deployed on EGI infrastructure and said that the performance was as good - if not better - than that running on GPFS/supercomputer at HBP.

It was agreed that the next steps (before the end of March) should be as follows :
- validation of the pilot against bigger datasets and enable a federated access of datasets
SH said that a 4TB rat brain was available for deploying onto EGI
- interoperability with HBP AAI
It was noted that Jeff Muller had said in the previous HBP/EGI collaboration meeting that the OpenID Connect release that would enable interoperation with EGI won't be released until the end of Feb.  Given only a month from this point to the end of March, if there are problems with integration, AAI won't be able to be included in the pilot - this was acknowledged.

- Acceptance criteria was discussed and agreed for the pilot:
. Speed is for the imaging software to deliver 8 tiles under a second, with support for multiple (>10) users via load-balancing the imaging software
. Ability to scale to larger datasets (4TB). Ideally federated access.
. AAI compatibility with HBP
. Full documentation of the pilot setup, including testing

- The costing model was discussed.  HBP need a costing for the following:
. 10 TB production running of hosting data accessible by visualization software over the next 12 months.  Potentially more, to be added on an agile basis, to meet rapidly changing requirements.
. Backup datasets on EGI infrastructure as the primary backup and/or cost of staging datasets to EGI for primary access
. Scale the serving of data to 100 concurrent users, using measurements made from the pilot.

11:00 → 11:10 Brief oveview of status of tests

Speaker: Mr Matthew Viljoen (EGI.eu)

11:10 → 11:30 Next steps: objectives and timeline for a fully functioning pilot by May

How do we know that the testing of HBP software on EGI is a success, and what are the expectations for a "fully functioning" pilot? - ability to establish new repositories quickly. What are the requirements? - what are the services expected? E.g. (1) a resilient visualization/imaging service deployed on EGI infrastructure (resilient here means deployed at >1 site) offering the capability of dealing with multiple concurrent users (e.g <10 users, depending on software limitations) - capability of dealing with larger dataset sizes (what amount of storage collectively? we currently only have 100GB, we would like to demonstrate that 1TB and more is achievable) and to transfer and replicate these datasets geographically? - authenticated access via an integration with the HBP OpenID Connect service - service performance levels

11:30 → 11:50 Discussion of cost models and project activities

- services, amounts, timeline - EGI service level agreements including . description of guaranteed services . minimum availability/reliability, monthly reporting of performance . service hours and support channels (collective SLA for all federated providers) - EGI offered models: combination of free at point of use, pay for use (one provider), free at point of use with participation to project activities for the recovery of operational/technical support work dedicated to HBP - EGI interested providers (high throughput computing, data management, cloud IaaS, data management expertise/development): DE- GeoGRID (GWDG), Juelich, DESY, FI - CSC, FR- France CNRS, GR- GRNET, IT- INFN Bari, PL- CYFRONET.