Notes of MoBrain CC teleconference, 10/May/2016

 

Participants:

- Alexandre Bonvin (chair)

- Antonio Rosato

- Alberto Redolfi

- Laura del Cano, Jesus Cuenca

- Jerome Revillard

- Gergely Sipos (notes)

 

Discussion points:

1. Cryo-EM in the cloud: bringing clouds to the data:

* Deployed Federated Cloud VM of Scipion installed. Image not yet public. Once finalised will be shared in AppDB.

* Image requires very large capacity: 265 GB memory and 64 cores. This was deployed at Surfsara site (direct deployment) and at CESNET (FedCloud deployment).

* Typical input for a Scipion run is in TB range. Vision: Have a VM that can be deployed on sites where big data is available, or can be staged in within reasonable time. We expect rather static setup because the run requires large resource allocation.  

* STFC could be another site where the VM would be used. What's the progress of setting up cloud site? (They committed in the MoBrain OLA)

* Currently testing solution for data upload/download. - Globus Connect/Online is tested. Planning to attend EGI Webinar on Open Data Platform which could be another solution.
* Testing solution for remote visualisation on the cloud - x2Go, Guacamole. Developing an automated (script-based) installation for Guacamole.
* To learn more about dynamic allocation of HTC resources we are planning to attend EGI Webinar on DIRAC

* The deliverable relating to this CC activity will be due in Month 21 (November 2016): D6.14-Scipion cloud deployment for MoBrai. Until this date it's enough to report about progress during our regular CC teleconference meetings. 

 

2. GPU portals for biomolecular simulations

* 2 portals. Both are also use cases in the INDIGO-Datacloud project, where portals are integrated with VM management

* AMBER portal: Recently was working on optimising the protocol to select most suitable computational architecture (CPU vs GPU vs Cloud). Currently setting up the production portal. GROMACS portal can also choose target platform in an optimised way.

* GPU infrastructure: Trying to get access to other GPU sites, so to use not only Florence. Marco could provide update on this (he was not present)

* No problem foreseen for reaching the related deliverable.

 

3. Integrated portal:

* Development was finished. However, the portal is not available at the moment. Jerome needs to check the VM in the FedCloud site. Setup monitoring and restart mechanism for this?

* Migrate the NeuGrid applications into the CERNVMFS server to move applications to the VO. Migration is done, now the server need to be activated to the VO.

* Authentication layer in Neugrid Web services has been replaced by Keycloak (http://keycloak.jboss.org/). This needs to be integrated now into the Neugrid Liferay portal. This would allow external applications to use the Neugrid web services without X509 certificates, but with e.g. OpenID, OAUTH. Also, this could allow the use of the same authentication mechanism on the Web portal and in the Web Services. At the moment this won't impact other activities in MoBrain.

* At the moment we still continue to use the 'old' X509 based authentication in the WeNMR portals, but in the long-term we should transform to something more sustainable.

* Gergely to try to arrange a webinar about the EGI AAI proxy service so CC members could see what's coming in terms of AAI services from EGI, and position/focus internal CC AAI activities appropriately.

 

4. User support, training, other events

* INSTRUCT course took place in Utrecht, Alexandre will circulate a report about this for EGI news channels

* Planning WestLife-MoBrain workshop contribution to iNEXT project meeting in second half of October, close to Madrid

* DI4R (Digital Infrastructures for Research) Conference: http://www.egi.eu/news-and-media/newsfeed/news_2016_011.html. Maybe co-locate a MoBrain tutorial with this to attact/engage with Polish users? (Contact for local organiser: Ɓukasz Dutka <lukasz.dutka@cyfronet.pl>)

* Gergely to circulate the EGI Training Plan Milestone document to the CC list, and CC members can contribute to (with training topics/modules, and with specific events targeted/considered for training)