...
Step | Description of action | Dependency on 3rd party services (EOSC-hub or other) |
|---|---|---|
UC1 | On submission of a containerised workflow, sufficient resources are provisioned on a remote site(s) to allow execution of that workflow. The users home credentials (or the services services credentials) must be accepted on the remote site(s) | Orchestrator/Kubernetes EGI Fed Cloud/Other cloud Suitable AAI mechanism (note the fusion community does not have a centralised IdP or AAI system, but each site provides its own authentication based on username/password) PROMINENCE cloud execution service |
UC2 | During workflow execution, intermediate files should be written to a location which will be accessible to later stages of the workflow. The final results should also be accessible at the users home institute | EGI FedCloud/B2DROP integration based on either user or service credentials |
UC3 | Data hosted off-site shall be accessible to code using the fusion Unified Data Access (UDA) middleware so that the same code can be used to access data regardless of locality. | B2SAFE with three way replication (preferred sites would be CINECA(for the MARCONI link), STFC (geographicaly close) and PSNC (another fusion site running this service)). Will require suitable AAI mechanism and integration with UDA. |
| UC4 | Any attempt to access data hosted off site should determine whether the data is 'open' or 'embargoed'. In both cases, users should authenticate themselves to allow traceability of who has accessed the data. In cases where the data is embargoed, the hosting site should deny access to unauthorised users. | B2SAFE and embargo periods |
| UC5 | Data placed at an offsite location should be replicated to CINECA and at least one other partner site within the fusion community |
...
This has undergone a paper based assessment and found unsuitable for user needs
INDIGO PaaS Orchestrator
In use. Have also tested Kubernetes which is both easier, more feature complete and more widely used. Lack of integration os Orchestrator with other components both in Indigo and EGI has caused issues, but the developers are very responsive
OneData
Under test
B2SAFE
Under discussion
EGI FedCloud
Heavily used. The real limitation is the access model being imposed. We don't want to 'order' CPU time, we just want to make opportunistic use of it, This would be akin to 'pre-empitble' VMs where the process could be killed if needed for other work. the curreny order based system would mean I would need to ask for 100 cores at each site but I could not guarantee filling them.
Requirements for EOSC-hub
...
Requirement number | Requirement title | Link to Requirement JIRA ticket | Source Use Case | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
Example | EOSC-hub to provide an FTS data transfer service |
| UC1 | ||||||||
RQ1 | Storage requests for WP8.2 Fusion CC | Request for B2SAFE Storage | UC5 | ||||||||
RQ2 |
Capacity Requirements
EOSC-hub services | Amount of requested resources | Time period |
|---|---|---|
| B2SAFE | 10TB at 3 sites (CINECA, STFC, PSNC) | 6 months |
| EGI FedCloud | 10 cores | 6 months irregularly |
| EGI FedCloud | 100 cores | 2 months |
...