RA policy survey

From EGIWiki
Jump to: navigation, search
Main EGI.eu operations services Support Documentation Tools Activities Performance Technology Catch-all Services Resource Allocation Security


EGI Resource Allocation menu:




Local polices regarding EGI Resource Allocation process

Why

On July OMB (see minutes) Resource Allocation team got an action to contact NGIs to find out the situation of the local policies in order to understand obstacles in taking part in EGI Resource Allocation process. The process goal is to put EGI Customers and Resource Providers closer together in order to reach agreement on using resources. It does not mean collecting resources by EGI.eu and claiming management rights over it.

Scope

Whenever we refer to "resource allocation" we mean allocation of resources integrated with EGI.

The survey is meant to be filled out by both NGI and site managers.

Survey

Identification - Site/NGI Name, Contact (Name/Surname)

Q1: What is the resource allocation policy document at your NGI/site and who defines it? (document specifying who can be supported, to what amount of resources, etc.)

Q2: Are there any specific requirements regarding who (what scientific domain etc.) can apply for resources? (including foreign scientists)

Q3: Are there any requirements for accounting of used resources e.g. acknowledgements in publications.

Q4: Who is in charge to make decisions about resource allocation requests at your site/NGI? (Site Manager)

Q5: What system your site/NGI uses for handling resource allocation requests? (face to face/paper/web/other)

Q6: What is the frequency of submiting/handling the requests and how long does it take to handle it? (on-demand/weekly/monthly/yearly/other)

Q7: Is it possible to share current resource allocation at your site/NGI? (i.e. how many core/CPU hours are given to customers)

Q8: What main obstacles do you see to include your resources in EGI Resource Allocation process? https://wiki.egi.eu/wiki/Resource_Allocation


NGI Name Contact (Name/Surname) Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8
NGI_AEGIS Antun Balaz Q1: There is no single allocation policy document. Sites regulate their allocations individually. Q2: No. Q3: Yes, acknowledgment to NGI is requested. Q4: Site manager(s). Q5: Face to face/e-mail. Q6: Weakly to yearly. Q7: Yes, smaller fractions. Q8: Funding for providing of additional resources to share, since the current ones are mostly dedicated.
NGI_ARMGRID Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_BA Mihajlo Savic Q1: No central policy, sites regulate independently. Q2: No. Q3: No. Q4: Site manager. Q5: Face/e-mail. Q6: On demand. Q7: No. Q8: Existing resources are dedicated. Currently there is no local funding for expansion of resources.
NGI_BG Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_BY Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_CH Sigve Haug Q1: There is no single allocation policy document. Sites regulate their allocations individually. Q2: No. Q3: No. Q4: The site managers. Q5: Face to face. Q6: Weekly to yearly. Q7: Smaller fractions can be shared, O(10%) Q8: Current resources are dedicated. Without scientific partnership or payment, usage from outside will hardly get accepted.
NGI_CHINA Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_CYGRID Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_CZ Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_DE Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_FI Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_FRANCE Jerome Pansanel Q1: A draft of single allocation policy document is available, but it is not complete. Therefore, Sites regulate their allocations individually. Q2: No. Q3: Yes, acknowledgment to NGI is requested. Q4: The site managers. Q5: Face to face or e-mail Q6: On-demand. Q7: Sites are currently not publishing the share. It may be feasible using the CECapability Glue tag. Q8: Current resources are dedicated. Without scientific partnership or payment, usage from outside will hardly get accepted.
NGI_GE Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_GRNET Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_HR Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_HU Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_IBERGRID Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_IL Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_IT Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_MARGI Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_MD Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_ME Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_NDGF Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_NL Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_PL Marcin Radecki Q1: There is no single document for the NGI. Sites have freedom in deciding which request to support. Their decision is mainly based on value of scientific outcome promised in the application. Q2: Sites are multidisciplinary. Resources can be used by people affiliated with a Polish research or education organization. In practice if a foreign researcher want to use our resource he/she needs to find a collaborator sufficiently affiliated. Q3: NGI_PL requires acknowledgements in scientific publications. Q4: Site manager. Some allocations can be done by PL-Grid operator. Q5: Web system developed by PL-Grid. Q6: On-demand. For simple requests they are handled within one day. For milion-hour requests up to two weeks. Q7: Yes it is. Q8: No obstacles. We already take part.
NGI_RO Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_SI Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_SK Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_TR Q1: Since sites are managed by the operation center of NGI_TR, operation center is regulates the allocations. Q2: No. Q3: Yes, acknowledgement to NGI_TR is requested. Q4: Site managers Q5: face to face and email Q6: on-demand Q7: Smaller fractions can be shared(10%) Q8: Since the resources are dedicated, funding is expected for providing of additional resources to share.
NGI_UA Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
NGI_UK John Gordon Q1: There is no published document. Policy is recorded in various management committee minutes. 
Current policy is that GridPP is prepared to allocate up to 10% of resources for non-WLCG use. Most of this is used by non-WLCG particle physics experiments supported by STFC but any other VO can request support and will usually be given a nominal allocation with the expectation that they will make use of unused resources with no guarantees. NGI_UK has a record of providing significant resources by this route. Individual resource providers in the UK might be prepared to enter into bipartite agreements with VOs with EGI providing the introductions.  

Q2:It should be scientific research or pre-competitive commercial pilots. Q3:All usage is accounted like any other and so identifiable. NGI_UK does not currently define such requirements, however the resource provider GridPP has a citation policy https://www.gridpp.ac.uk/acknowledging.html. Q4:NGI Operations Manager with advice from the NGI Management Board, the sites involved, and the Project Management Board of GridPP – the biggest user community and owner of the majority of resources.   Q5:Small level requests are dealt with via email. Any VO that begins to use a significant fraction of resources must submit requests in writing to a review board. Q6:On-demand. Usually handled weekly but no published guarantees. Allocations are reviewed quarterly. Q7:Yes. NGI has overall 10% target but no hard allocations. CPU is managed on a fair-share basis. Disk is more problematic - for small requests joint pools are used, but beyond a certain level the data management process requires space tokens to be considered for the VO. Q8:Our understanding of the EGI requirement is that we would have to provide a guaranteed allocation (irrespective of usage) that EGI can broker usage over and fully decide on who gets what. The UK allocation policy outlined in Q1 ensures (a) that allocated resources can be efficiently used; (b) that the science is aligned with UK interests; and (c) that we are in a position to handle periods of high demand from our principle clients, the LHC experiments. Our funding agency would not support any other approach. 
AfricaArabia/ZA-UJ Simon Connell Q1: There is no general resource allocation policy document. Q2: None at present Q3: Nothing explicit. Q4: Simon Connell Q5: e-mail / personal contacts Q6: on-demand, weekly Q7: All VOs use one common Grid infrastructure. Faire-share policy is in effect with priority given to existing VO's Q8: We are already well saturated. ~90% of our resources are fully used by supported VO's and the rest are allocated to local users
AsiaPacific Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
CERN Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
IDGF Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
ROC_CANADA Di Qing Q1: There is not central policy, and site defines it individually. Q2: No. Q3: No Q4: Site Manager Q5: face to face/email
Q6: Not clear Q7: No Q8: The resources are dedicated.
ROC_LA Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Russia Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:


Site Name Contact (Name/Surname) Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8
DESY-HH / NGI_DE Andreas Gellrich Q1: There is know general resource allocation policy document but sites
have commitments according to MoUs with the VOs.
Q2: No, but is intended that the VOs which demand resources fit the
scientific portfolio of the site.
Q3: Nothing explicit. Q4: Site manager Q5: e-mail / personal contacts Q6: on-demand Q7: All VOs use one common Grid infrastrucuture. Typically 2/3 are
guaranteed to MoU partners.
Q8: We are already well saturated.
UK typical John Gordon Q1:Individual UK sites usually have additional resources

not pledged to GridPP and they have freedom to make these available to whomsoever they wish subject to local policies, in the grid, cloud, or outside.  In most cases this is done for VOs with which the site has some affiliation or synergy.

Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8:
Q1: Q2: Q3: Q4: Q5: Q6: Q7: Q8: