Operations/Resource Allocation

From EGIWiki
< Operations
Revision as of 14:02, 7 September 2012 by Dobrisa (talk | contribs)
Jump to: navigation, search
Main EGI.eu operations services Support Documentation Tools Activities Performance Technology Catch-all Services Resource Allocation Security

DEADLINE for reply: 07 September 2012

NGI Resource Allocation Policy and applicable VOs Does the NGI own resources? Comments
Example: NGI_XY 10% of the resources in each NGI site is allocated to Life Science VOs. This amount increases to 20% in site XY. All other disciplines access the remaining 80% of the provided capacity. The NGI owns a pool of 500 cores and 10 TB to be allocated to VOs according to the NGI priorities. The policy expresses a recommendation but is not enforced at site level.
NGI_AEGIS VOs supported by the AEGIS sites: national VO (aegis), High Energy Physics VOs (atlas, cms), regional and projects specific VOs (see, seegrid, sgdemo, ops.vo.egee-see.org, meteo.see-grid-sci.eu, seismo.see-grid-sci.eu, env.see-grid-sci.eu, desktopgrid.vo.edges-grid.eu, vo.aginfra.eu). 30% of the resources is allocated to the national VO, 50% to regional and projects specific VOs and 20% to High Energy Physics VOs. The NGI does not own any resources (resources are owned by the sites).
NGI_CH Supported VOs by the Swiss EGI sites: ATLAS, CMS, LHCB + SMSCG project VOs (LTPC, CRYPTO, ECON, LIFE, EARTH, TUTOR). Cores: 2144 cores, ATLAS 58%, CMS 32%, LHCB 8%, SMSCG 1.4%. Storage: 2082 TB, ATLAS 50.6%, CMS 48.4%, SMSCG VOs 1.1% The NGI does not own any resources. SMSCG cores are unprivileged i.e. they are not dedicated at all times to the SMSCG VOs. A minor part of the ATLAS cores (330 cores) is also unprivileged.
NGI_CZ VOs supported on sites are mostly defined by sites themselves (NGI doesn't own these resources), based on involvement of their users (for example LHC VOs or Auger), with small fraction usable by other VOs promoted by NGI (VOCE, Auger) On resources provided by NGI we support mostly smaller VOs, either backed up by some local group of users or VOs which are in early stage of building. Yes, thery are used mostly as seed resources for new VOs.
NGI_GE Taking into account still low demand there is no need in Resource Allocation Policy. The NGI owns 12 cores and 1 TB storage resource. According to the project submitted to the National Science Foundation we are planning to upgrade existing infrastructure and attract new users, after that we will implement Resource Allocation Policy.
NGI_HR Resources integrated into EGI infrastructure are currently allocated to HEP, biomed and regional (VOCE) VOs. We use fair share model for allocating all resources. Requests from new VOs will be evaluated by the NGI board. NGI owns resources and enforces policies.
NGI_HU 5% allocated to the Hungrid VO; the rest depends on the sites, mainly used by LHC VOs. Resources are owned by the sites.
NGI_NL The Dutch NGI BiGGrid pledges resources to a few international VOs like ATLAS, LHCb and ALICE and LOFAR. Apart from those VOs, BiGGrid supports a number of smaller, mostly national, VOs. Typically, individual users and user groups apply for resources (computational, storage, support) to BiGGrid for the period of one year. If awarded, the users are often added to one of the national VOs that fits with their type of research, for example, a life scientist will be added to the life sciences grid VO lsgrid. If no VO fits, a new VO will be created. The awarding of proposals is done by BiGGrid in close cooperation with the sites. In this process, it is also determined how much a site will contribute to the resource request. Here we mention that the organisational infrastructure of BiGGrid consists to a large extend out of site representatives. yes
NGI_PL Resources can be allocated only for research done in association with Polish science, which includes international VOs with Polish participation. Resource Allocation is done based on SLAs negotiated with national VO representatives (usually Polish member of this VO). The SLA specifies allocation at sites. In case when new national VO representative need to be established they should contact NGI_PL through either GGUS or Helpdesk@plgrid.pl. The resources are owned by sites, however the NGI coordinates and has some influence on allocation in a pool of 25 000 cores and 1 PB of storage space.
NGI_SK Resource allocation depends on sites. 75% of computing resources and 96% of storage capacity is managed by sites dedicated to LHC VOs. Resources are owned by sites.
NGI_TR 80% of the resources in TR-03-METU site is allocated to CMS VO, 80% of the resources in TR-10-ULAKBIM site is allocated to ATLAS VO, 80% of the resources in TR-01-ULAKBIM site is allocated to local VOs. All other disciplines except these can access the remaining 20% of the provided capacities in related sites. The NGI owns a pool of 12000 cores and 1.2 PB to be allocated to VOs according to the NGI priorities. NGI_TR infrastructure is supported by Ministry of Development this year and will be supported in the next three years.