The London Tier 2
 |
 |
|||
 |
 |
 |
- Brunel University
- Imperial College London, (Group Information)
- Queen Mary
- Royal Holloway
Group Information - University College London
Group Information
| Number of Running Jobs In London | |
The structure of LT2 was described for the GridPP2 Proposal as part of the Tier 2 Planning Document:
"The colleges of London University plus Brunel make a natural grouping from which to form a distributed Tier-2 Centre. Not only are they geographically close, they are well connected via the London Metropolitan Area Network. The benefits of working together and sharing resources were recognised before the 2001 SRIF round when each institution signed the Memorandum of Understanding (MoU) of the London e-Science Consortium. In which the partners agreed to "make resources obtained through the 2001 SRIF exercise available to each other and to collaborate on the development of new e-science technologies". Each of these institutions subsequently allocated SRIF money to improving their research computing facilities. The HEP groups at these institutions are involved in a wide spectrum of experiments, all of which will, as far as is reasonable, be supported by the Tier 2 Centre.
The London e-Science Centre (LeSC) is the core e-Science Programme regional centre for London and is based at Imperial College. The LeSC, in collaboration with the Imperial College HEP group will play a central role in the operation of the London Tier 2. UCL has also been awarded e-Science centre of excellence status (in networked systems) and will play a key role in e-Science network sector.
Brunel were one of the founder members of the London e-Science Consortium and are members of CMS, BaBar and MICE. A significant computing resource is currently under construction with a combination of SRIF1 and HEFCE funds. This will provide 64 dual-processor hyperthreading Xeon nodes plus a 1TB SCSI disk array. PPARC rolling programme grant funds are providing an additional 2TB of IDE disk store. Currently the Grid interface is being tested using a dedicated cluster with eight worker nodes being commissioned. Brunel HEP and distributed computing expertise is being used to integrate the Brunel resources into the distributed London Tier-2 centre. The intention is to provide a further 128 dual-processor nodes, making 192 in total, to enhance the performance of the Brunel cluster. This additional resource has University approval for SRIF2 funding and HEP will receive a considerable share of these resources.
The Imperial College part of the Tier 2 Centre will be a collaboration between the LeSC and the HEP group. Both groups are very active in Grid computing and have collaborated since the start of the e-Science programmes. Both groups have in the past benefited from JREI and JIF as well as the 2001 SRIF funding. Imperial College allocated £3M to e-Science from the 2001 SRIF spending, of this ~£1M is being spent on networking and buildings and ~£2M on equipment which will be provide resources mainly for HEP and two other user communities. This equipment is being purchased in a rolling programme to ensure maximum utility and value for money. The first machines arrived in September 2002 and since then have been used to generate Monte Carlo simulated data for D0, CMS, LHCb, UKDMC and MICE (generating well over 2.5TB of data for D0 alone). The HEP group and LeSC will be part of a consortium bidding for further resources in the next round.
The QMUL HEP group plays a lead role within the e-Science SRIF project at QMUL. This project received £1.2M of SRIF funds to provide infrastructure, including a modern high capacity machine room, high speed networking as well as clusters of commodity computers. The machine room is nearly complete and a 16 CPU cluster is currently being prototyping and the necessary software infrastructure set up to facilitate easy upgrades and management. The aim is to provide an interface to the existing DataGrid gateway as well as well as providing a more general purpose web-service based portal. The intention is to expand the prototype to about 70 CPUs and 2TB of disk storage for Spring 2003 and anticipate having 200-300 CPUs and a commensurate amount of storage available by Sspring 2004. In addition there is an existing 80 CPU farm which is mostly used for BaBar simulations. A recent further award of £630k from SRIF2 will further enhance these facilities.
RHUL has benefited from JREI, JIF and SRIF awards in the last few years. They are operating a farm of 80 CPUs and 0.5 TB file server which is heavily used for official BaBar simulation and local data analysis jobs. They are in the process of procuring equipment with their part of the London e-science SRIF award and are currently using a small test bed to prototype the SRIF facility and get experience with Grid software which is also used for local ATLAS simulation and analysis. They expect to build a facility with around 100 CPU and a few TB of disk storage.
The UCL part of the Tier-2 centre will be provided as part of the College wide e-Science Centre. UCL has over the last year formed a Grid executive and users board which oversees the deployment of the £1M SRIF-1 funding. UCL is also an e-Science Center of Excellence (CoE) in networked systems, recognising the existing strong collaboration between HEP, Computer Science, Electrical Engineering and the college Education and Information Systems division. Through the UCL e-Science Centre & CoE they are purchasing and installing a ~200 node Grid compute cluster with a significant amount of network attached storage. This will become part of the UK level-2 e-Science Grid as well as part of theEDG production Grid. It will benefit from direct Gbit/s network connections to the peering point with the London MAN. This facility will form the core of the existing resources, part of which can be offered as part of a Tier-2 centre.
In addition UCL also
- Have a 5-CPU Beowulf cluster running Globus 2 and 3 within Computer Science.
- Have a pilot EDGidPP testbed node in HEP.
- Will soon be commissioning a SGI Onyx2 visualisation supercomputer and are constructing a Virtual Reality-plus-AG suite in Chemistry.
- Have an access Grid node under construction.
In terms of SRIF-2, they are preparing a bid for substantial sum to augment the Centre facilities by at least a factor of 2-3 as demand dictates."
The management structure proposed for LT2 was:
"The distributed Tier 2 Centre will need to define its operational goals (by taking input from its constituent communities) and then ensure that agreed goals are implemented in a timely and affective manner. These two objectives will be achieved through:
- Management and Policy Group (MPG): Responsible for setting and policy decisions and resolving any issues concerning fractional usage, acceptable usage policy etc. within the Tier-2. The MPG will be meet quarterly to review compliance and to decide policies for the future including constitutional issues such as groups joining and leaving the Tier-2. The MPG will also decide and review the location of posts funded by GridPP2. While the MPG will normally reach a consensus, in the event of a vote the weight given to each institute will be according to the amount of funding provided by that institute into the Tier 2.
- Technical and Support Group (TSG) Responsible for the technical co-ordination between the different institutes within the Tier-2 so as to enable the effective operation of the Tier-2 Centre as defined by the MPG. Their remit is to ensure sufficient support throughout the Tier-2 to resolve technical issues, provide user liaison training and support, etc. In particular the Chair of the TSG will be employed by the Tier-2 in a support role and as well as co-ordinating the support structure this person is ultimately responsible for the successful resolution of user generated problem reports."
The Tier-2 Expression of Interest from the London Tier 2 institutions to GridPP2 will be made available here in .doc and PDF formats.
LT2 Management Board
LT2 Technical and Support Group
Last modified Thu 28 January 2010 . View page history
Switch to HTTPS . Website Help . Print View . Built with GridSite 1.4.3