MidPlus: A Centre of Excellence for Computational Science, Engineering and Mathematics

我们建议建立一个计算科学、工程和数学卓越中心（MidPlus），服务于从伦敦到中部地区的M1/M6走廊，最初是基于四所在跨机构合作、工业伙伴关系和计算研究方面具有杰出资质的顶尖大学：华威大学、伯明翰大学、诺丁汉大学和玛丽女王大学。我们把重点放在这个地区，因为地理位置接近极大地促进了与工业伙伴的联系和持续互动，特别是对中小企业而言。MidPlus地理位置优越，可为英国汽车、航空航天、生物医学、材料和创意产业的许多组织提供服务。随着MidPlus的发展，我们将把这种伙伴关系扩展到这些公司和其他大学。该卓越中心将建立在300万英镑的电子基础设施初始投资（160万英镑来自EPSRC呼叫和140万英镑来自合作伙伴大学），将提供：*高性能计算（HPC）通过能力集群（华威；2700核心，infiniBand，一些GPU和大内存SMP节点）与华威现有的集群（委托5/2011）相结合，创建一个6000核心集群，从而最大限度地扩大大规模并行作业的范围；以及高吞吐量集群（QMUL， 2900核），以方便需要多次运行以跨越大参数空间的项目。*数据存储和归档设施（镜像在伯明翰和诺丁汉的数据完整性），使中期和长期存储的研究数据（最初~ 1pb容量），和管理结构，使基于元数据的搜索和检索与安全实施范围用户指定的隐私级别。从长远来看，我们将：扩展数据存储的容量；开发一种自动数据老化协议，将数据依次迁移到适当的长期存储技术；扩展我们支持的第2层HPC架构的范围；并在数据和计算硬件之间发展更大的集成和更快的区域网络连接。我们的集体研究专业知识以及管理和使用e-Infrastructure的掌握对MidPlus的成功至关重要，就像我们将安装的设备一样。因此，我们将在电子基础设施之上建立一个智力上层建筑，这将：*积极促进跨学科和机构边界的合作；*提供一个协调的系统及行政支援队伍，让具备现有专业知识的业界使用这些设施，以应付内部对电脑设施的高峰需求，或作为自行建立电脑设施的替代方案；*提供专业知识基础，以培育电子基建的新工业用途；*制订策略性架构，以维持和发展区内的电子基建。这种智能上层建筑将使MidPlus能够提供比单纯的电子基础设施或工业云计算服务更有价值的服务。

We propose to establish a Centre of Excellence for Computational Science, Engineering and Mathematics (MidPlus) that serves the M1/M6 corridor from London to the Midlands, initially based on four leading universities with outstanding credentials for cross-institutional collaboration, industrial partnership, and computational research: Warwick, Birmingham, Nottingham and Queen Mary. We focus on this region because geographical proximity greatly facilitates outreach and ongoing interactions with industrial partners-especially for SMEs. MidPlus is well located to serve many organisations within the UK's automotive, aerospace, biomedical, materials and creative industries. We will extend this partnership to such companies, and other Universities, as MidPlus develops.This Centre of Excellence will be established with an initial investment in e-Infrastructure of £3M (£1.6M from this EPSRC call and £1.4M from the partner Universities) that will provide:* High performance Computing (HPC) through a capability cluster (Warwick; 2700 cores, infiniBand, some GPU and large-memory SMP nodes) to be combined with Warwick's existing cluster (commissioned 05/2011) to create a 6000 core cluster and so maximise scope for large massively parallel jobs; and a high throughput cluster (QMUL, 2900 cores) to facilitate projects that require multiple runs to span large parameter spaces.* Data storage and archive facilities (mirrored at Birmingham and Nottingham for data integrity) to enable mid- and long-term storage of research data (initially ~1 PB capacity), and the management structures to enable metadata-based search and retrieval with secure implementation of a range of user-specified levels of privacy.In the longer term we will: extend the capacity of the data store; develop an automated data-aging protocol to migrate data, successively, to appropriate longer-term storage technologies; extend the range of tier-2 HPC architecture we support; and develop greater integration of, and faster regional network connections between, the data and compute hardware.Our collective research expertise and mastery of managing and using e-Infrastructure is as crucial to the success of MidPlus as is the equipment we will install. We will therefore build an intellectual superstructure on top of the e-Infrastructure that will:* actively promote collaborations that cross disciplinary and institutional boundaries;* provide a coordinated systems and administrative support team to enable industries with existing expertise to use these facilities-either to deal with the peaks in their internal demand for computer facilities or as an alternative to establishing their own;* provide an expertise-base to nurture new industrial use of this e-Infrastructure;* create a strategic framework within which to sustain and develop the regional e-Infrastructure.This intellectual superstructure will enable MidPlus to offer services that add much more value than could be obtained from the bare e-Infrastructure or, indeed, from industrial cloud computing services.