Personal tools
You are here: Home User Support DEISA Primer

1 Introduction

1.1 Project Presentation

DEISA, the Distributed European Infrastructure for Supercomputing Applications, is a consortium of leading national Supercomputing centres that aims at fostering the pan-European world-leading computational science research.

DEISA deploys and operates a persistent, production quality, distributed supercomputing environment with continental scope. It aims at delivering a turnkey operational solution for a future European HPC eco­system. And by extending the European collaborative environment in the area of supercomputing, DEISA is paving the way towards the deployment and operation of a persistent cooperative European HPC ecosystem, as suggested by ESFRI.

DEISA2, funded by the European Commission in FP7, continues to develop and support the pan-European distributed high performance computing infrastructure established since 2002 within DEISA, the EU-funded predecessor project. The DEISA infrastructure is currently based on a tight coupling of eleven national supercomputing centres (DEISA principal partners) from seven European countries, using dedicated network interconnections of GÉANT2 and the NRENs. Since 2008 four associate partners (dashed lines in the figure) have joined DEISA being in the process of integrating additional ressources into the DEISA HPC infrastructure.

The DEISA consortium is currently consolidating the existing HPC infrastructure and services. Activities and services relevant for Applications Enabling, Operations, and Technologies are continued and enhanced, as these are indispensable for the effective support of computational sciences in the area of supercomputing. In addition, DEISA has extending the service provisioning model towards the inclusion of non-localized Science Communities. Accordingly, collaborative activities are expedited with further European and international initiatives.

DEISA principal and associate partners

1: DEISA principal and associate partners

1.2 Infrastructure and Resources

DEISA is operating a heterogeneous HPC infrastructure currently formed by eleven European national supercomputing centres that are tightly interconnected by a dedicated high performance network. The term heterogeneous refers to the variety of HPC system architectures, operating systems, batch schedulers and local file systems provided by the DEISA supercomputing centres and that is typical for such an HPC ecosystem.

DEISA is structured as a layer on top of the national supercomputing services by providing generalized interfaces and services that allow to access and utilize this pool of computing resources in a consistent way and, therefore, more efficiently. That way, the DEISA HPC infrastructure and services combine – for users and user communities – the advantage of having access to a variety of supercomputing architectures for different demanding computing purposes with the advantage provided by consistent interfaces across these different resources and services.

The DEISA supercomputing resources incorporate several different platforms and operating systems: IBM AIX or Linux on Power5 and Power6, IBM Linux on PowerPC, IBM BlueGene/P (Linux), SGI Linux on Itanium, Cray XT (Linux), and NEC vector systems (Super-UX). The DEISA infrastructure's aggregated computing power is several hundreds of Teraflops:

Table 1: Infrastructure and Resources
Site Architecture TF/s1 #Cores2 Memory3
BSC IBM PowerPC 94.2 10,240 20.0
CINECA IBM Power6-575 101.0 5376 20.0
CSC Cray XT4 AMD-Opteron 4-core 32.3 4,048 4.4
Cray XT5 AMD-Opteron 8-core 86.7 6,816 6.1
CSCS Cray XT5 212 22,128 29
ECMWF IBM P5-575+ 33.0 4,552
EPCC Cray XE6 360 44,544 59.4
Cray XT4 114.4 12,288 24
Cray X2 2.9 112 0.9
FZJ BlueGene/ P 1000.0 294,912 144.0
Intel Xeon X5570 QC 207.0 17,664 53.0
HLRS NEC SX-9 19.2 192 6.0
Intel Xeon X5560 62.0 5600 8.4
IDRIS BlueGene/ P 139.0 40,960 20.0
IBM Power6-575 67.3 3,584 10.8
KTH AMD Opteron 2354 86.02 9,800 19.81
62.3 9,728 39.0
RZG BlueGene/ P 54.0 16,384 4.0
IBM Power6-575 120.0 6,560 18.5
SARA IBM Power6-575 62.6 3,328 15.3
Totals All machines 2821.72 508,576 482.61

1.3 Classification of the DEISA Production Services

DEISA operates a set of services, called the DEISA Production Services, to allow all users to achieve their scientific objectives conveniently and efficiently.

Not all of these services are available on all platforms, due to technical limitations or local policy restrictions. However, to have both a global coherency and relative flexibility across all its  distributed infrastructure, DEISA has defined a three-level classification of all its services, according to the service level targeted. They are respectively named the "Core", "Additional" and "Optional" DEISA Services.

  • The Core Services are implemented and supported at each site, where technically possible. They are highly robust and will persist in the long term. DEISA provides a high support level for these Core Services.
  • The Additional Services are also supported at every DEISA site when technically possible, but are services for which not the same level of reliability and long term availability can be guaranteed as for the Core Services. They are supported on a best effort basis with a medium priority to solve the reported problems.
  • Finally, the Optional Services might be supported on each DEISA platform, but are not guaranteed to be supported in the near future, and have a low priority level of support.

The important Core Services for the users are:

The Additional Services are UNICORE and DESHL (a line command interface and API to UNICORE), which are uniform job and workflow managers which allow to submit jobs on heterogeneous platforms in a common and seamless way.

The Optional Services are the SRB and iRODS data management tools, and the RFT (Reliable File Transfer) tool. For the compute services, WS-GRAM from the Globus toolkit, which allows to submit jobs on heterogeneous platforms in a uniform way, is available on some sites. For the AAA services, the SSH and UNICORE-SSH alternative tools to provide an interactive access to the platforms are available in some places.

[1] - Peak performance of the total system in TeraFlop/s
[2] - Total number of CPU-cores
[3] - Total memory available in TeraByte
Document Actions