e-BioGrid E-Science infrastructure for life

science research

E-BioGrid workshop, November 10, 2011

Introduction

Big Grid infrastructure

High-performance computing and storage resources,

Grid – Life Science Grid

High Performance Cloud

Hadoop (soon)

Non-Big Grid but available via NWO/NCF:

Huygens, LISA, GPU

How to apply for access?

All applications go into the NWO/NCF IRIS system. Basically: specify what you need, why And give an estimate how much you need. All applications go to Big Grid when Big Grid infrastructure is involved. For Huygens and Lisa, NWO/NCF will consider all applications

Please contact us for assistance. We are working to simplify this procedure.

What does SARA do?

SARA doesn’t own the systems. SARA doesn’t decide on user access. SARA doesn’t host websites, small scale services, wiki’s etc. SARA maintains HPC systems, provides assistance and consultancy. SARA works in scientific projects like eBioGrid to assist in the use of HPC hardware and e-science environments. SARA works on parallelization projects (funded by NWO): parallelize your software for use on HPC hardware.

The Life Science Grid

WHAT IT IS

•Several clusters at academic medical centers

and universities.

•Open to all life scientists based in the

Netherlands.

•Clusters can be used separately or together

by using Grid middleware.

•Currently more than 3700 job slots for

concurrent processing.

•Distributive storage capabilities and

automatic replication of data.

•Maintained and supported by SARA.

The Life Science Grid sites

LUMC – Leiden WUR – Wageningen UU – Utrecht NKI – Amsterdam AMC – Amsterdam Radboud Universiteit – Nijmegen ErasmusMC – Rotterdam Technische Universiteit - Delft Rijksuniversiteit Groningen Keygene - Wageningen Each of these clusters have either 16 or 32 cores and 18 Tb of storage space.

Grid nodes

Currently 6 small clusters :16 job slots each, soon to be

upgraded

5 clusters: 32 job slots each

32 job slot each- 20TB storage

All clusters run CentOS 5– 64 bits

Gina cluster SARA 2800 job slots

Philips 1470

Nikhef 2160

Groningen 164

High Performance Cloud

19 nodes of 32 cores each (608 cores)

Each node has 256 GB RAM

400 TB storage.

Nodes are connected by four 10 Gb interfaces to a non-blocking switch.

So maximum 40 GB access to storage (per node).

High Performance Cloud

When to use cloud?

Start up your own Virtual Machines. Be your own boss. Users have their own Virtual Cluster, running their OS and software of choice.

Expand when you need to. You must be able to organize system administration yourself.

Huygens National Supercomputer

Huygens is an IBM pSeries 575 system. The number of IBM Power6 cores is 3456. The total peak performance is 65 Tflop/s. The total memory is 15.75 TB. The inter-node InfiniBand interconnect is 160 Gbit/s. The net disk space is 700 TB. The system runs Linux (SLES11 SP1)

Huygens: when to use

Often used for large scale simulations

Parallel systems with interdependencies and a need for high speed interconnects.

Software typically uses MPI or openMP.

Parallel jobs communicate heavily with each other.

Existing sofware is often ported in NWO funded

parallelization projects.

LISA: Description

Total number of cores 4480

Total amount of memory12 TB

Total peak performance20 TFlop/sec

Disk space 48 TB

Operating systemDebian Linux AMD64 OS

Affiliates of UvA and VU can use this system directly.

LISA is a popular PBS cluster.

Hadoop (soon to come)

Hadoop (soon to come)

Finally

Which infrastructure to choose for your problem?

Ask eBioGrid.

How to get access?

NWO/NCF Iris system

Ask eBioGrid/SARA for help.

How to get assistance?

Ask eBioGrid for support.