Using the VL-e Grid infrastructure for Life Sciences

Eindhoven 02 April 2008For IBM and the Vrije Universiteit Amsterdam, Alex de Landgraaf presented some results of using the VL-e Grid infrastructure for a Life Sciences application. He presented his results at the Gridforum.nl Business Day in Eindhoven. VL-e - Virtual Laboratory for e-Science - is a 20 million euro Dutch project that aims at creating a virtual laboratory environment for e-Science and industrial research in The Netherlands. The project is running for several years and has a working Grid infrastructure. Hence the idea to see whether the VL-e Proof-of-Concept Grid could be used by the Acadamic Medical Centre, Amsterdam (AMC) for routine investigations of genome analysis.

Advertisement

AMC has a large genome sequencer, the FLX from Roche. The storage requirements of raw data are about 10 Tbyte/year. The machine is rather expensive to operate. Currently two runs per day are done, which cost 5000-7000 euro each. The data are sensitive, hence need to be treated with high levels of security.

The data is analysed by non-IT specialists: researchers and clinical personal. Hence it is necessary to integrate the Grid infrastructure into the existing infrastructure that is used on a day-to-day basis. Therefore, Alex de Landgraaf chose an architecture where there was a Grid Access Point at the AMC that was connected via the Grid to Grid Storage Resources at SARA Computing and Networking Services and the Dutch National institute for Subatomic Physics (NIKHEF).

The Grid Access Point takes care that the files on the Grid are presented to the user as part of his file system tree. To him the Grid storage looks just like an attached disk, like any other disk attached from the network. The Grid can take care of on-the-fly encryption.

The solution proved to meet the security, privacy and usability requirements, said Alex de Landgraaf. However, performance and availability of the solution were not that good. The speed from local application to the Grid was only 7,4 Mbit/s. Copying a 200 Mbyte file took on average 25 seconds.

Availability was tested by listing and copying files using gLite utilities directly during a 4 week test period. Copying a 100 Kbyte file showed an uptime of 93,34 percent. This means 1 out of 12 file copy commands failed. The latency measured was 4,5 seconds.

These were disappointing results. Alex de Landgraaf said he thought this was due to single points of failure due to lack of redundancy in key Grid services provided by gLite. Hence he concluded the VL-e Proof-of-Concept Grid is good enough for e-Science, but not yet for (medical) business.

In the discussion following his presentation, it was suggested to also have a look at the VL-e production Grid as this could have better availability characteristics. More information can be found at the Gridforum.nl website.


Ad Emmen

[Medical IT News][Calendar][Virtual Medical Worlds Community][News on Advanced IT]