hankimage9 - Fotolia

UK supercomputer is now world’s biggest Panasas scale-out NAS cluster

Jasmin network supercomputer storage expands to 20PB of Panasas scale-out NAS, which has helped to cut analysis times for environmental data from days to hours

The UK’s environmental research supercomputer and data repository, Jasmin, has expanded storage capacity to 20PB of Panasas ActivStor scale-out NAS appliances.

This makes it the world’s largest deployment of Panasas clustered NAS storage and one that has cut data analysis run times from days to hours for Jasmin users as it processes between 1PB and 3PB a day.

Jasmin is funded by the Natural Environment Research Council (NERC) and the UK Space Agency (UKSA) and is delivered by the Science and Technology Facilities Council (STFC).

It handles environmental data from research that spans space, climate and the world’s oceans and provides data storage and analysis to academics and industry researchers, including for IPCC climate change modelling.

This creates workloads that comprise mostly very large files – multi-GB or TB, with 10s of TB expected soon.

With an addition of 1.6PB of capacity, Jasmin now has about 20PB in total, with around 5,000 compute cores available for data processing. There is also about 30PB of data held on tape.

The throughput of the Panasas cluster is 318GBps. IOPS is not important to the Jasmin cluster because the workload is made up of sequential processing of large files. Network bandwidths are measured in terabits.

Scale-out NAS allows nodes of storage to be added to a cluster to grow capacity and/or performance, usually with a parallel file system that scales to billions of objects.

Read more on NAS

STFC expects to have to buy a further 10PB of Panasas storage this year to keep up with what is described as data with “gravity” and persistence. In other words, these are large datasets – too large to be moved elsewhere – which must be retained for a very long time.

STFC had had previous experience with Panasas storage when it went to open tender and selected Panasas for the first phase of Jasmin in 2012. It deployed 5PB initially, with further phases bringing it up to the current 20PB.

The Panasas deployment comprises about 5,000 disk spindles, with some flash storage on newer units. Flash, however, is not vital to Jasmin operations because of the tendency towards large files and sequential processing and is used only for metadata.

Read more on Computer storage hardware