ICC Builds Solution for NCSA “Dark Energy Survey”

Share on Google+Tweet about this on TwitterShare on FacebookShare on LinkedInPin on PinterestEmail this to someone

International Computer Concepts, with partner LSI, has developed a supercomputing solution for one of the most challenging astronomy projects ever conceived.

The National Center for Supercomputing Applications (NCSA) is an organization that develops and deploys national-scale cyberinfrastructure that advances science and engineering, providing high performance computing (HPC) resources to researchers across the country. Leading-edge computing, data storage, and visualization resources come out of the NCSA, and it is home to some of the most complex and computationally-intensive projects.

One of these projects is the “Dark Energy Survey,” an attempt to solve one of today’s greatest scientific mysteries: why does the expansion of the universe appear to be speeding up.

The Challenge

The “Dark Energy Survey” will collect digital data from astronomical observations in a database expected to grow by 400GB daily over the one and a half year life of the project to approximately 200TB of raw image data.

The Solution

ICC delivered a configuration that would maximize the in-box data throughput and provide the scalability for the clustered configurations the projects will generate. The system was built around the LSI MegaRAID SAS 9260-8i low-profile MD2 eight-port internal 6Gb/s SATA and SAS PCIe RAID controller and the LSI MegaRAID CacheCade software. The addition of the CacheCade software allowed ICC to accelerate the performance of HDD arrays by enabling 3 x 160GB SSDs to be configured as an additional read cache resource available to the controller.

The Result

The combination of the 6Gb/s RAID card from LSI and its CacheCade software solved a real performance problem. It was taking NCSA four to six hours to create database indices. With the new 9260-8i card and CacheCade software, that time fell to about 15 minutes, a factor of roughly 20:1.

There are two cool things about the case study: the fact that the cluster will eventually be used to test some of Einstein’s theories about the universe, and the way the storage system gets around the I/O bottleneck problem. The storage solution that we developed for the NCSA take advantages of the CacheCade technology and allows solid-state drives to be used as RAID controller cache, helping to overcome the I/O bottleneck plaguing storage technology today so that DES can transfer the approximately 400GB of data that will be generated each day by the new telescope in Chile to storage systems in the United States. The plan is to eventually add more than one controller per node so that the controllers can also pool their SSD-based cache. Initial tests showed significant performance gains.

To learn more about this check out the full case study.

Comments

  1. Grayson Manley says:

    Very interesting solution. SSDs are definitely the future, and combining that with advanced raid technology seems like a win win. Interesting post, love the blog, tons of interesting material, thanks!

Leave a Reply