Data Science Field Trip to Spain’s Most Powerful Computing Cluster

Everyone needs to take a break once in awhile. Unburying ourselves from a pile of Latin and Greek letters, most notably q’s and delta’s and little red g‘s, the 2015-16 candidates for a Master’s in Data Science at the Barcelona Graduate School of Economics visited MareNostrum, Spain’s most powerful computer cluster located on the western edge of inland Barcelona.

Aside from its impressive technical specifications (see MareNostrum by the Numbers below), MareNostrum is hosted in a beautiful space: the Torre Girona chapel, where the meditative ambiance is set by Gregorian chanting. I like to think the chanting soothes and cools the compute units as they tirelessly process and “crunch” data.

12049126_10153931328919311_2485919657190857984_n

MareNostrum by the Numbers,:

  • 3,056 computing nodes
  • 48,896 Sandy Bridge processors
  • 10,240 CPU’s
  • 100.8 TB of memory
  • 2 PB of disk storage
  • 1.1 Petaflops per second, peak performance
  • Each compute node hosts a 500GB storage disk alongside 32GB of memory.

MareNostrum’s supporting infrastructure:

  • The first two racks on the far left are dedicated as the “System Queue” where researchers enqueue data processing jobs.
  • The racks located to the right of the system queue are dedicated to long-term storage. After a job is finished, data is moved from a local disk to these storage racks.
  • Two “research” racks located in the back of the cluster host microprocessors as proof of concept in the use of such processors as part of a distributed computing architecture.
  • Compute nodes run a Linux Operating System and networking is enabled by gigabit ethernet and Infiniband interconnection.

Proof of Concept Research in Microprocessors

20 years ago, the largest computers had the same computing power as a modern-day tablet. The next generation of supercomputing may be in the use of microprocessors which were inspired by mobile technologies in our phones and tables. This seems like a good thing because Torre Girona is a limited space which needs to upgrade its hardware every 4 years to keep up with demand and modern technology.

Projects on MareNostrum

The Barcelona Supercomputing Center won the 2012 Scientific Visualization Challenge for Alya Red, their “Computational Cardiac Model”. In other words, they created a scientifically realistic visualization of the human heart using electric impulses. This work was made possible by MareNostrum. In addition to their computational muscle, the scientific visualization team at BSC creates media to relate how they did accomplish such feats and the value such work offers the scientific community.

BSC is located next door to MareNostrum and engages in many large-scale projects with both academic researchers and industry. See more about these projects on their youtube channel.

A big thanks to Oriol Riu for giving us a tour and Fernando Cucchietti for speaking with us about the projects going on at BSC!

Advertisements

One thought on “Data Science Field Trip to Spain’s Most Powerful Computing Cluster

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s