iHub Cluster Goes Live

It’s with great excitement and anticipation that we can formally say that the is finally up and running!

The iHub Cluster was a project thought of by an iHub Community member (Idd Salim) and was kindly funded by Google (through a grant) and Intel (through an infrastructure donation – Intel gave the iHub cluster a blade chassis populated with 3 blade servers).

Unfortunately, due to a myriad of reasons, the project was delayed over the last couple of months (primarily lags in shipping for the infrastructure we got).   There are two ideas behind the Cluster. One of the ideas is to act as a sandbox for parallel programming. Given the nature of mass market hardware today, it’s incumbent on the software developers in Kenya to catch up with the trends in the rest of the world and learn how to write software for multiple processor environments.

Centre for High Performance Computing

The second idea is to simply create a centre for high performance computing – there’s none in Sub Saharan Africa (with the exception of South Africa). The main problem with this situation is that a lot of research in fields both within and without computing cannot be conducted due to lack of requisite infrastructure. A simple example is, for example, weather forecasting, It’s not as accurate as it could be in Kenya.  Additionally, analytics is moving from the fringes of technology and is taking up a more mainstream role with regards to use and adoption in Kenya.

Eric Schmidt at the iHub Cluster with Phares Kariuki, iHub Cluster LeadEric Schmidt at the iHub Cluster with Phares Kariuki, iHub Cluster Lead

The biggest hindrance to both the aforementioned examples is compute power.   The current setup consists of 1 ‘head node’ which is basically the Intel Blade Chasis that Intel donated to the iHub (has been setup as a CPU cluster) along with two GPU clusters. Total output as it stands is 23 Tera Flops.   Within the next two weeks, another two nodes (both GPU Clusters) will be powered up, bringing the total number nodes (excluding the master node) in our deploy to four. We will have a total output of roughly 42 Tera flops.   We have a (soft) target of getting to 90 Tera Flops within the next twelve months.

What does this mean for the community? 

We unfortunately are not yet taking up requests for processing (apart from those who are involved in testing the platform) as we try and figure out how the platform will work – billing, scheduling, priority etc. This should be done by the 11th of February, 2013 and at that point, we will be open to the community.   This project will eventually morph into a traditional “Center for High Performance Computing” where we will be taking specific research requests, so partnerships with Universities are being worked on for sustainability in the medium to long term.

We would also like to thank all community members and our funders for their patience, especially because this process has been unnecessarily painful and drawn out. We are, however, learning from our mistakes. Keep reading the iHub blog and follow the iHub Cluster Twitter account (@iHubCluster) to get a feel of what we are doing/planning to do!

 

Related Stories

http://www.ihub.co.ke/blog/2012/09/the-ihub-cluster/

http://www.ihub.co.ke/blog/2012/05/the-ihub-ux-lab-and-supercomputer-cluster/

http://cluster.ihub.co.ke/

https://twitter.com/iHubCLuster

iHub Cluster

Go to Source

zv7qrnb
Tags:

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>