IBM Kittyhawk
Kittyhawk is a new IBM supercomputer. The proposed project entails constructing a global-scale shared supercomputer capable of hosting the entire Internet on one platform as an application, whereas the current Internet is a collection of interconnected computer networks.[1][2]
In 2010 IBM open sourced the Linux kernel patches that allow otherwise unmodified Linux distributions to run on Blue Gene/P. This action allowed the Kittyhawk system software stack to be run at large scale at Argonne National Lab. The open source version of Kittyhawk is available on a public website hosted by Boston University.[3]
In 2012 the Kittyhawk project was made a part of the United States Department of Energy fault oblivious execution (FOX) project, and ported to run on the Intrepid supercomputer at Argonne National Laboratory.[4]
In 2013 researchers used the Kittyhawk project to demonstrate a novel high-performance cloud computing platform by merging a cloud computing environment with a supercomputer. [5] [6]
Specifications
IBM Research published three papers[7][8][9] detailing the project. The Kittyhawk will be based on the previously developed IBM supercomputer called Blue Gene/P. In theory the Kittyhawk will have up to 16,384 racks, providing a maximum of 67.1 million cores with 32 PB (32 PiB) of memory.[10]
References
- ^ IBM Proposes One Computer to Run Entire Internet
- ^ One computer to rule them all
- ^ Open Source Kittyhawk
- ^ At exascale, being oblivious to a fault keeps apps running
- ^ Researchers Implement HPC-First Cloud Approach
- ^ Researchers Describe Project to Merge Cloud Computing and Supercomputing
- ^ Project Kittyhawk: Building a Global-Scale Computer
- ^ Kittyhawk: Enabling cooperation and competition in a global, shared computational system
- ^ Providing a Cloud Network Infrastructure on a Supercomputer
- ^ IBM explores 67.1m-core computer for running entire internet
See also
External links
- Current Kittyhawk Homepage (Boston University)
- Towards a Global Scale Public Computer — lecture given by IBM Engineer, Jonathan Appavoo at Stanford University (video archive).