Jan. 28, 2021

Pazy is a joint foundation of the Israel Atomic Energy Commission (IAEC) and the Council for Higher Education (VATAT).
Prof. Danny Hendler together with his colleagues, Prof. Hagit Attiya (Technion) and Dr. Gal Oren (Technion and Nuclear Research Center Negev) have won a Pazy grant for their research proposal entitled "Recoverable Algorithms using Non-Volatile Memory for Simulations in High Performance Computing Systems".

High performance computing (HPC) systems are a critical resource for scientific research and advanced industries. As the demand for compute-intensive calculations and simulations increases, huge amounts of data should be analyzed, manipulated and stored. HPC architectures are becoming more complex, thus fault rates incurred by large-scale computations increase while the penalty (in terms of runtime) incurred by each fault become costlier. This increases the need for efficient algorithms that correctly recover from such faults.

Our research investigates ways of utilizing emerging Non-Volatile Random-Access Memory (NVRAM) technologies and devices in efficient recoverable high-performance scientific applications. Towards this goal, we devise recoverable versions of common algorithms and leverage them to implement recoverable scientific applications. Such applications would obviate the need for checkpointing and may also even avoid Distributed File Systems entirely. This would lead to enormous savings in money, power and computation time, thus constituting a significant step forward for scientific ​computing towards the exascale era.


Danny_Hendler (002).jpg