Prof. Assaf Schuster

Prof. Assaf Schuster of the Computer Science Department at the Technion published over 250 papers in the areas of Machine Learning, AI, Streaming Data, Big Data Systems, Parallel and Distributed Computing, Scalability. He is a Fellow of the ACM and the IEEE, and is heading MLIS, the Technion center for Machine Learning and Intelligent Systems.

Speech Title：Unleashing affordablecloud resources with SpotOS

Abstract

Community wisdom tells us that while the cloud early vision promised economy of scale, in practice, renting cloud resources is an expensive business. 

There are many reasons for this unfortunate state of affairs, some of which can be attributed to non-utilized opportunities.

SpotOS is an effort which is aimed to use state of the art tools to help cloud users make better rental choices.

In particular, SpotOS enables wider, easier, and better use of spot instances, which are “empty” resources sold in high discount with the pitfall that they may have to be evacuated in short notice.

SpotOS can be viewed as a new kind of a distributed “operating system” which extends the OS traditional tasks by optimizing task allocation to resources, optimizing batch resource rental from the provider, and supporting failover mechanisms.

To achieve these goals SpotOS uses advanced modern tools such as optimization algorithms, predictive machine learning, efficient external storage, and stateful task management in distributed environments.

The talk will survey the challenges and opportunities in SpotOS.