Date: 10:00am, March 6, 2017
Location: Costa Commons (CEPSR 750)
Speaker: Dr. Ankit Singh Rawat, Post-Doctoral Associate, Massachussets Institute of Technology
Abstract: Storage on the cloud is an increasingly popular choice for storing and analyzing vast amounts of data. Besides serving as backbone systems for large institutions such as CERN, Google, and Microsoft, cloud storage systems have been instrumental in the emergence and rapid growth of modern cloud computing framework. It is imperative that such systems enable efficient repair mechanisms and allow fast access to data. Furthermore, the distributed structure of cloud storage systems and the valuable and confidential nature of information stored in these systems make them susceptible to various security and privacy threats.
In this talk, I'll present a comprehensive coding theoretic framework to address and answer the aforementioned key challenges arising in the context of designing cloud storage systems. I’ll describe new constructions of codes that ensure resource efficient regeneration of the content lost due to server failures. These codes provide practical regeneration schemes without compromising fault-tolerance of the storage system. Focusing on the issue of fast access to data for seamless computation, I’ll then present a simple approach for designing high rate batch codes which support a large number of parallel accesses to the information stored in the system. I’ll conclude the talk by presenting optimal coding schemes that securely store information against eavesdropping attacks.