Doctoral candidate Mehran Elyasi has been recognized as a 2018 Facebook Fellow. The Facebook Fellowship Program awards its fellows with tuition and fees for the academic year (up to two years), a stipend, travel funds, and the opportunity to meet with and present their work to other researchers at Facebook’s annual summit.
Established seven years ago, the Facebook Fellowship program seeks to support doctoral students engaged in research that is promising, innovative, and pertinent to the fields of computer science and engineering. This year 17 doctoral students have been awarded the Fellowship. The topics from this year’s cohort range from natural language processing to networking and connectivity hardware, to security and privacy. The Fellowship application is open to doctoral students pursuing research in an accredited institution in any country.
Mehran’s fellowship winning work is his design of a novel coding scheme called Determinant Coding, for Distributed Storage Systems (DSS). The scheme seeks to overcome some of the challenges currently faced by such systems.
A large volume of digital data is generated by Internet users everyday. At the same time, the number of users seeking access to such data through applications such as Facebook, is also rapidly increasing. Distributed Storage Systems (DSS) are widely used in such large scale storage systems to maintain data availability and reliability. However there are several concerns faced by such systems, key among them being their unreliability. While the issue can be overcome by building in redundancy in the data, it leads to storage overheads. Additionally, with failures being common in large scale storage systems, a significant volume of network traffic is diverted to repair failed storage nodes. While the ideal goal is to minimise the repair bandwidth while also maximizing the storage efficiency of the system, currently one can be optimized only at the cost of the other.
To counter these challenges, Mehran has designed a novel coding scheme, called Determinant Coding, for Distributed Storage Systems. The construction of the scheme provides encoding/decoding algorithms for storage as well as an efficient mechanism for the repair of failed storage units. These universally structured codes can operate in all the optimum points of the storage-bandwidth trade-off.
Mehran earned his Bachelor of Science in Electrical Engineering and Mathematics from Isfahan University of Technology, Iran, in 2014. His areas of interest include information theory and its applications in communication, distributed storage systems, and statistical machine learning. He is pursuing his doctoral degree under the guidance of Prof. Soheil Mohajer.