Abhishek Verma

Interned in the Storage and Information Managements Platforms Lab and researched on MapReduce performance modeling. By profiling MapReduce jobs, we built a compact performance model representing their execution. It is used to optimize the overall infrastructure utility and achieve specified service level objectives.

Contributed to the design and development of Direct Object Repository Architecture (DORA), which is a high performance, horizontally scalable and reliable object based back-end storage. Specifically, I worked on distributing the meta-data for fault tolerance and higher availability, wrote a FUSE (Filesystem in User Space) implementation and a Distributed Filesystem for Hadoop, the open source MapReduce implementation.

Worked on Orkut, an online social networking website. It involved understanding technologies like MapReduce and the distributed file systems, collecting data for research on Orkut and developing new features.