Bilkent University
Department of Computer Engineering


Load balancing performance in distributed storage with regular balanced redundancy


Dr. Mehmet Fatih Akta
MathWorks, MA, USA

Contention at the storage nodes is the main cause of long and variable data access times in distributed storage systems. The offered load on the system must be balanced across the storage nodes in order to minimize contention, and the load balance in the system should be robust against the skews and fluctuations in content popularities. Modern systems replicate data objects across multiple nodes to allow for load balancing. However redundancy increases the storage requirement and should be used efficiently. In this talk, we will evaluate the load balancing performance of a set of natural storage schemes. Specifically, we will investigate how the load balance in the system changes with respect to the important system parameters such as the number of objects, the number of nodes and the replication factor. We will also extend our analysis to the storage schemes that use XOR's rather than object replicas.

Bio: Dr. Mehmet Fatih Aktas is currently an engineer working on the execution engine of MATLAB. Theme of his work at MathWorks is to make MATLAB faster. His research interest is broadly to make distributed computer systems faster and more robust. More recently, he also started working on preserving user anonymity on the Internet. Along these lines, he has been doing research in both systems development and theoretical analysis. While solving research problems, he mostly relies on probabilistic modeling and on tools from applied probability such as queueing theory, order statistics and reinforcement learning. He previously completed an MS and a PhD in Electrical and Computer Engineering at Rutgers University, and a BS in Electrical and Electronics Engineering at Bilkent University.


DATE: 10 December 2020, Thursday @ 16:00