Saba Ahmadian

Ph.D. student

Cost-Effective Storage Architectures for Emerging Big Data Applications

  Halls department, Hall 3
  Thursday, 28 December 2017
  17:00 - 18:00

Abstract

With increasing performance requirements of data-intensive applications in data centers, storage subsystems have become performance bottlenecks of computing systems. Hard Disk Drives (HDDs), which are used as main media to store user data in storage systems, provide large capacity and low cost, but they suffer from low performance, particularly for random access workloads in Big Data applications. The low performance of HDD-based storage systems can be avoided by employing high-performance storage devices such as Solid-State Drives (SSDs). Compared to HDDs, SSDs provide higher performance due to their non-mechanical structure used to retrieve and store data. SSDs, however, impose up to 10X higher cost and support only a limited number of reliable writes, which makes the replacement of HDDs by SSDs usually prohibitively expensive. In order to take advantage of the merits of both HDDs and SSDs, several studies from EMC, Facebook, and VMWare employ high-performance SSDs as a caching layer for high-capacity HDDs in storage systems. Applying such I/O caching schemes requires a proper cache management mechanism in order to achieve higher I/O performance and reliability and also preserve the endurance of the SSDs. In this talk, I will briefly introduce major performance and reliability challenges of storage systems for emerging I/O intensive applications. Then, our recent research work to tackle these challenges are presented.

Bio

Saba Ahmadian received the B.S. and M.S. degrees in computer engineering from Sharif University of Technology (SUT), Tehran, Iran, in 2013 and 2015, respectively. From 2011 to 2012, she was a member of Energy Aware Systems Lab (EASY), SUT, where she researched on power reduction techniques on embedded CPUs. From 2012 to 2015, she was a member of Embedded Systems Research Lab (ESR), SUT, where she researched on low power and reliability-aware techniques on automata-based embedded systems. Currently she is a Ph.D. candidate at Data Storage, Network, and Processing (DSN) Lab at SUT under supervision of Dr. Hossein Asadi. Her research interests include storage systems design, virtualization platforms, fault tolerant design, and Low power systems design.