Date

Presenter

Topic

May 25

Monia Ghobadi

Trickle: Rate Limiting YouTube Video Streaming

YouTube traffic is bursty. These bursts trigger packet losses and stress router queues, causing TCP's congestion-control algorithm to kick in. In this paper, we introduce Trickle, a server-side mechanism that uses TCP to rate limit YouTube video streaming. Trickle paces the video stream by placing an upper bound on TCP's congestion window as a function of the streaming rate and the round-trip time. We evaluated Trickle on YouTube production data centers in Europe and India and analyzed its impact on losses, bandwidth, RTT, and video buffer under-run events. The results show that Trickle reduces the average TCP loss rate by up to 43% and the average RTT by up to 28% while maintaining the streaming rate requested by the application.

Presenter Bio: Monia is PhD candidate in Systems and Networking group at University of Toronto. Her research interests are in the general area of computer networking, including data center networking, transport protocols, switch and router architecture, resource management, network measurement, and online social networks. In the past she worked on sizing buffers in Internet routers.

Jun 1

George Amvrosiadis

Practical Scrubbing: Getting to the bad sector at the right time

Latent sector errors (LSEs) are a common hard disk failure mode, where disk sectors become inaccessible while the rest of the disk remains unaffected. To protect against LSEs, commercial storage systems use scrubbers: background processes verifying disk data. The efficiency of different scrubbing algorithms in detecting LSEs has been studied in depth; however, no attempts have been made to evaluate or mitigate the impact of scrubbing on application performance.

In this talk, we provide the first known evaluation of the performance impact of different scrubbing policies in implementation, including guidelines on implementing a scrubber. To lessen this impact, we present an approach giving conclusive answers to the questions: *when* should scrubbing requests be issued, and at *what size*, to minimize impact and maximize scrubbing throughput for a given workload. Our approach achieves six times more throughput, and up to three orders of magnitude less slowdown than the default Linux I/O scheduler.

Presenter Bio: George is a Ph.D. student in the Computer Systems and Networks group at the University of Toronto. His current research revolves around storage reliability and component idleness characterization, detection and utilization. He completed his B.Sc. at the University of Ioannina in Greece, with a B.Sc. thesis studying namespace management techniques for federated filesystems.

Jul 6

Sergey Blagodurov

Towards the contention aware scheduling in HPC cluster environment

Contention for shared resources in High-Performance Computing (HPC) clusters occurs when jobs are concurrently executing on the same multicore node (there is a contention for shared caches, memory buses, memory controllers and memory domains). The shared resource contention incurs severe degradation to workload performance and stability and hence must be addressed. The state-of-the-art HPC clusters, however, are not contention-aware. The goal of this work is the design, implementation and evaluation of a virtualized HPC cluster framework that is contention aware.

Presenter Bio: Sergey Blagodurov is a Ph.D. Candidate in School of Computing Science at Simon Fraser University, Vancouver, Canada. His research focuses on resource contention-aware scheduling in High-Performance Computing (HPC) clusters of multicore machines and exploring new techniques to deliver better performance on non-uniform memory access (NUMA) multicore systems under Linux. He is also a Research Associate at HP Labs, where he studies the design and operation of net-zero energy data centers.

Jul 20

Leonid Ryzkyk

Automatic Device Driver Synthesis

Faulty device drivers are the leading source of reliability problems in modern operating systems. Automatic device driver synthesis is a radical approach to creating drivers faster and with fewer defects. In this talk I will give an overview of the game-based approach to device driver synthesis, where a correct driver implementation is obtained as a winning strategy in a game played by the driver against the device and its physical environment. I will list the main challenges involved in synthesising drivers for complex real-world devices and will present our driver synthesis tool called Termite.

Presenter Bio: Leonid Ryzhyk completed his Ph.D. in 2010 under Gernot Heiser at the University of New South Wales and NICTA. His PhD and subsequent work (mostly in collaboration with Intel Labs) has been on making device drivers more reliable. Some of his work was published in HotOS, SOSP, ASPLOS, and EuroSys.