Elie Krevat's Home Page

Elie Krevat

Elie Krevat

Graduate Student
Department of Computer Science

Carnegie Mellon University
GHC 6221
ekrevat at cs dot cmu dot edu

Hi there! I'm a Ph.D. student in computer science at Carnegie Mellon, researching many flavors of distributed systems, storage, networks, virtual machines, and cloud computing. I'm also a member of the Parallel Data Lab where I am advised by Greg Ganger.

Back to Top

News

Back to Top

Research

My research interests include designing large distributed systems with better resource partitioning and coordination that can react to changes online, with a focus on cloud computing, storage, and networks. Some of my current and past research projects are listed below.

Automated performance problem mitigation with efficient resource allocations

Responding to resource-sensitive performance problems is becoming increasingly difficult for system administrators, and expensive in the amount of unnecessary overprovisioning, as distributed and cloud computing applications are built across larger numbers of interconnected shared services. Performance problems are common from many sources such as service upgrades and configuration errors, but root cause problem diagnosis efforts can take hours or days to isolate and fix the problem, even for system experts. This project proposes an automated approach to mitigate performance problems through the reactive provisioning of machines. This "quick fix" leverages end-to-end flow analysis to determine the critical path of requests, to help classify performance issues, and to direct an efficient allocation of resources to the services that have more effect on client-perceived delays. In many cases, problems can be mitigated in a few minutes after a problem is detected, returning client performance to acceptable levels and allowing any other necessary problem diagnosis efforts to continue unconstrained.

Seeking Efficient Data-Intensive Computing

New programming frameworks for scale-out parallel analysis, such as MapReduce and Hadoop, have become a cornerstone for exploiting large datasets. However, there has been little analysis of how these systems perform relative to the capabilities of the hardware on which they run. We developed a simple model of I/O resource consumption and applied it to a map-reduce workload to produce an ideal lower bound on its runtime, exposing the inefficiency of popular scale-out systems. Using a simplified dataflow processing tool called Parallel DataSeries (PDS), we also demonstrated that the model's ideal can be approached within 20%, and explored the reasons for the gap between ideal and actual performance that are faced by any DISC system built atop standard OS and networking services. We found that disk stragglers and network slowdown effects are the primary culprits for lost efficiency.

Incast: TCP Throughput Collapse in Cluster-based Storage Systems

Building cluster-based storage systems using commodity TCP/IP and Ethernet networks is attractive because of their low cost, ease-of-use, and the desire to combine routing infrastructures for LAN, SAN, and high performance computing. However, an important barrier to their use is the TCP Incast problem, where bursty traffic from synchronized reads in cluster-based storage systems produce a one to two order magnitude TCP throughput collapse. We have studied the network conditions that cause this TCP throughput collapse in both simulation and real-world deployments, examined the effectiveness of TCP- and Ethernet-level solutions, and with our latest publication we have found reasonable solutions to the problem with high resolution timers that implement a microsecond-granularity TCP retransmission timeout. This solution is both feasible and practical for fast storage networks while also safe for wide area networks, revisiting an older assumption on spurious TCP retransmissions that no longer holds true.

Back to Top

Publications

Back to Top

Other Projects and Presentations

Back to Top

Teaching

This semester I am TAing 15-712: Advanced Operating Systems and Distributed Systems.

I TAed 15-213: Introduction to Computer Systems in Fall '08.

I also TAed 6.033: Computer System Engineering at MIT when I was earning my M.Eng. degree.

Back to Top

Background

Before CMU, I completed a B.S. and M.Eng. in computer science at MIT, with a minor in economics. My master's thesis included work from a few summers and a semester of research at IBM T.J. Watson Research Center on system software for the Blue Gene supercomputer. I also spent a few years working for Microsoft as a software design engineer, where I played around with pre-alpha Vista technologies and developed the first two versions of Office Accounting Professional, a stand-alone product and third-party development platform for small business accounting.

Back to Top

Fun

The CSD summer softball team has been getting better and better. In 2013 we were the #1 seed in the regular season standings and lost a tough playoff game in the rain to finish #3.

I first got excited about sailing just before I left MIT, and after taking sailing lessons in Seattle. Sailing options in Pittsburgh are a bit more limited.

I returned to playing another sport that I haven't attempted since my undergraduate days -- ice hockey! I'm not ready to crack the Penguins roster just yet, but there's no better way to improve your athletic agility than avoiding large speeding human missiles on ice.

When I can find the time, I enjoy traveling to exotic world destinations. Some of my longer trips have included Spain, Italy, Greece, Thailand, Israel, Brazil, and Argentina (where I went on a month-long volunteering and solidarity trip during their economic crisis).

I'm always on a search for good sushi and fish restaurants - in Pittsburgh, Penn Fish Co. is one of my favorites.

I have two wonderful and very talented sisters, Ariela and Rina, but when I tell them that Rina says "Ariela is both of them." Ariela gets many kudos for designing this web page on the fly (she's a User Experience Designer at Amazon). Rina just graduated from UPenn's Digital Media Design program. Check out Ariela's web site here and Rina's here.

Back to Top