SCS Talk: Ymir Vigfusson, Assistant Professor, Reykjavik University

Add to Calendar
March 27, 2014 11:00 am - 12:00 pm
Atlanta, GA


Ymir Vigfusson, Assistant Professor, Reykjavik University


Dynamically Profiling the Distributed Cache Tier in Data Centers


Scalable data replication protocols and layers, such as streaming, multicast and caching, enable many large distributed systems to be practical. As a concrete example, large-scale in-memory object caches like memcached are now widely used to accelerate popular web sites and to reduce burden on backend databases. Yet operators still have limited visibility into how these caches should be set up to optimally accommodate the workloads they see. How much would the cache performance improve from additional cache space, or by adding more cache servers to the pool? Since resources come at a cost, to what extent would request latencies deteriorate if cache memory were repurposed for a different service?


In this talk, I'll focus on some of the latest research questions pertaining to scalable data replication and large-scale distributed caches. In particular, I'll home in on the challenge of providing online monitoring of the cost and benefits of memory space in a large-scale cache, enabling cache operators to answer the questions above without requiring extraneous trace collection and manual offline tuning. I will introduce general and efficient algorithms for dynamically estimating hit rate curves -- histograms of cache hit rate as a function of memory size -- which can be plugged into cache replacement policies such as LRU. 

Extensive simulations on cache benchmarks indicate that these methods provide accurate estimates of hit rate at different cache sizes. Experiments on an implementation of these methods in memcached showed that hit rate curves were dynamically estimated at over 98% accuracy with only a small drop in throughput. The results are encouraging and suggest that exposing hit rate curves can be a practical method for improving provisioning and metering of large-scale caches.



Prof. Ymir Vigfusson has been an Assistant Professor in the School of Computer Science at Reykjavik University since 2011. Ymir received his B.Sc. in Mathematics from the University of Iceland (2005) and his Ph.D. in Computer Science from Cornell University (2009), where he researched approaches to exploit group similarity and improve scalability in distributed systems. His dissertation was nominated for the ACM Doctoral Dissertation Award by Cornell. Ymir was a post-doctoral scientist at IBM Research (2009-2011) where he worked on caching services in the cloud and the security of financial institutions. Ymir's research projects include creating and optimizing systems and algorithms for distributed settings, and getting data replication to work in a variety of environments. He founded Syndis in 2013, a growing start-up focused on offensive security and attack services for high-value clients. Ymir's work has been partially supported by a Fulbright Scholarship, a Yahoo! Research grant and a Grant-of-Excellence from the Icelandic Research Centre.