Replication is a key approach for scaling wide-area applications and for achieving high-performance and high-availability. However, because of wide-area latency and potential
Internet congestion/failures, consistency overhead in a naive replication system may actually lead to degraded overall performance and availability relative to a centralized architecture.
This duality currently limits the deployment of replicated services across the wide area.
In this talk, I will discuss why system consistency is continuous rather than binary for many applications, and how these applications can significantly benefit from exploring the semantic space between strong consistency and optimistic consistency. Specifically, I will describe the TACT continuous consistency model that allows applications to dynamically tune replicated system consistency. I will also briefly discuss the TACT prototype and performance evaluation across the Internet.
In the second part of my talk, I will explore the consistency model's effects on system bound under continuous consistency. This protocol-independent and implementation-independent upper bound reveals the inherent effects of continuous consistency on availability. Next I will describe how we measure the availability of existing consistency protocols in both WAN and
a local emulation environment and compare the results to the upper bound.
Monday, April 22 at 3:00pm in Duncan Hall 1070
A reception in DH 3092 will follow the talk.
About Haifeng Yu
Haifeng Yu is a Ph.D. candidate at Duke University. His research interests lie in the general area of distributed systems and operating systems, with current work focusing on data replication and consistency in WAN. He received his M.S. from Duke University and his B.E. from Shanghai Jiaotong University, China. More information about his research is available at http://www.cs.duke.edu/~yhf.
![[RiceCS]](http://www.cs.rice.edu/Database/media/CSLogo.gif){kind=small}
![[Rice]](http://www.cs.rice.edu/Database/media/Owl.gif){kind=small}