IQ-Twemcached is one aspect of a broader effort to provide cache augmented database management systems with physical data independence. Hieu, Jason, and I designed and developed the I and Q leases to prevent undesirable race conditions from causing the cache to produce stale data. This provides strong consistency which is known to simplify a programmer's reasoning about the behavior of an application. Our proposed leases are in support of simple operations that read and write a small amount of data from big data. While the I lease is implicit, the programmer must extend the application with Q leases explicitly. They are trivial to use and their impact on correctness is immediate.
As a PhD candidate, I have been working on transparent cache consistency in cache-augmented database management systems (CADBMS). Caches like memcached are widely used throughout the industry to reduce database load and greatly improve system throughput. While the performance benefits of CADBMSs are evident, utilizing these systems effectively is a laborious and error prone process. We developed SQLTrig to provide a seamless caching solution that leverages the CADBMS architecture and offers physical data independence to the application developer. One of the critical components of this solution is handling undesirable race conditions with the IQ framework.
I am a first-year Ph.D. student at the University of Southern California. My research interest is on big data in support of operations that read and write a small amount of data. A key research direction is strong consistency in cache-augmented database management systems (CADBMS). Current techniques suffer from stale data produced by undesirable race conditions. The IQ framework with leases addresses this limitation by preventing those race conditions.