I am Gowtham Kaki, a Ph.D candidate at Purdue University, USA. I am excited to present my paper on Safe Transferable Regions at ECOOP this year! The paper highlights the performance problems created by language-level Garbage Collection (GC) in a distributed data processing system (such as Microsoft’s Dryad), and proposes a language framework (embedded in C#) called Broom that mitigates the problem. Broom does so by safely getting rid of the GC in favor of Region-based Memory Management wherever possible, resulting in a significant improvement of latency (10%-36%) in dataflow computations.
The ECOOP paper is primarily focused on the region type system and type inference that let us enforce memory safety in Broom even without a GC. In general, I am interested in reasoning about the safety and functional correctness properties of concurrent and distributed programs. I consider questions such as: What sort of programming abstractions make it easy to build complex distributed applications? What reasoning techniques are most useful to reason about such programs? How can we build program analyses to detect bugs, or verify that there are no bugs of a particular kind? etc. One of the things that I look forward to during ECOOP is meeting like-minded researchers who work on similar problems, and having engaging discussions with them. I always enjoy research discussions at PL conferences, and I expect ECOOP to be particularly rewarding since it gives me an opportunity to meet my fellow researchers in the Europe, whom I otherwise may not get to meet. Furthermore, ECOOP this year has a host of talks in the areas of concurrency, type systems, and formal methods, which are of interest to me, and which I plan to attend.
My travel to Amsterdam is funded in part by the ECOOP committee, for which I am very grateful. Among other things, the funding allows me to extend my stay beyond the day of my talk, and attend all the sessions of the conference to which I look forward.