Maged Michael

Non-Blocking Synchronization and Memory Management 02-26-2015 @ 11:10 - 12:10

Maged Michael — Maged Michael is a Research Staff Member at the IBM Thomas J. Watson Research Center. He received a PhD degree in computer science from the University of Rochester. His areas of interest and experience include concurrent algorithms, distributed computing, non-blocking synchronization, memory management, multicore systems, transactional memory, cache coherence, and multiprocessor performance. His algorithms and libraries have been included in many commercial products and standard libraries. He was elected member of the Connecticut Academy of Science and Engineering. He was designated by the Association of Computing Machinery as an ACM Distinguished Scientist. He received the 2014 ACM SIGPLAN Most Influential PLDI Paper Award for his paper on Scalable Lock-Free Dynamic Memory Allocation.

Non-blocking data structures offer high availability and strong progress guarantees. They are guaranteed to be available regardless of thread scheduling, including thread termination and indefinite delay. This guarantee is most useful in cases such as sharing between threads and signal handlers, systems where processes could be terminated asynchronously, and soft real-time applications such as media players.

The inherent characteristics of non-blocking algorithms pose intricate problems that need to be solved to achieve the desired progress guarantees. Many of such problems are related to memory management, including dynamic memory allocation, the determination of the safety of reclaiming dynamic memory blocks, the ABA problem, and bounding memory use of non-blocking algorithms. This talk presents an overview of non-blocking synchronization and the memory management problems that arise in non-blocking algorithms and their solutions.


Video (1:00:18)