Synchronization and consistency in concurrent systems

Licentiate thesis, 2004

This thesis investigates aspects of synchronization and coordination in concurrent systems. In such systems synchronization and coordination are highly important as they form a basis for how a set of entities can collaborate to solve a task. In systems without shared knowledge of global time logical clocks provide a way to track how events are related to each other. Small-sized logical clocks with high causal-ordering accuracy is useful, in particular where (i) the precision of the knowledge of the causal dependencies among events implies savings in time overhead and (ii) the overhead of Full Vector clock timestamps is high. We introduce the Non-Uniformly Mapped R-Entries Vector (NUREV) clocks, a general class of plausible clocks that allow accuracy adaptation and we analyse the ways that these clocks may relate causally independent pairs of events. Our analysis resulted in a set of conclusions and the formulation of new, adaptive plausible clocks algorithms with improved accuracy even when the number of clock entries is very small, which is important in peer-to-peer communication systems. Concerning system services in concurrent systems, such as scheduling and resource allocation, the potential of efficient synchronization methods is large. In particular, in multiprocessor systems certain synchronization methods, such as lock-based ones, may limit the parallelism. It is significant to see the impact of wait/lock-free synchronization design in key services for multiprocessor systems, such as the memory allocation service. Therefore efficient and scalable memory allocators for multi-threaded applications in multiprocessor systems is a significant goal of recent research projects. We propose a lock-free memory allocator, to enhance the parallelism in such systems. Its architecture is inspired by Hoard, a successful concurrent memory allocator, with a modular and scalable design that preserves scalability and helps avoiding false sharing and heap blowup. Within our effort on designing appropriate lock-free algorithms for the synchronization in multiprocessor systems, we propose a new non-blocking data structure called flat-sets, supporting conventional ``internal'' operations as well as ``inter-object'' operations for moving elements between flat-sets. Our experiments indicate that the scalability properties are enhanced even further with the help of the lock-free synchronization, without compromising the other good properties provided by the Hoard architecture.