Hierarchical Clustering: A Structure for Scalable Multiprocessor Operating System Design

We propose a simple structuring technique based on clustering for designing scalable shared memory multiprocessor operating systems. Clustering has a number of advantages. First distributed systems principles are applied by distributing and replicating system services and data objects to increase locality, increase concurrency, and to avoid centralized bottlenecks, thus making the system scalable. However, since there is tight coupling within a cluster, the system performs well for local interactions. Second, a clustered system can easily be adapted to different hardware configurations and architectures by changing the size of the clusters. Third, clustering maximizes locality which is key to good performance in large NUMA systems. Finally, clustering leads to a modular system composed from easy-to-design and hence efficient building blocks. In this paper, we describe clustering and how it is applied to a micro-kernel-based operating system. We present measurements of performance, as ..