Portal de Periódicos da CAPES

Collaborative Memories in Clusters: Opportunities and Challenges

2014; Springer Science+Business Media; Linguagem: Inglês

10.1007/978-3-642-54212-1_2

1611-3349

Ahmad Samih, Ren Wang, Christian Maciocco, Mazen Kharbutli, Yan Solihin,

Distributed and Parallel Computing Systems

Highly-integrated distributed systems such as Intel Micro Server and SeaMicro Server are increasingly becoming a popular server architecture. Designers of such systems face interesting memory hierarchy design challenges while attempting to reduce/eliminate the notorious disk storage swapping. Disk swapping activities slow down applications' execution drastically. Swapping to the free remote memory - near by nodes, through Memory Collaboration has demonstrated its cost-effectiveness compared to overprovisioning memory for peak load requirements. Recent studies propose several ways to access the under-utilized remote memory in static system configurations, without detailed exploration of dynamic memory collaboration. Dynamic collaboration is an important aspect given the run-time memory usage fluctuations in clustered systems. Furthermore, with the growing interest in memory collaboration, it is crucial to understand the existing performance bottlenecks, overheads, and potential optimizations. In this paper we address these two issues. First, we propose an Autonomous Collaborative Memory System (ACMS) that manages memory resources dynamically at run time, to optimize performance, and provide QoS measures for nodes engaging in the system. We implement a prototype realizing the proposed ACMS, experiment with a wide range of real-world applications, and show up to 3x performance speedup compared to a non-collaborative memory system, without perceivable performance impact on nodes that provide memory. Second, we analyze, in depth, the end-to-end memory collaboration overhead and bottlenecks. Based on this analysis, we provide insights on several corresponding optimizations to further improve the performance.