Abstract
With the deployment of innovative memories such as non-volatile memory and 3D-stacked memory in distributed systems, how to improve the application performance by utilizing the unique characteristics of these hybrid memories remains an active research direction. For instance, the Intel Knight Landing (KNL) processor incorporates a High Bandwidth Memory (HBM) using 3D-stacked technology with traditional DRAM onto the same chip. HBM achieves much higher bandwidth than traditional DRAM when the application exhibits high parallelism and sequential access. In this paper, we propose a new metric SP-factor to guide the data scheduling in distributed system using hybrid memories such as HBM and DRAM. The SP-factor incorporates the data access patterns including data block size and data access parallelism, which leads to better data scheduling decision for higher performance. We apply SP-factor to several data eviction policies on the hybrid memory system, which achieves better performance. Moreover, an adaptive data scheduling method (ADSM) is proposed for such hybrid memory system with HBM and DRAM. ADSM can dynamically adjust scheduling decisions based on runtime performance metrics so that it can adapt to workloads with different data access patterns. Our experimental results show that ADSM can significantly improve the performance of the representative workloads. For SQL query application with mixed access pattern, the cache hit ratio increases by 10.4% and the execution time reduces by 14.6% using ADSM compared to ARC policy.
Highlights
With the development of innovative memories, emerging memory technologies such as non-volatile memory [1], [2], 3D-stacked [3], [4], memory have been integrated into the memory system, effectively complemented traditional DRAM
LRU-SP policy combines the temporal locality and the characteristics of high bandwidth memory (HBM)-DRAM hybrid memory system, which ensures that the most recently accessed data blocks residing in HBM, and the data blocks with high recency are re-ordered by SP-factor
The constraint is the capacity of HBM, and the target of the data block scheduling in the HBM-DRAM hybrid memory system can be expressed as Equation 13
Summary
With the development of innovative memories, emerging memory technologies such as non-volatile memory [1], [2], 3D-stacked [3], [4], memory have been integrated into the memory system, effectively complemented traditional DRAM. For hybrid memory using HBM and DRAM, the existing scheduling methods cannot fully exploit the high bandwidth of HBM and the low latency of DRAM, which wastes the potential for further performance improvement of big data applications. HBM and DRAM, we propose a new data scheduling metric SP-factor, which represents the impact on application performance to store data block in HBM It overcomes the limitations of existing scheduling methods by considering the impact of both data block size and data access parallelism, which improves the effectiveness of data scheduling methods. Both the design overview and implementation details are provided.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
