Authors: Sayan Ghosh and Nathan Tallent (Pacific Northwest National Laboratory (PNNL)); Marco Minutoli and Mahantesh Halappanavar (Pacific Northwest National Laboratory (PNNL), Washington State University); Ramesh Peri (Facebook); and Ananth Kalyanaraman (Washington State University, Pacific Northwest National Laboratory (PNNL))
Abstract: Because of cost, non-volatile memory NVDIMMs such as Intel Optane are attractive in single-node big-memory systems. We evaluate performance and cost trade-offs when using Optane as volatile memory for huge-graph analytics. We study two scalable graph applications with different work locality, access patterns and parallelism. We evaluate single and partitioned address spaces; Memory and AppDirect modes; and compare with distributed executions on GPU-accelerated and CPU-based supercomputers.
We show that AppDirect can perform and scale better than Memory for the largest working sets (12%), even when dominated by irregular access patterns, if most accesses are NUMA-local and Optane accesses are frequently reads. Surprisingly, between Memory and AppDirect, processor-cache performance can change due to line invalidations; updates to the caching policy (via non-temporal hints) can make a 25% improvement. We observe that single-node graph analytics frequently has >4–10x cost/performance advantages over distributed-memory executions on supercomputers.
Back to Technical Papers Archive Listing