2019/04/15 by Gurbinder Gill, Gill, Gurbinder, Roshan Dathathri +7 · 1 citation
Computer Science · #B.3.1 #C.2.4 #D.1.3 #Distributed #Distributed systems and fault tolerance #FOS: Computer and information sciences #Graph Theory and Algorithms #Parallel #Parallel Computing and Optimization Techniques #and Cluster Computing (cs.DC)
paper · pdf · doi:10.48550/arxiv.1904.07162
openalex publication_date 2019/04/15 · openalex created_date 2022/07/29 · openalex updated_date 2026/07/28
Intel Optane DC Persistent Memory (Optane PMM) is a new kind of\nbyte-addressable memory with higher density and lower cost than DRAM. This\nenables the design of affordable systems that support up to 6TB of randomly\naccessible memory. In this paper, we present key runtime and algorithmic\nprinciples to consider when performing graph analytics on extreme-scale graphs\non large-memory platforms of this sort.\n To demonstrate the importance of these principles, we evaluate four existing\nshared-memory graph frameworks on large real-world web-crawls, using a machine\nwith 6TB of Optane PMM. Our results show that frameworks based on the runtime\nand algorithmic principles advocated in this paper (i) perform significantly\nbetter than the others, and (ii) are competitive with graph analytics\nframeworks running on large production clusters.\n