layout: post title: “Crail Storage Performance -- Part I: DRAM” author: Patrick Stuedi category: blog comments: true

Hardware Configuration

The specific cluster configuration used for the experiments in this blog:

  • Cluster
    • 8 node OpenPower cluster (for Crail)
    • 2 node X86 cluster (for RAMCloud)
  • OpenPower Node configuration
    • CPU: 2x OpenPOWER Power8 10-core @2.9Ghz
    • DRAM: 512GB DDR4
    • Network: 1x100Gbit/s Ethernet Mellanox ConnectX-4 EN (Ethernet/RoCE)
      • RDMA send/recv latency, ib_send_lat (RTT): 3.1us
      • RDMA read latency, ib_read_lat (RTT): 2.3us
  • Software
    • RedHat 7.2 with Linux kernel version 4.10.13
    • Crail 1.0, internal version 2842
    • Alluxio 1.4
    • RAMCloud commit f53202398b4720f20b0cdc42732edf48b928b8d7

Anatomy of a Crail Data Operation

Sequential Read/Write Throughput

crail.buffersize     1048576
crail.slicesize      524288

Random Read Latency