ToDo/Block/Qcow2PerformanceRoadmap

From QEMU
Revision as of 15:01, 23 November 2010 by Stefanha (talk | contribs)

The following FFSB benchmark results model QCOW2 performance:

  • QCOW2 (2010/9/22) is the upstream QCOW2 code including Kevin's work to avoid flushes when possible.
  • QED with L2 Flush is QED with a flush on L2 update, the best QCOW2 can get since a flush is required after refcount update.
  • QED is the QED v1 patchset, a fully asynchronous block driver implementation.

All results are throughput in MB/s.

FFSB Scenario Threads QCOW2 (2010/9/22) QED with L2 Flush QED
Large File Creates (Block Size=256KB) 1 95.1 101.0 121.5
8 110.0 95.2 125.0
16 89.4 96.8 134.5
Sequential Reads (Block Size=256KB) 1 68.2 154.0 149.0
8 262.0 849.0 817.5
16 276.0 908.0 922.0
Large File Creates (Block Size=8KB) 1 11.8 13.9 16.3
8 18.4 29.7 41.0
16 17.7 30.2 43.3
Sequential Reads (Block Size=8KB) 1 20.8 26.3 22.2
8 13.8 136.0 131.5
16 13.8 160.0 160.0
Random Reads (Block Size=8KB) 1 3.8 3.5 3.6
8 13.8 26.0 25.7
16 13.7 48.8 48.4
Random Writes (Block Size=8KB) 1 22.6 22.5 22.6
8 14.0 113.1 112.7
16 14.0 125.2 126.7
Mail Server (Block Size=8KB) 1 7.8 8.8 8.3
8 10.8 45.8 46.4
16 11.8 72.0 73.2
Mixed I/O (70% Reads, 30% Writes, Block Size=8KB) 1 10.2 9.5 8.8
8 24.8 63.5 62.9
16 28.6 114.1 111.9

O_DIRECT benchmark execution. Image file on ext4 on LVM on 8 LUNs.