STAC Report: kdb+ 4.0 sharded across Optane PMem under STAC-M3 (tick analytics)

Solution with 57 TiB of data in Optane PMem sets records

29 October 2021

STAC recently performed STAC-M3 Benchmarks on the first stack to hold five years of data (57 TiB) in Optane Persistent Memory (PMem). The solution leveraged kdb+ sharded mode to distribute the data across 10 nodes.  The results are available here.

STAC-M3 is the set of industry standard enterprise tick-analytics benchmarks for database software/ hardware stacks that manage large time series of market data ("tick data"). In this project, we ran both the baseline benchmark suite (code named Antuco) and the optional scaling suite (code named Kanaga).

The stack under test (SUT) was KX’s kdb+ 4.0 database system configured to run in sharded mode on a 10-node cluster of 2-socket DELL PowerEdge R640 servers, each with 2 x Intel® Xeon® Gold 6240L (Cascade Lake) 18-core CPUS @ 2.6 GHz and 6TiB Intel® Optane DC Persistent Memory 100 Series. The PMem was configured in Storage over App Direct Mode with Filesystem-DAX namespaces. The SUT used Compatibility Revision H of the kdb+ STAC Packs. Intel chose to highlight that this solution:

• Outperformed all publicly disclosed results in 16 of 24 STAC-M3 Kanaga benchmarks
  • 7 of the benchmarks had speed-ups exceeding 2.2x compared to the previous best result
    (STAC-M3.β1.1T.YR{1,2,3,4,5}VWAB-12D-HO.TIME,
    STAC-M3.β1.50T.YR{1,2}VWAB-12D-HO.TIME)
• Was faster in 20 of 24 Kanaga benchmarks than a solution involving a parallel filesystem with 14 database servers and 18 storage servers, kdb+3.6, and a previous version of the kdb STAC Pack (SUT ID KDB200401), including:
  • 57x to 60x speed-up in the four 10-user market snapshot benchmarks
    (STAC-M3.β1.10T.YR{2,3,4,5}-MKTSNAP.TIME)
  • 3.9x to 7.0x speed-up in the 5 single-user 12-day VWAB benchmarks.
    (STAC-M3.β1.1T.YR{1,2,3,4,5}VWAB-12D-HO.TIME)
  • 4.8x speed-up in the 50-user year-one 12-day benchmark
    (STAC-M3.β1.50T.YR1VWAB-12D-HO.TIME)

For details, please see the report at the link above. Premium subscribers have access to the code used in this project as well as the micro-detailed configuration information for the solution. To learn about subscription options, please contact us.