Pure STACs up benchmark wins against Dell, WekaIO, AWS

Pure Storage STAC-M3 benchmark results show its all-QLC flash FlashBlade//S500 beating Dell PowerScale and Weka/AWS when running the Antuco and Kanaga test suites.

Update. Pure comment on storage capacity in its STAC configuration added. 22 Jan 2024.

The STAC (Securities Technology Analysis Center) benchmarks are designed and run specifically for the financial services industry, with audited results by the STAC Benchmark Council, which consists of over 400 financial institutions and 50 vendor organizations. They are rigorously controlled with extensive documentation and none of the potential extreme tuning dodges used in some other storage benchmarks. There are two main test suites looking at time series and tick-by-tick data: The baseline Antuco suite with a year’s worth of simulated stock market data as its dataset, and the scaling suite Kanaga, with five years of simulated stock market data.

Diane Saucier.

Diane Saucier, Pure’s Financial Services Director, blogs: “In the fast-paced world of high-frequency and quantitative trading, every microsecond counts. Financial firms are constantly on the lookout for data platforms that offer better performance, scalability, and reliability for their quant analysts, data scientists, risk managers, and others supporting their trading teams.”

She talks of “groundbreaking results” with this first system Pure Storage has submitted for a STAC-M3 audit, saying it “demonstrate that our customers can achieve maximum performance and simplicity when choosing Pure Storage as their time-series or tick-by-tick market data platform.”

The Pure STAC benchmark test can be found on STAC’s website, reference KDB231122, and its test stack featured:

  • KX Systems kdb+ 4.0 DBMS
  • NFS v3
  • FlashBlade//S500 chassis with 10 x Flashblades with 2 x 24 TB flash modules each – totaling 480 TB – 266 TiB (292.5 TB) usable.
  • 8 x Dell PowerEDge R740xd servers each with:
  • 2 x Intel Xeon Platinum 8260 CPUs
  • 256GiB memory
  • Compatibility Rev 1 of the STAC-M3 Pack for kdb+ (benchmark implementation code).

Pure said that, in its results, compared with a Dell PowerScale all-flash system:

  • 13 of 17 STAC-M3 Antuco mean-response time benchmarks witnessed improvements, including:
    • 7x speedup in the 10-user theoretical P&L (STAC-M3.β1.10T.THEOPL.TIME) 
    • 5.8x speedup in 10-user market snapshot (STAC-M3.β1.10T.MKTSNAP.TIME)
  • 17 of 24 STAC-M3 Kanaga mean-response time benchmarks exhibited enhancements, such as:
    • 1.3x-1.5x speedup in 50-user 12-day VWAB (STAC-M3.β1.50T.YR{1,2,3,4,5}VWAB-12D-HO.TIME).
    • 1.2 – 1.4x speedup in 1-user year high bid (STAC-M3.β1.1T.{2,3,4,5}YRHIBID.TIME)

The Dell PowerEdge server + all-flash PowerScale file storage result – KDB220506 – featured: 

  • KX Systems kdb+ 4.0, running in distributed mode
  • 3 x Dell EMC PowerScale F900 All-Flash Scale-Out NAS nodes for 251 TiB (277.2 TB) total physical capacity
  • 8 x Dell PowerEdge R7525 server, each with:
    • 2 x 64-core AMD EPYC 7H12 CPU @ 2.6GHz
    • 512 GiB DRAM (549.8 GB)
    • NFS v3
  • Compatibility Rev H of the STAC Pack

When compared to a WekaIO/AWS system, Pure’s results “outperformed in 9 of 17 Antuco benchmarks and 12 of 24 Kanaga benchmarks, showcasing significant speed advantages.”

Pure said it was faster in:

  • 9 of 17 Antuco mean-response time benchmarks, including:
    • 11x speedup in 10-user market snapshot (STAC-M3.β1.10T.MKTSNAP.TIME)
    • 8.4x speedup in 10-user volume curve (STAC-M3.β1.10T.VOLCURV.TIME)
  • 12 of 24 Kanaga mean-response time benchmarks, including:
    • 4.4 – 5.3x speedup in 10-user market snapshots (STAC-M3.β1.10T.YR{2,3,4,5}-MKTSNAP.TIME)

The WekaIO/AWS result – KD210507 – featured: 

  • KX Systems kdb+ 4.0 in distributed mode
  • Amazon Linux 2
  • DB cluster: 15 x Amazon EC2 C5n.9xlarge, each with 36 vCPUs using Intel Xeon Platinum 8124M CPU @ 3.0GHz
  • Weka IO WekaFS 3.10.1
  • Storage cluster: 40 x Amazon EC2 i3en.6xlarge, each with 2 x 6.82 TiB (7.5 TB) SSD and 24 vCPUs using intel Xeon Platinum 8259CL CPU @ 2.5GHz totalling 545.6 TiB (599.9 TB)
  • Compatibility Rev E of the STAC Pack

We noticed there were considerable differences in the storage capacity of the three systems tested:

  • Pure – total 480 TB with 292.5 TB usable
  • Dell – total 277.2 TB
  • WekaIO/AWS – total 599.9 TB

THis will affect the price/performance characteristics of the systems tested. A Pure spokesperson said: “Pure has extensive experience supporting high performance tick analytics and time-series based workloads, empowering the financial sector to make the most of their data for over a decade. The STAC- M3 test focuses on performance results and configuration plays a significant role in both outcomes and associated cost factors for firms to consider when determining which data solution best fits their requirements.  

“As stated in the STAC configuration disclosure, we chose a baseline configuration that specifically highlights our solution in terms of being able to achieve groundbreaking performance.  We specifically conducted the test with a conservative configuration to demonstrate that there is ample room for capacity and performance growth in a single chassis. When using this FlashBlade//S500 configuration, firms have the additional benefit of effectively quadrupling capacity through expansions within the same chassis or scaling out with additional chassis to meet the current and ever growing demands of their tick analytics workloads.”

You need to be a registered STAC member to be able access the full details of the three test runs covered here.