MDA protocol and library improvements
The CMF performance has been enhanced by various MDA protocol and library improvements. The following throughput and latency improvements were recorded in laboratory conditions:
Throughput (orderbook updates/second) | CMF 5.1.1 | CMF 5.1.1 | CMF 5.2 | CMF 5.2 |
| 1 context | 4 contexts | 1 context | 4 contexts |
Windows | 3880 | 7140 | 6650 (+71%) | 16400 (+130%) |
Linux | 4590 | 10160 | 6460 (+40%) | 25580 (+151%) |
Latency (orderbook milliseconds 99th percentile) | CMF 5.1.1 | CMF 5.1.1 | CMF 5.2 | CMF 5.2 |
| Main context | Non-main context | Main context | Non-main context |
Windows | 0.82 | 0.87 | 0.60 (-27%) | 0.60 (-31%) |
Linux | 0.81 | 0.85 | 0.60 (-26%) | 0.60 (-29%) |
The main/single context results show the performance improvements due to protocol and library improvements.
The non-main/multi-context results also show the performance and scaling improvements due to the use of channels.
These measurements were recorded on:
Windows machine specifications: Intel® Xeon® CPU E3-1270 V2 @ 3.50GHz 8 Cores, 16Gb RAM, Windows 7 64-bit
Linux machine specifications: Intel(R) Xeon(R) CPU E5-2690 @ 2.90GHz 32 Cores 64Gb RAM, RHEL 6 64-bit