We continue to grow our client base most recently securing a major North American financial organisation who have signed up to take Chronicle FIX to replace their current incumbent supplier. Chronicle FIX was benchmarked within the client environment against several other FIX Engines and was shown to be consistently faster, other contributing factors to the decision included the simplicity of the Chronicle FIX pricing model and the provision of source code that enables greater flexibility in development and deployment.
Normally we would start an engagement with a short workshop which helps us to understand the client environment and what they are looking to achieve, however in this instance that wasn’t the case but we were given a lot of transparency into the client test environment which meant that during the pre sales process we were able to ensure that the performance of Chronicle FIX was optimised.
This continues a very successful year for Chronicle FIX in which we have expanded our installed base including a Tier 1 bank who are currently in the process of deploying Chronicle FIX globally .
Benchmarking
As part of our demo repository for Chronicle FIX we provide several benchmarking scenarios these included multiple clients to multiple endpoints, warm up benchmarks and a standard performance benchmark
The performance benchmark initiates a FIX acceptor (server) and a FIX initiator (client) and transmits messages between them. At the end of the test a comprehensive set of statistics are generated are measuring :-
End to end – time from NewOrderSingle being generated in the initiator, and initiator reading the ExecutionReport response from the socket
End to end latencies for 20K/s on a i7-7820X CPU @ 3.60GHz
Percentile run1 run2 run3 run4 % Variation
50: 12.77 12.80 12.81 12.76 0.25
90: 13.44 13.40 13.43 13.40 0.16
99: 15.57 15.42 15.23 15.16 1.15
99.7: 19.34 18.89 17.99 17.75 4.09
99.9: 21.42 20.06 19.67 19.56 1.66
All timings are in microseconds
As the table shows that the mean time is around 12 microseconds and even at the 99.9 it is around 20 microseconds, a full breakdown of all the processes within this benchmark can be found on our GitHub site plus you can get access to our Java Latency Benchmark Harness (JLBH) which is a tool built by Chronicle for performance testing and benchmarks.
For more information about Chronicle FIX and access to our GitHub repositories contact andrew.twigg@chronicle.software
