At Chronicle our background is in consultancy and we have a strong consultancy and development team with on average over 20 years in experience, led by Peter Lawrey a Java Champion with the most answers on Stack Overflow, we have been able to offer significant latency gains to client infrastructure regardless if they use Chronicle open source, Enterprise or no Chronicle Software at all:-
1) Tier 1 Bank- Recently we completed a three week engagement to provide some performance tuning consultancy with a Tier 1 Bank based in London this was done through our partnership with MThree. The objective was to improve the longest delays in the OMS with minimal changes to the underlying framework and the key challenge was to identify where the delays were. After much analysis the main source of the delay was identified and with a number of configuration changes we saw a factor of 25 reduction in latency for the 10 most common tasks and a 40% reduction in latency overall
2) Hedge Fund- We were retained for a 3 day workshop in New York in early February, they were users of Chronicle open source Queue and were now looking to upgrade to Queue Enterprise, in conjunction with the upgrade they asked Peter to provide training plus insight on the most efficient way to deploy Queue Enterprise. The overall engagement proved very successful for the client with an initial read/write time of approx. 260bytes being 173 milliseconds they ended the workshop at the conclusion of the workshop this was reduced to 6.5 microseconds an improvement of a factor of over 20,000
We would be delighted to discuss how our team of experienced consultants can help your business
There are two major factors behind our development of a C++ version of our products, one being interoperability between the two languages allowing for seamless data flow and the second one being the latency improvements that can be gained by deploying C++. As part of the release cycle for Chronicle Queue C++ we have carried out extensive benchmarking between the two languages which we are happy to share below is just one example:-
The following test writes 48-byte messages at a sustained rate of 100K messages per second, the plot shows the end to end latency between a writer updating the queue and the change being visible to a reader on the same host. The latency axis is a logarithmic scale
This illustrates a significant performance gain across the whole benchmark and we can see specifically once we get past the 99.9% percentile that the performance improvement is by a factor of around 100.
We have several other performance tests including single write , bulk write, read/poll etc. which we would be happy to discuss in further detail.