Quantitative analysis is easy for latency and performance alone, but not so much for productivity. Since my goal is performance over limited amount of engineering effort, I didn’t do a thorough analysis (which itself will take effort).
However, I can share that my HFT program in Java is definitely one of the fastest on those exchanges (and I suspect that I might be the only one using Java).
Thanks, Zheng, for this post and sharing with us your insights!
As an outsider of the area of HFT (but a Java programmer for a long while), I am curious if there is some (high-level) quantitative analysis on the competitiveness of Java based HFT systems, comparing to systems using other languages.
Well, one day it could just be natural human language thanks to ChatGPT and LLMs :-)
Quantitative analysis is easy for latency and performance alone, but not so much for productivity. Since my goal is performance over limited amount of engineering effort, I didn’t do a thorough analysis (which itself will take effort).
However, I can share that my HFT program in Java is definitely one of the fastest on those exchanges (and I suspect that I might be the only one using Java).
Got it! Thanks for the additional note!
Thanks, Zheng, for this post and sharing with us your insights!
As an outsider of the area of HFT (but a Java programmer for a long while), I am curious if there is some (high-level) quantitative analysis on the competitiveness of Java based HFT systems, comparing to systems using other languages.
Well, one day it could just be natural human language thanks to ChatGPT and LLMs :-)
Very insightful !!! Thanks for sharing Zheng