Crypto HFT - In Depth Guide to Optimisation (III)
In the first post, we discussed where latencies come from in cloud-hosted infrastructures for crypto hft mm. We discussed how ALBs can add jitter
and how to find the best ‘front door’. Upon finding the right endpoint (if any) - we discussed AWS box placement strategies and search heuristics to increase probability of reducing the physical distance to server racks, such as using partition and spread placement groups.
We want to round this up with some more AWS-native notes, and talk a little bit on the importance of network topology and writing a custom gateway for multi-asset/regional arbitrage. I wanted to include some notes regarding on-prem server tuning, but I think it’s not really crypto-specific MM so I will put that in a separate post.