Mitigating clearing latency and slippage in high-volume MT5 bridge integrations
We have been scaling our retail FX flow over the last two quarters and are hitting a wall with our bridge provider during high-impact news events (primarily around non-farm payrolls and CPI releases on $EURUSD and $GBPUSD). Even with tier-1 prime-of-prime feeds, our execution latency at the bridge level is spiking from a baseline of 4ms up to over 180ms during peak book depth updates. This is causing unacceptable slippage on limit fills, which we are currently absorbing to keep our B-book clients happy, but it is eating into our net margins. I am trying to figure out if we should bite the bullet and develop a proprietary FIX engine wrapper in-house, or if there is a specific multi-asset liquidity aggregator on the market that actually handles raw L2 market data parsing without choke points under high-concurrency scenarios. For those of you running a hybrid A/B book model, how are you benchmark testing your bridge throughput under synthetic load, and what is your tolerance threshold before you reroute flow to secondary LPs?
Are you sure the bottleneck is actually the bridge itself and not the MT5 gateway queueing up requests during those spikes? I've heard some groups had to switch to direct FIX connections for their heavy flow because the MT5 architecture just couldn't handle that level of concurrent updates.