I previously use Smart as destination. Having found bad fills, I tried to use Island. However, the fill is no better. Do you know how to configure Smart to send to the most-volumed exchange?
As @sle mentioned, don't expect it to help much, but here it is. You actually may benefit by including routing to Dark pools ... no easy solution to the problem.
do you know what kind of latency the Lightspeed/Lime is, for both quote and execution, for US stocks?
Latency is very slippery subject. You have cross instrument strategy (different geographical location) which makes it more difficult. I can't give you actual numbers, but I think half millisecond is easily achievable (for equities with Colo in New Jersey). Of course, at this speed, you will be light years behind your competition. I don't mean to be negative on you, it's just the reality as we are far from the early stages of this now and the low hanging fruit had been mostly picked.
Each software is different and your ping times will vary based on your location and internet providers. The fastest would be the Lime Trading Gateway-this is rated ultra-low latency and used by other broker dealers, hedge funds and prop firms. This is a hosted solutions. https://www.lightspeed.com/technology/trading-software/. VERY FAST. Lightspeed, Sterling Trader Pro, Realtick and Silexx are all desktop applications with available APIs. You would need to ping their servers from live software to get ping times. I'm currently home-about 45 miles from the NYSE data center and 800 miles from Chicago-these are my current ping times on the Sterling server. (QCS-Options is in Chicago). The second picture is from my NYC office. As you can see, some is distance and some of the delay is from your internet provider and hops to the server. The NYC office is faster, not just from distance.
I got some data on the latency of IB quotes. Using their tick-by-tick market data request, the return data is about 26 millisecond delayed, including my VPS's 2 ms delay to their server. It is done by comparing the timestamp of the returning data and local PC's clock. However, its time stamp has only precision to seconds. So, the latency is only an estimation. What is your statistics?
This is making the assumption that every broker receives raw date from the exchange, normalizes it and delivers this in the same process and with the same delay as the raw data is made available. There is no easy process to test the delay from raw to delivered at the quote server, but each broker is different.