[thibautx]

I would say that even options/futures/and other products markets are just as hard to market make compared to the equities markets. In my opinion, you need some sort of edge not through hardware but through your model or your market outlook, combined with market-making/hft/statistical-arbitrage in order to be profitable and compete.

1) Capital in the magnitude of >$500k seems reasonable for a small operation

2) Lime Brokerage (a friend has started a hft strategy with a professor on campus and has tested out lime) is okay, I'd look into some other execution services as well

3) Yes you should be backtesting on tick-level data from multiple exchanges ideally. Be wary of lookahead bias and try to build some sort of execution model to account for how your executions might actually be filled and how your orders might affect the market microstructure

Ultimately, I wouldn't rely on pure hft and latency sensitive strategies...latency is always important but you should try to find your edge elsewhere.