What my trading algo does is placing limit orders and preferably complete the trade after a fill in less than 30 seconds. Obviously, getting my limit orders filled is the crucial part of this strategy. I thought this means being as fast as possible, but maybe, after reading all the replies, I was wrong.
30 seconds? dude, the hft algos have traded 52,000 times in 30 seconds. Made more in that time then you will in a year, if you're good. what you want is to do is stay out of their way, not try to compete with them.
TT, TWS, OEC and just about every trading platform on the market can handle your trading requirements. Most platforms support automated trading and signal processing offering demo accounts. Load a few up and see which ones are easiest to load your algo.
The success of my trading algo is critically depending on how fast limit orders are filled, for example when the prices touch the limit price just briefly. Would co-location be an advantage compared to running the algo over the internet?
Depends on what you trade. Read up on CME's matching algo and price banding. Most of their products trade FIFO. http://www.cmegroup.com/globex/introduction/ Other exchanges operate differently. For CME you always have the option to send market orders which are actually limit orders with protection. When you are co-located you get a low latency fiber connection feeding you up to 70 updates per second per instrument. Over the internet you have 15 - 200 ms delay. Advantage has their act together technically - 3ms located at equinix Chicago. Sounds like you are discretionarily scalping rather than running a HF Algorithmic trading system.
I wasn't aware that the term HF is exclusively related to the ms realm; I thought everything under 1 min would be called HF. Thanks for the education and your reply; I think that's what forums are for too.
HFT is a bastardized term... my point was based on your time frames co-locating is probably an overkill. I'd focus on streamlining your order execution and if your system can take advantage of higher frequency data than co-locate. For example: you are trading 50 lots of NQ. Higher frequency data allows you to monitor level II and build your own "executable" market feed based on this size. Your computers monitor the limit order book and maintain calculated ask and bid price to fill 50 lots. Using this higher frequency calculated feed you are able to precisely time your orders to get in and out instantly while filtering out much of the noise. Charting this 50 lot feed will produce a much different look of the underlying contract.