I'm currently working on a project for quant trading using IB. Today, I'm located at West Coast. My program should trade approx 50-100 trades a day (23 hour) for a single futures symbol. No, I'm not HFT. My questions are: [1] Where do I Colo or get a dedicated server for the best latency with IB. I could afford 1-2U only, no I can't afford CenturyLink in getting half a cage [2] What do you think the best latency I could get in terms of Tick Data and Order entry? In 5ms, 10ms, etc? [3] Is IB America Central Server very close to CME? [4] Does IB America Central Server route Futures order to CME directly or it still goes thru Grenwich CT ? I'm also starting to think whether IB really fit what I'm trying to do. Thanks in advance.
Does latency of, say, 50ms from west coast to the exchange really matter if your holding time is at least a few minutes?
I'm experiencing around 300ms+. I time with the PC clock against the Tick Data timestamp. I do sync time with time server. Yes, the Tick Data time resolution is to the second but you could tell. I have Comcast 1Gbps internet but ping to ndc1.ibllc.com (America East) took around 80ms. The system is not live yet but with 1/3 of second behind everybody else, I really worried about the slippage - instead of selling at ask1, I might have to pay ask2 to be sure to close long position (sell).
[1] There are several firms that offer servers in Aurora (e.g. Netsource is proximity hosted for like a 120/month) and some offer shared hosted servers for like 1-2k a month (virtualized or dedicated OMS instance). With that setup, your mileage will vary a lot, but I recon you can get flight times in hundreds of mikes or better. [2] This is like asking what's the best 0-60 you can get with "a car". You have so many variable - colocation, hardware, EMS/OMS, your own software, risk checks etc. I'd start by asking yourself "what is the highest latency I can live with without compromising my alpha?" and going from there [3,4] My understanding is that (specific to IB) colocation is irrelevant - their risk checks query a global server and that hop is going to be the limiting factor in most cases. I never traded through IB, so this is via word of mouth. If you are in any shape or form sensitive to latency, you will probably want a DMA solution with a gateway check. Just IMHO, of course.
That's very optimistic, IMHO. The speed of light is about 300 m/mike in vacuum and straight line distance from SF to Chicago is about 3k kilometers. Since the speed of light in fiber is about 2/3s of that in vacuum, just the flight time is about 15 milliseconds. Once you start adding switches, non-optical connections to the backbone and (most importantly) various hops to and fro, you probably talking 100 ms at least and probably mean of 200-250 ms. You might as well send a carrier pigeon to the exchange.
Hmm. NYLon distance is 5.5kkm, so the flight time should be over (5,500,000m / 300mus) * 3/2 ~= 27 ms. I'd be surprised if you really got 30ms ping times (here and back?). Unless I am mistaken in my math somehow.
If you are asking questions that are related to latency/colocation, IBKR is not for you. They have a (relatively) slow system and risk-checks (already mentioned). Their business model is "universal access", not single-venue speed. There's "proper" HFT, "reasonable" latency - and then IBKR/etc. Why don't you ask around some smaller Chicago FCM's - like Advantage etc. With traditional Chicago HFT's dropping dead, there's plenty discounted rack-space in Aurora. And even Equinix colo on Cermak is going to be better than IB.