It was RT. On second thought I think it was 33ms to Chicago and 20 or so to NYC. In any case, there was some latency. RT from my FiOS connection was in the 90s from western pennsylvania. I haven't looked into google app engine.
The broker is not going to trade something, unless they have a local bid. And if you are trading anything of volume it means you can't trade. Regarding the comment of products going offline, and not exchanges, well it is the same thing. If you are trading say NDX, and NDX goes offline the exchange might as well go offline since you are not able to trade it. Regarding CBOE, it was SPX that recently (last week) I could not trade for about 30 minutes. And for those 30 minutes I thought, ok time to get a coffee... As sappjason said, 95% of the time you are humming... This brings me back to my original point. Having redundancies, and thinking about cloud computing does not make sense. Since the exchange to start off will be flaky and hence your application will be flaky and trying to compensate that is a waste of money. My solution to use a home DSL, with backup cell network has worked out exceedingly well... And it is inexpensive to boot...
As has already been pointed out by Aitch Eff Tee, this is very very different to an exchange or product going offline. You are talking about liquidity issues. No it is not the same thing. From your perspective they might have the same impact on your trading strategy but you can't expect to post a comment saying that you see exchanges going offline for hours every day without someone calling it out as bullshit. What you said was factually incorrect. The SPX has been a joke for a long time. Most the volume is institutional and done off exchange. The CBOE screwed up with this instrument many years ago by not allowing any competition. Depending on your trading needs you might find the SPY options a better bet... Best of luck.
If we are going to mince words I was somewhat incorrect in the initial phrasing, which I clarified later. I clarified that I regularly get messages like "exchange A will not be able to trade products A to J". I did not say hours at a time, and I do see these messages very very often. The reason is because with IB I am subscribed to about 30 exchanges. Thus the probability that an exchange will have problems is pretty high. And it is no BS regarding the problems with the exchanges. Everybody assumes that exchanges just work and you will never have problems. Take a look at the following: http://batstrading.com/alerts/ See all of the alerts and the problems? Now imagine being subscribed to 30 exchanges? I see lots of messages, and it is a bit annoying because IB keeps pinging.. I am not even including the market data issues, which are quite common as well. As was said previously 95% of the time it does not affect you. But sometimes it does affect you. And if your algorithm or trading style assumes no exchange problems then you are asking for trouble. I used to run a trading algorithm that traded based on second increments, and keeping the runtime data clean was a HUGE problem. In the end what I ended up doing was getting every 15 minutes a historical tick data refresh. Otherwise I would have skew where I would miss trades or enter a trade on bad data. The problem was not that I overfitted my algo, but when the market moves fast data might come to fast or too slow. My point is that if you are trying to minimize round trip time, or keeping the system running without having a running system you are wasting your time. The hardest part with a 24x7 trading system is making it robust to deal with a noisy market data... If you figure that out you will make money... Christian
Nobody with an iota of sense assumes this. Again. Nobody with an iota of sense assumes this. Data cleansing is an ongoing problem for everyone. If you are trading pit contracts then your problems will be exponentially greater than trading purely electronic contracts. Anyway...I have no wish to piss on your fire. I just felt that I had to question what I saw as such an outlandishly incorrect statement. Best of luck with your trading...
I only had to wait, oh 37 minutes before the first outage occurred in the American exchanges... http://www.happyasahippo.com/images/exchangeproblems.jpg
The Eagle has landed. I decided to dedicate some time to this and after a couple of tries, I was able to launch a small windows 2008 server instance and Remote Desktop into it. Looks like I got one core and about 30GB of disk space. If I save market data (for example, an NxCore image), how do I make sure that I don't lose that data in the future? I mean, once the instance is gone, don't I lose what was saved to the local disk of that machine? I suspect there is some way to attach personal storage dedicated to me to each instance, but I haven't delved that far into it yet.
After you have the installed software you need installed, you bundle a new instance that is yours alone. You can save data without bundling an instance (which seems to take like 20-30 minutes or so(. To do this you create a volume and attach it to your instance. Also, you can lease a static IP and map it to a hostname (using your own DNS services). This makes it nice so you can remote desktop to the same hostname.