Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

As far as I understand it, firms that colocate their servers with a financial exchange do so only at the primary site. (Correct me if I'm wrong).

So what happens if the primary site goes up in flames and operations switch to the DR site? Are colocated servers then effectively excluded from trading, or how does this work in reality?

share|improve this question
If the primary site goes down, the exchange is almost certainly closed anyway... –  Michael Hampton Sep 3 '13 at 2:18
@MichaelHampton Basically. –  ewwhite Sep 3 '13 at 2:26
@MichaelHampton I guess they can do this if extreme weather or similar affects everyone in the area, not just one exchange, but most or all exchanges, and more importantly, most local customers who won't go to work because of the weather. A fact that your article btw specifically mentions. But what if only one exchange's primary DC goes up in smoke, isn't that when they would fail over to DR? –  Eugene Beresovsky Sep 3 '13 at 6:46

2 Answers 2

up vote 4 down vote accepted

Yes, I've co-located in the facilities of the respective exchanges... There are many reasons for this, with latency being the primary driver. But access, comms lines, cost and other factors come into play.

As far as DR, that has never come into play. I've never thought about it from a trading strategy perspective. Why? Well, there will be some internal redundancies from the exchange/matching engine/feeds... If the primary site goes, trading more or less halts. These are not necessarily 24/7 critical computing environments. Maintaining stability during the market day is what's important.

In reality, I've had systems go down in a colo and definitely experienced line failures that have taken us out of the market. It is what it is. Your trading strategy needs to incorporate that risk and have the safeguards to handle those scenarios (and a panic button).

As for even gaining access to an exchange's DR site, it may not even be possible.

More details will help us help you, though. What are the players/organizations involved here?

share|improve this answer
This is a general question, not particular to any specific exchange. Re: gaining access to an exchange's DR site, it may not even be possible. Well, non-colocated line handlers would switch to the DR site, wouldn't they? How else do orders end up in the DR site? I've implemented line-handlers for reporting (not order submission), and in that case I would switch to DR if primary is inaccessible. You say if primary goes, trading more or less halts. Then what is the purpose of the DR? –  Eugene Beresovsky Sep 3 '13 at 3:20
I don't think DR is a consideration. I'll dig a little and see if things have changed much since I left the industry. –  ewwhite Sep 4 '13 at 3:04
The (reporting) line handler I was talking about was done in 2007. And considering ongoing low-latency-fication of the industry (which isn't really an issue when doing reporting), I'd doubt in recent times they'd consider adding load/latency by talking to DRs, albeit asynchronously. Still doesn't answer my question: Why do they invest in DRs when they don't use them. Of course extreme weather events, like mentioned by @MichaelHampton are somewhat different from disasters affecting just one DC - is the latter maybe a case where they'd switch over? Or wouldn't they because of the colo customers? –  Eugene Beresovsky Sep 5 '13 at 1:50

From my understanding, the benefit of coloing at or near an exchange is that you have extremely low latency, which is necessary for high frequency trading. By failing over to a DR site, you will lose this benefit which may or may not impact that firm's ability to successfully execute HFT. I'm not sure that there's any actual requirement to be colo'd in a specific building to trade.

I'm sure @ewwhite will show up here and write 20 pages on this for you, so stay tuned.

share|improve this answer
Nah, you can trade wherever. There was a big push to get cabinets at NYSE's Mahwah, NJ facility. $10k/month/cabinet... and I was limited in how I could connect Chicago to Mahwah. It was a racket... so we took a latency hit and placed systems at another datacenter, but had more flexibility as a result. –  ewwhite Sep 3 '13 at 2:28

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.