Slashdot Mirror
Wall St. Trading Servers To Power Off-Hour Clouds?
miller60 writes "As cloud computing gains traction, some Wall Street firms running armadas of servers to power high-frequency trading operations are contemplating leasing out their excess computing capacity after the trading day ends at 4 p.m. 'Once 4:30 rolls around, we don't need those machines,' said one CTO of a market data firm. 'There may be an opportunity there.' A similar revelation led to the creation of the cloud computing operation at Amazon.com, which built its infrastructure to handle peak Christmas-season loads that lasted just a few weeks each year."
We've already got Virutal Machine tech and several other ways of saying "Don't cross this line!" so unless it's implemented by an idiot I'd say we'll be okay.
Location is the reason these system have spare cycles. They build data centers across the street from the market so that they have the lowest latency access to the market (it matters for high frequency trading). A data center on Wall Street is useless for other markets.
" nobody needs their clock cycles 24/7"
What about computers the service cloud computing?
The Kruger Dunning explains most post on
These High-Frequency trading systems require sub-second responses from the market servers. They're usually collocated in the same structure as the marketing servers, and firms pay handsomely for that kind of space and network access.
When that market closes, you don't want these machines trading overseas, the latency to reach those servers would negate the entire purpose of these machines.
Why aren't you encrypting your e-mail?
All trading instructions, no matter how technologically implemented, come from people. Somebody has to write and fund the program that says "Buy XXXX when its price reaches $Y" even if they're not attending it at the time it happens.
Just look at what a mess after-hours markets are. Sometimes they're offering tomorrow's price today, but sometimes they get bent out of shape. Don't you dare buy a stock Jim Cramer promotes on Mad Money in after hours... somebody who owns that stock would love to jack up the price on you. Worse yet, somebody can short that stock and likely have a profit by 9:30am the next business morning.
If you want 24/7 markets, try currency trading. The US Dollar is up for trades at nearly all hours of the day.
Is letting 'cloud users' access the servers that run out financial markets really a good idea?
No.
Citation needed.
The NAS and compute cores just need to be separate systems
Is letting 'cloud users' access the servers that run out financial markets really a good idea?
No.
Citation needed.
http://www.answers.com/topic/common-sense
As a former developer from a very successful HF trading firm, I can tell you that's not how it works. You are forgetting one crucial item, which is that there is a spread in the market. At any time, the price at which people are willing to sell, and which other people are willing to buy, is different. I'm willing to sell a share for $3, but you only want to pay $2 to buy it. (This spread is unrealistically wide to make it easier to illustrate the point). Unless one of us changes our mind and is willing to sell for less or buy for more, there's no transactions going on.
There are many ways of doing high frequency trading, but here's one common way. The high frequency trader, call him T, steps in and says that he willing to sell it for $2.75 and buy it for $2.25. I think, hmm, I would really like to get $3 for my share, but unless I can sell it, I'm not making any money. $2.25 is better than $2, so I'll sell to T for $2.25 instead of to you for $2. Two things have happened here: Now I'm able to sell my share, and I wasn't able to before because $2 was just too low. Also, I got a better price than if I bought from you.
T now owns 1 share, and offers to sell it for $2.75. You decide that you really need to buy a share. At least $2.75 is a better price than $3. It is not as good as the $2 price you really wanted, but you can't always get the price you want, so you buy the share for $2.75. You got a better price from T than you would have gotten from me.
T gets $0.50 for his trouble. It is not risk free, but I'll get to that later.
One way of looking at this is that T is a parasite. After all, you could have been willing to buy for $2.25 and made the transaction. But the point is that you weren't willing to do it. You were free to offer 2.25, but you didn't do it. I was free to sell for $2.75, but I didn't do it. Without T's intervention, you and I could have stared at each other for days without buying or selling.
Thus, the other way of looking at it is that T made the market more liquid, and decreased the spread between bid and ask. I made more than I would have without T, and you paid less than you would have without T.
T's transaction is not without risk. For instance, the market may move the wrong way, leaving T to take a loss. I may sell my share to T for $2.25, but then the market drops, and no one wants to buy it for $2.75. T drops his price to $2.65. Still no takers. He drops it to $2.55. Still can't sell. At this point, he can either chase the price down bit by bit, which is dangerous because it can lead to a huge loss, or give up now and sell at whatever price buyers name, which is probably more like $1.80 at this point. Instead of a $0.50 gain, he takes a $0.45 loss. There are other ways to loose money, too.
Also, T is not alone in the market. Other people are doing this, too. Someone will see T's juicy profits, and offer to buy for $2.30 and sell for $2.70, taking less profit, but getting the business instead of T. But someone else will offer to buy for $2.40 and sell for $2.60. The high frequency traders pile on, and the spread shrinks until it is so small that only the most efficient traders can make any money, and they usually make only fractions of a cent per share. As a retail investor, this is great. With more competition, I can sell for more and buy for less than without these guys.
Your wrong. The stock market has a profound impact on everyone's lives throughout the world. While it doesn't impact us directly, it does through many other financial abstract layers. Everything from your IRA, savings, to corporate reinvesting which including hiring of new employees to expand a business.
Have none of you learned from the Great Depression. Take away the stock market and watch what happens!
Life is not for the lazy.
1) The trader T is virtually alone in the pre 1 second market, when compared with the size of the original market. Thus the original market has been split into two markets, one market for a small number of privileged players and one market for everybody (privileged and unprivileged players). That always leads to suboptimal prices. To get optimal prices in the market, it's necessary to coalesce these two markets.
2) The idea that making an extra sale is always better than not making a sale is wrong. In the market, the nonexistence of a transaction is just as important as the existence of a transaction, so by facilitating a deal the trader T is biasing the natural outcome, which again results in suboptimal price movements.
3) You are implicitly treating T as a regular market player, when you describe T's profit and risk strategies. Regular players are good for the market, but T is not a regular player, instead T is a player in two markets, which cannot interact until we find out how to do time travel. The interaction is only one-way. The actions of T in the 1 second market lead to certain advantages in the initial portfolio of T when viewed in the later market. These advantages would not exist if T did not have the 1 second privilege, due to all the other (privileged and unprivileged) players in the market. But these other players cannot act, as they are not part of the first market. T therefore plays the role of a middle man which cannot be routed around.
4) The total costs of joining the high frequency trading markets are nontrivial, and as such represent an inefficiency in the market that results in suboptimal outcomes for the market as a whole (eg pre and post 1 second). Since it is physically impossible for everybody's servers to be collocated, the only way to fix this inefficiency is to allow nobody to perform high frequency trading.
5) The privileged (1 second) market is necessarily automated. This leads to a market in which all players are algorithms. From the economic perspective, this is again inefficient, as the algorithms are not the players, the companies are. An algorithm (especially one that needs to be fast) is only a limited representation of an economic preference function.
I would love to tell where I work or identify myself to lend credence to what I'm going to say. But I value my job too much... I work as a trade support specialist and systems engineer at an electronic trading firm dealing with many exchanges in a few countries.
We are finishing the process of scrapping Windows from over 100 machines simply because they are too slow and cannot be optimized enough at a workable cost. When speaking to some of the exchanges I have a good laugh with them about Windows machines and we definitely see poorer performance from the dark pools (exchanges with no market data) which are built on Windows. Large exchanges typically do not have Windows servers anywhere near the order entry or market data systems as a whole.
As to the notion of using the off-hours cycles, we have given a passing thought to using them for SETI or the like, but always came to the conclusion that the power bill from dozens of colocations worth of servers being run at full tilt squashes all real intent at doing this. These are some really powerful machines! Even if we charged just for electricity and cooling it would be hard to make a profit from it. Seriously, one colocation has 5x the service rating than does most people's relatively large homes. And we certainly don't run at a low load!
Well, I'm in the business too, and I'm not going to break my NDAs and specify whose MS software I service. But I do write and maintain apps for some pretty big traders (and directly related financial businesses). There's lots of MS platforms in their core business ops. Lots of Windows server farms, particularly running SQL Server and business objects. Tremendous horsepower, both in-house, and colo at telco hotels for low latency to exchanges - and at leasable datacenters. And starting to move some services to clouds. They're interested in Azure, and waiting to see what it's like when it's ready for prime time.
But their biggest obstacle is letting their data and algorithms, or anything in their critical path, live at Microsoft. If they had an Azure cloud distributed among the locations they control themselves, they might be closer to moving their apps to that model.
And if they could make money off their sunk hardware costs while it's "sleeping" (except for some hefty datamining and OLAP procs), they would. They'd sell their grandmother if it had a chance at a profit. And computing services rarely sell for less than the power and other ops costs to run them.
So there's probably a future in this. I wonder who else has seen more of it than I have yet. So far, only people who haven't have chimed in here.
--
make install -not war
Because the entire idea of the SEC is that all information is to be known by all parties at roughly the same time, it's why insider trading is illegal for instance.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Getting value out of unused resources is cool. You may recall that when AOL bought the Compuserve service, a company called Worldcom (later MCI Worldcom) bought the Compuserve network infrastructure. What is less well known is that Worldcom was originally known as LDDS (Long Distance Discount Service), a telecom company that was spun off from a gas company.
The idea was that this regional gas company (I don't recall the name) had overbuilt their urban pipeline infrastructure, and so they had all these empty tubes, and someone had the bright idea of filling them with *something*... Fiber optic lines!
Well, the board of directors and stockholders in the gas company were conservative and didn't want to get into that business, so a team from the company spun off and started their own telecom business, and bought the pipes and right-of-way from the gas company. The parent saw steady growth as most utility companies do, and of course LDDS eventually fell under good old Bernie Ebbers' control (as MCI Worldcom), and he, along with his buddies at Enron, practically destroyed the stock market around the turn of the century!