Zero Latency: Making Trading Risk Checks Compliant And High-Speed
Author: Marcus Perrett
Nobody likes choosing between two unsatisfactory alternatives. For example, which would you prefer: being fired from your job or taking a 50 percent pay cut? Not much to like there. Yet that's precisely the kind of choice many trading companies have been forced to make while trading electronically — until now.
What if you could get a promotion and a raise? That's precisely the kind of "best-of-both- worlds" solution that Fixnetix, a DXC company, has recently delivered.
First, a little background. In 2008, U.S. regulators passed rules that added controls to "naked access," or trades that occurred when a firm placed orders through an investment bank's infrastructure and exchange membership. In these trades, investment banks were liable for client losses, but they had virtually no control over a client's trading activity. In response, the U.S. Securities and Exchange Commission (SEC) gave investment banks additional control over their clients' flows to the exchanges.
To accommodate this change, IT solution providers developed systems that helped investment banks gain control over clients' trades. The systems offer a choice between two kinds of solutions — passive and active — and both have trade-offs in performance and risk-checking capabilities.
Passive systems let banks either catch order elements for risk decisions or swap a symbol to change the semantics of an order. Active systems, in contrast, perform a risk check by actually intercepting trade flows and control risks on behalf of the bank in a pretrade fashion.
Both approaches add some latency, and neither meets the needs of every use case. While passive systems are the faster of the two, they may also violate some regulatory recommendations. Active systems address that concern, but until now have been comparatively slow. While the added latency time for a passive system can be as little as 250 nanoseconds, the added latency time for an active approach can reach about 1.5 microseconds, or six times as much. In today's high-frequency trading marketplace, those fractions add up fast — and they can cause traders to suffer losses or miss important deals.
Faster, Safer Trades
To create a third, better option, Fixnetix developed iX-eCute Zero Latency (ZL). It's a risk-management and trading gateway that supports electronic trading for investment banks, prime brokers, hedge funds and proprietary trading firms. The product fundamentally alters the flow of an order through the pretrade risk process, providing the client and exchange alike with additional protection against market manipulation, without a latency penalty. ZL conducts more than 40 different risk checks according to SEC, Investment Industry Regulatory Organization of Canada ( IIROC), Canadian Securities Administrators (CSA) and Australian Securities and Investments Commission (ASIC) requirements. For the first time, an active trading product independently offers both robust risk checks and high-speed order submissions, effectively removing the need or use case for passive inline systems ZL is built on the existing technology of iX-eCute, which essentially rips apart a trade order and holds the pieces up to a light. If the elements meet an investment bank's regulatory demands, the order is reformatted and submitted to the exchange. If the order doesn't meet demands, it doesn't reach the exchange at all. In other words, exchanges never see orders that would fail a risk check.
What's more, active systems such as iX-eCute keep a firm's trading history in memory. That way, if a trader sends an amendment to an earlier trade, the system can compare the two orders and check them against daily-limit and other restrictions. That's in keeping with the spirit of guidance from the SEC and other regulatory bodies, but as you can imagine, all that work takes time. That's why active approaches such as iX-eCute have, until now, imposed a larger latency penalty than passive systems.
Passive systems, by contrast, have addressed the need for extremely high trading speeds without requiring additional programming. Most do so by offering one of two approaches. In the first, the system spots the elements needed to infer a risk decision as the order flies past — a bit like trying to spy a person wearing a red shirt in a speeding train while you're standing on the platform! If an order fails the test, it still goes to the exchange, but in a malformed, scrambled state. It's as if you had somehow pushed over the last car of that metaphorical train; it's still moving, but now partially capsized.
In the second approach, passive systems change the semantics of an order, allowing the exchange to trade the order, but in a way that avoids any material impact. For example, if an exchange offers a "test" symbol, a passive system could swap out the "trade" symbol on an order for a "test," send the order on its way and then have it run as a mere test order. The problem here is that the system actually changes the client's message. That's a questionable approach in the view of some regulatory bodies. Also, clients could easily miss the fact that their order has been changed, not knowing that an order sent as a buy was actually fulfilled as a test.
Best of Both Worlds
Enter the Fixnetix ZL approach. It delivers the best of both worlds: the high speed of naked access combined with the rigorous and compliant risk-checking of active approaches. It's a trading gateway, a system investment banks can use to better manage the risk of trades conducted by clients using the bank's identification.
Built on Fixnetix's existing iX-eCute solution, the ZL gateway is nonetheless an independent system. It can be used on its own or in tandem with iX-eCute, depending on which setup works best for a bank.
ZL offers a unique solution to the challenge of reducing the latency penalty while maintaining robust risk-checking. While more conventional solutions have essentially reached the limits of physics — they really can't get much faster — with Fixnetix, we asked, "How can we move a bank's risk checks closer to the client?"
The answer, as embodied in ZL, is to actually put the risk checks in the client's server. This way, the client's system becomes self-governing. Orders that would fail a risk check in a more conventional solution are not even generated with ZL.
More specifically, the client's intention of placing an order, rather than the message itself, is passed along to ZL. In parallel with the client algorithm generating the order, ZL uses the client's spare latency to perform the actual risk check. In fact, the risk check is done at the same time the client generates the order; hence the "zero latency" approach.
The ZL approach delivers two major benefits. One is that a bank can be fully compliant in its risk checks without any latency impact. The second benefit is speed: Because ZL's architecture bypasses all standard x86 transfers — including Peripheral Component Interconnect Express (PCIe) — orders can be placed on the wire faster than clients could ever do it themselves. As a bonus, ZL operates at a "raw" (rather than market) level, so trading on different markets requires little or no additional programming. This means ZL can be deployed on either an asset-class or region basis.
With investment banks under increasing regulatory scrutiny, many in the industry have wondered where we go from here. Passive systems are clearly not the long-term answer; the SEC recently provided guidance requesting an explanation of message scrambling and the use of the test system. The answer, we believe, is in solutions such as ZL, which deliver secure risk checks with almost immediate feedback.
Marcus Perrett is chief technology officer at Fixnetix, a DXC company that provides outsourced managed services to investment banks, hedge funds and trading groups.