Understanding the Limits of Complex Event Processing

CEP solutions are limited in their out-of-the-box functionality.

100x102stock-ticker.gif
Harrell Smith

Harrell Smith

A.E. Fletcher Photography

In their ongoing quest for competitive advantage, many firms that trade on exchanges are looking to implement high-performance algorithmic trading solutions. As a result, numerous complex event processing (CEP) vendors are marketing themselves as high-frequency, low-latency analytical engines that can form the backbone of a firm’s trading infrastructure.

These products, however, are severely limited in terms of their out-of-the-box functionality. Since CEP vendors target multiple verticals (e.g., defense, telecom, commercial banking, retail, etc.), they do not provide extensive support for the type of core components that make up an algorithmic trading solution.

Numerous technical issues must be addressed to deploy a truly complete algorithmic trading platform. Among these are:

Market Data Handlers and Load Balancing Technology. Market data is obviously a key component of any algorithmic trading solution. Firms looking to deploy a high-frequency solution have to ensure that they have rock-solid feed handlers for their data provider(s) of choice. In addition, many firms employ multiple feed handlers, either for asset-specific strategies or to use “first in the door” real-time data scrubbing to ensure their algorithmic strategies have the most accurate view of the market.

Sponsored

Complete State Management. For an algorithmic trading engine to operate properly, it must be able to maintain and act on a complete, real-time view of a trade’s life cycle. Getting answers to key questions up front can help determine how an order should be handled. Is it a single ticket entry? Is it a “child order” — a small part of a larger order that might itself be part of a larger basket-level strategy? Has the order been accepted? If so, is it subject to cancel/replace? Was there already a partial fill that would affect one’s position? All of these issues influence how the entire trading strategy should be managed. Collectively, this kind of technology is known as state management. Without it, an algorithmic engine would be flying blind.

Certified Connections to Trading Destinations. In today’s marketplace, literally hundreds of electronic execution venues are available. Broker algorithmic suites, exchanges, electronic communication networks, automated trading systems and crossing networks have created a complex, highly fragmented trading environment. Building and certifying connections to global market centers is an enormous undertaking, requiring ongoing development to comply with emerging standards and changes to liquidity providers’ messaging logic.

Front-End Trading Interface. Traders who use stand-alone CEP engines and wish to develop, launch and monitor algorithmic trading strategies in a graphical environment must develop a custom solution or integrate with a third-party front end.

Does this mean that event-processing technology has no application in automated trading? Of course not, it is an integral part of any high-performance trading system. But firms considering using stand-alone CEP engines for algorithmic trading should be aware that these products are just a small part of the overall solution. If they wish to leverage a high-performance trading solution, they’ll need to consider the enormous cost and time associated with this kind of piecemeal development.

Harrell Smith is the head of product strategy at Portware, a developer of broker-neutral, automated trading software for global equities, futures, options and FX.

Related