Is Your Firm Ready for Hyper-Contextual Trading?
By Andrew F. Bach, Chief Architect for the Financial Services Team, and Juniper Networks
Fast is good, but smart is better. As profits from high-frequency trading have eroded, capital market firms are shifting their focus to high-intelligence trading.
Microseconds, milliseconds and nanoseconds will always matter, but a significant advantage can be gleaned from making better trading decisions based on the exploding volumes of information publicly available. Financial firms are increasingly scouring a broad variety of sources, including market data, industry news, company press releases, as well as the tsunami of social media, to assess information credibility in an instant and swiftly move this intelligence into the decision stream.
Hyper-contextual trading, or the ability to combine big-data analytics with increasingly powerful computing, is fast emerging as the next market advantage. It’s an emerging opportunity, and to succeed, capital market firms need a data center infrastructure that can deliver at levels that have not yet been achieved. Hyper-contextual trading is predicated on an ultra-high-performance, ultra-low-latency data center infrastructure.
Financial firms need the ability to analyze and process massive quantities of data at a breakneck pace literally billions of messages every second. The race to zero network latency has gotten the industry partially there, but as we reach the physical limits of how fast packets can be processed, a different approach is required to cross the finish line first.
One emerging approach is applications-embedded networking. In this model, switches are enhanced with additional compute, storage, and processing power, allowing application functions to be run directly on the switch. Applications-embedded networking can be used to distribute computing across the data center fabric to achieve both the ultra-high performance and predictability that’s necessary in today’s capital markets. With innovative, ultra-high-performance switches coming to market, firms can significantly improve performance for market data feeds, order execution routing, big data analytics, and many other highly demanding applications.
Application-embedded networking also promises to reduce IT costs significantly. Firms can eliminate countless servers and switches, which reduces the size and complexity of the trading infrastructure. And there’s less demand for power and cooling. With ultra-high-performance, applications-embedded switches, the trading infrastructure becomes smaller, faster, and easier to manage.
But that’s just the start. Other innovations are needed to fully harness the power. First, the ability to run network functions in a virtual machine on the switch itself increases application performance and minimizes interference from the network operating system itself.
The second innovation is the integration of a field-programmable gate array (FPGA) into the switch. An FPGA is essentially a chip that can be programmed by the customer after the product is shipped. Traditionally, highly skilled hardware engineers were needed to program FPGAs not exactly a core competency at capital markets firms. But with recent advances, developers now can use Java or other popular programming languages to port their applications to run directly on these powerful switches. This allows firms to reduce their development cycles as well as use and reuse code for other applications, accelerating the time-to-value.
This allows firms to take advantage of compute-integrated, high-performance data center switching to accelerate data flow processing and, at the same time, deliver the highly deterministic performance to ensure fairness in a highly regulated industry. Applications-embedded switches do not replace servers and programmable NICs in the data center, but rather this approach optimizes the computing resources that are distributed across the data center infrastructure to achieve maximum performance.
Additionally, these applications-embedded switches can be integrated easily into current data center architectures. With powerful switches built on industry standards, the business’s future requirements won’t be limited by proprietary lock-in.
Three Ways to Change the Game
The possibilities of applications-embedded networking are transformative across capital markets, including equities and commodities exchanges, market data providers, high-frequency trading, risk management, insurance, and credit card processors.
Ensuring the predictability and consistency of order routing and execution especially as the industry shifts to hyper-contextual trading is critical to attracting and retaining customers. Firms can achieve a significant performance advantage by integrating these core applications directly into the data switching fabric, allowing a host of contextual information to be weighed with every trade. With applications-embedded networking, firms will have the speed and power to process billions of messages every second.
Applications-embedded networking is also ideal for the aggregation of compression of market data feeds. Financial information providers and exchanges handle hundreds of data feeds, with the size of feeds growing by leaps and bounds every year.
Risk analytics is another area where firms can use applications-embedded networking to gain an edge. The demands on risk analytics have grown exponentially, driven by new products as well as by an onslaught of regulatory requirements. Firms may need to consider tens of thousands of factors in an instant. With so much at stake, even small improvements in risk management models can have a big payoff. Performing these complex calculations over massive data sets either in real-time or on-demand requires an ultra-high-performance data center infrastructure. And applications-embedded networking can deliver.
The race to zero latency is coming to an end, as high-frequency trading ebbs and ultra-low-latencies derived from silicon advancements level off. To gain an edge, firms must make trading decisions based on split-second analysis of mountains of data. And that requires a fresh approach a compute integrated, high-performance data center infrastructure that has both the brains and the brawn for today’s unforgiving market.