Jan 1, 2018

Accelerating the Development of Real-Time Trading and Risk Systems: The Dependency Graph Advantage

The OTC derivatives markets are moving faster and faster—increasingly demanding real-time responsiveness. Notably, as we see electronic trading become more pervasive in these markets, the quicker a firm can react to market changes, as well as calculate price and risk information, the greater advantage it has. Staying competitive in OTC markets demands timely decision making, and that requires real-time intelligence. But building trading and risk management frameworks capable of doing this can be overwhelming, especially when considering the time, budget, and development expertise needed to make it happen.

Numerix Chief Marketing Officer James Jockle asked Bill Dwyer, Vice President of Business Development, Americas at Numerix, to share his viewpoints on the specific challenges technology leadership and development teams in financial institutions face as they rapidly scramble to build the modern OTC trading and risk platforms that can meet current and future market demands. A veteran of the technology and financial services industries, Bill has deep capital markets experience and has helped some of the world’s largest sell-side and buy-side institutions overcome their most complex business challenges.

In this interview, we will address what’s driving this shift to real-time trading and risk management platforms, the issues faced, and the type of technology needed to quickly deliver solutions that will provide institutions developmental value and a competitive edge.

James Jockle: What are the core industry challenges facing capital markets institutions?

Bill Dwyer: I think we can safely say that for capital markets participants business is harder today than ever. The structural changes stemming from post-crisis regulation have placed market participants under increasing pressure. We are witnessing in the OTC derivatives markets, an area where Numerix has traditionally been active, that the speed of business is accelerating as trading is moving to electronic platforms and the business is evolving to be more flow oriented.

The market has changed from being a voice driven, highly bespoke bilateral market to a higher velocity, electronically traded market with lower margins. So, staying competitive in this kind of changing market requires participants to be able to do things efficiently in an automated way; to price trades in real time and to calculate risk on a continuous event-driven basis throughout the trading day.

Jockle: So, to put it plainly, decision makers need more and faster access to the right information. What challenges does that present for CTOs and development teams in charge of adapting and building the technology systems needed to meet real-time requirements?

Dwyer: I think that having a real-time infrastructure for pricing and risk is going to be table stakes for OTC market making going forward. The problem is that the real-time calculation of pricing, risk, margin and the like present significant technology challenges for most organizations. For one, the commercial OTC trading and risk systems available on the market today are not adaptable to real-time, event-driven operations. Most of these designs date back to the mid-90s and were built to solve the business problems of a different era. They cannot be easily re-engineered or retrofitted for the “new normal.”

We have seen people try to attack these issues using more general-purpose middleware or CEP tools, but this software was not designed specifically for use in the capital markets, so their value is more limited. Typically, these tools solve only part of the real-time pricing and risk problem, and often come with unforeseen limitations that limit their practical usefulness.

Some have tried building their real-time calculation engines in-house from scratch using open source tools and the like. I think this has proven to be a viable approach at very large Tier 1 institutions, essentially the eight to ten largest investment banks in the world. But in my view, these banks are the only ones who can have the scale to afford the multi-millions of dollars in expense. These implementations typically require 3-5 years and demand large development teams with very hard-to-find expertise. So, the internal build is likely feasible for a very large player but probably a bad idea for the “rest of the world.”

Jockle: If traditional trading and risk systems, messaging middleware stacks, and CEP Engines can’t rise to the challenge, are there technologies out there that can?

Dwyer: From what I have seen, the most common approaches are to use an off-the-shelf trading system and live with its limitations, build something using a middleware product and live with other kinds of limitations, or build from scratch and spend enormous amounts of time and money.

Knowing these limitations, our experts set out to build a very compelling alternative. It’s essentially a development toolkit, a set of building blocks for constructing real-time event-driven pricing and risk engines. The underlying infrastructure of the framework looks a lot like middleware in that it supports things such as high-performance communication, fault tolerance, load balancing, and data source consistency.

However, it also contains a dependency graph layer which acts as a sort of “brain” that dynamically orchestrates all the various nodes and the complex calculations they perform. Hence, the name Oneview Graph Framework. The result is that calculations are performed intelligently, on an event-driven basis, and in a way that is highly efficient and highly scalable.

The thing that is differentiating, at least for me, is that our graph framework was purposely built to solve the problem of real-time calculation for capital markets participants. It is proven in real live production use cases such as real-time risk based pricing of RFQs, real-time PnL calculation, real-time greeks and margining.

Technology of this kind is not unusual for large market makers, and, in fact, our software was built by a team who had direct experience doing this for a Tier 1. To be frank, I don’t know of any other software vendor who is offering a commercial product or anything remotely like this.

Jockle: What makes Oneview Graph a solution for the challenges the industry faces now?

Dwyer: As I mentioned, the actual dependency graph layer of Oneview Graph sits on top of the event-driven communication and functions as a sort of “distributed brain” that orchestrates the interaction between the various calculation nodes on the graph. It intelligently reacts to events and allows each node to perform its specific task, but only when necessary. That’s key to its effectiveness and efficiency. For example, if we receive an interest rate tick on a market data feed, the graph will know which curves need to be bootstrapped, know what trades need to have their NPVs and greeks recalculated—all without any human intervention.

This is just a very simple example; in practice dependencies become very complex to manage. A large market making operation will typically need to perform millions of calculations throughout the trading day to support processes like auto quoting, risk or margining. If you don’t have something like the Graph orchestrating this process, there is no way to do it efficiently, in real-time, and in an event-driven way.

Jockle: What benefits will development teams walk away with integrating or leveraging a graph component as part of their trading architecture?

Dwyer: It sounds a bit like a software cliché, but a good development framework acts as an “ROI accelerator” by providing a set of building blocks that allow developers to focus on the business logic and business processes that are actually differentiating and unique to their business. So, there is an incredible gain in terms of productivity, time to market, and reduction of project risk.

In terms of integration, Oneview Graph has a light footprint and can thus integrate into virtually any existing architecture. For example, it does not impose a data model, nor does it come with a UI but has all the hooks to integrate with the existing architecture. A banks’ IT team can choose to plug in their own pricing analytics or business servers, or can optionally use off-the-shelf components from providers like Numerix.

The result, we think, is a solution for the real-time pricing and risk engine problem that is easy to integrate into virtually any existing architecture. We think this provides benefit in terms of potentially extending the life of existing trading systems that lack this capability. Instead of throwing out the baby with the bath water by doing a rip-and-replace of an existing system, banks can add a real-time risk and pricing service that leverages existing models and existing trade processing systems.

Jockle: What benefits will an institution realize with a graph framework?

Dwyer: Oneview Graph has been used to solve a number of different business problems including real-time, high volume risk-based automated quoting of RFQs at a Tier 1 Bank, real-time risk across trading desks at hedge funds and banks, as well as real-time margin calculation and limit management.

For organizations needing to solve these kinds of problems, the challenge can be very daunting. If you are a Tier 1, you could choose to throw a large team of engineers at the problem for three years, but most organizations don’t have this luxury.

To summarize the benefit, a graph framework can solve a very challenging set of real-time computation problems for market participants, while de-risking and accelerating the actual implementation. The business benefits that flow from this can depend on the problem solved. For one market maker leveraging our graph framework as a quoting engine, it has meant an increase in market share. Changes like that can lead to literally multiple hundreds of millions of dollars in revenue growth over a multi-year period.

Bill Dwyer is Vice President of Corporate and Business Development at Numerix. He has over 20 years of experience in the financial services technology sector and has helped some of the world’s largest sell-side and buy-side institutions overcome their most complex business and regulatory challenges. Currently based in New York, Bill’s past experience has included stints with SunGard and Reuters in Paris, and Teknekron Software Systems (now Tibco) in Luxembourg and Washington, D.C.

Need Assistance?

Want More From Numerix? Subscribe to our mailing list to stay current on what we're doing and thinking

Want More from Numerix?

Subscribe to our mailing list to stay current on what we're doing and thinking at Numerix

Subscribe Today!