Ajai Murali - Are you Being Served?

Article URL (PDF):


Are You Being Served? Article Link

This article by Emily Fraser discusses various types of programs and their advantages and disadvantages in front office, middle office, and back office environments.

Firms are looking for further ways to reduce latency in application timings in a high volume market in which sub-millisecond times play a significant role.

SOA is one design philosophy that is mentioned throughout the article, short for service-oriented architecture. Service-oriented architecture allows firms to reuse software, by allowing applications borrow pieces of code and interact with any other application, to save time and money. It is apparently very efficient in running middle and back office functions, and works well across legal entities and operating units.

A shift needs to take place, according to the author, from service-oriented architecture to something new for front-office applications, since extremely low latency is desired with real time updates and streaming data. A shift needs to take place to higher-performance computing based on market/transaction data.

SOA runs primarily on XML, and though XML provides a dynamic environment for application development, it is not particularly well suited for low latency applications. It may be too loaded with information/brackets/slashes that are not necessary with basic data figures from daily market transactions.

One solution the article mentions is a new algorithm known as event stream processing. HSBC is currently developing a platform in which the back and middle offices are controlled by SOA and the front office is run with ESP. The head of HSBC argues that ESP is really only very useful for the low=latency requirements of the front office, but SOA contains powerful elements that support middle and back offices in a more productive manner.

ESP is an emerging technology that is already at the forefront of becoming a standard application amongst trading firms, and Is becoming more useful for things like data monitoring, compliance, TCA, and risk management. The next big mainstream thing is supposed to be event-driven architecture much like ESP, but it still has a long way to go until smooth implementation possibilities.

There are some advantages to SOA, however. SOA is more useful for applications in which instant response is not absolutely critical. It is also more flexible and able to extend life from legacy systems. Since SOA borrows code from other applications, legacy systems can borrow code for updated functionality. Since mergers and acquisitions are ever-present in the financial world, SOA is easier since it borrows code and makes it easier for firms to consolidate applications.

I believe that SOA is a good move for its functionality, flexibility, diversity. Though event stream processing is a new technology, it must be regarded as that and not implemented until it reaches full maturity. New firms can experiment with it and eventually it will reach old systems. SOA is too common to replace at the moment, and replacing it could sum up to be an immense amount of money. Software algorithms will always build on themselves, and thinking back 10 years, SOA was not even common. So, ESP will eventually become a norm, but SOA is too easy and efficient at the moment.