Advertisment

Spiderman on a stock-market skyline: Could NSE manage it?

author-image
Abhigna
New Update

NEW DELHI, INDIA: THE walls can be too slippery. Having a firm grip on all the sharp edges is very very crucial. Plus, if you are acrophobic, you better stay in the plains.

Advertisment

Did N Muralidaran, Director NSE Infotech Services Ltd. love heights too much? Or was it his hate for cobwebs flanking certain trading spaces in India? Something in between certainly because he surely jumped towards a huge skyscraper when he perched himself on this ambitious innovation endeavour.

For the National Stock Exchange of India, being at par and even ahead of the international markets has been a key objective in its efforts to improve the market mechanism and cater to the growing needs of the Indian market. As direct market access and algorithmic tradingbecame popular, and the market started adapting to new technology ,NSE realized that the number of transactions has increased manifold. In a short span of four years, the number of messages processed by NSE's trading system increased 10 times.

It jumped to an average of more than 500 million messages in a day from 50 million messages.NSE also realized that the low-latency race continued to intensify despite the throughput growing at a fast pace. NSE has been at the forefront on both these parameters because of various strategic initiatives in the past few years.

Advertisment

NSE has always attempted to provide world-class trading services to market participants at the lowest possible transaction cost. With a new generation trading platform Mr. Muralidaran and his team took another giant leap forward. The world had entered a new era of algorithms and automatons, and it was a given that India should be in sync with the international developments.Mr. Murlidaran explains that for any organization to keep pace with change, it is essential to re-invent itself at every stage of evolution. "NSE is no exception to this. NSE had put in a strategy to re-engineer the legacy, mainframe-based proprietary trading system to a more agile, new-generation technology that is capable of running on commodity hardware, along with the non-functional objective of addressing throughput and latency needs of the market," he says.

Besides, the earlier architecture had reached its capacity of vertical scalability as the single-core processors had reached their maximum processing speed. The new era of multi-core processors with higher processing speed called for a redesign of the application architecture to cope with the growing volumes and performance expectations.

Jumping vs. Crawling

Advertisment

The new architecture moved away from a single, monolithic, tightly-coupled architecture to a more layered and loosely-coupled distributed architecture. The functional split of the trading application into multiple layers allowed each layer to scale independently. Each layer could be scaled to any desired capacity by adding hardware. Hence, the architecture allowed for unlimited horizontal scalability. This initiative called for a massive change, and new goalposts had to be conquered. Many fundamental changes had to be made to create the new generation platform.

Some Major Changes:

- A new messaging fabric to communicate within the various layers of the distributed system

- A new operating system from proprietary to open platforms

- A different hardware platform from mainframes to commodity

- A standard programming language

- A new application design concept for low latency

- A custom built in memory database

The team had set sight on a tough and intimidating task that would raise the bar for everyone. But it knew that it had the firepower to bring about this great change. Past achievements reinforced its strength to ascend new heights with ease. In 1994, for instance, NSE had introduced a new electronic trading system in India.

Advertisment

This was supported by India's first exchange trading ecosystem. The core components compriseda trading workstation, centralized limit order book, VSAT-based nation-wide network, real-time risk management systems and a clearing system. Since 1994, NSE's volumes have been growing by leaps and bounds, also helped by regulatory reforms and liberalization of thetrading mechanism.

"If we jump right into what a trading landscape looks like today, it is very different from what it was just five to seven years back," Mr. Muralidaran reminisces. "Today, as part of the ecosystem, a large number of transactions is generated by machines rather than human traders. These machines are neck-to-neck with the exchange's trading system, generating orders at hundreds of thousands of messages per second, and expecting a response equally fast. At the same time, like all organizations, cost effectiveness is equally paramount, to be able to deliver to market participants at the lowest costs , as is the responsibility of running a service for the nation's financial back bone," he says.

DNA transformed

Advertisment

The electronic trading highway was a major achievement. NSE's nationwide network for the market ecosystem was created to connect all trading entities across the country.This was all about designing and building a reliable, secure, scalable, high-bandwidth, wide-reach and cost-effective network in BOOR model, i.e. Build, Own, Operate and Refresh, which was a ‘first of its kind' in India. The closed user group network was delivered through 13 POPs (Point of Presence) strategically located across the country.

It enabled a communication path to the primary and disaster-recovery data centers of the exchange. It also supported scaling up to add more connections as well as POPs to maintain cost effectiveness. On the other hand, the network supported latest communication protocols in line with cost-effective skillsetsavailable with the telecom and application builders.

This translated into management of increased bandwidth and security risks, but these concerns were addressed by innovative flow control protocol implemented through Trading Access Point (TAP).

This allowed the members to consolidate their IT infrastructure, resulting in huge cost savings. Saving costs became one of the most important achievements of this team. "In the earlier setup of X.25, because of the bandwidth restrictions of 64 kbps, members had to handle high throughput by deploying hundreds of 64 kbps lines. They also needed Eicons to connect these lines to the servers, and had one server for each of these lines. The new system could handle this traffic with just a couple of lines under a single CTCL server that was connected to multiple communication lines using TAP. These changes resulted in cost saving in IT infrastructure, and human resource."

Advertisment

Four Key Functional Architectural Layers of the New-Generation Trading System:

1. Communication management - For user authentication, session management and connections management with end user terminals

2. Master management - An orchestration layer for product flexibility, multiple sessions and trading flexibility besides being the gold copy of market member session and state information

3. Matching - The core of trading engine for matching of the order messages and disseminating the same to post matching layer

4. Post-Matching - For processing and dissemination of market data to market participants, risk management systems, clearing & settlement systems

Saving more damsels

Advertisment

Apart from the implementation of the new platform,there were innovations in other areas too. For example, the new pricing model for end users and service providers was introduced as a part of this project. For end users, the flexible upgrade or reduction options were introduced to facilitate easy scale-up and scale-down, depending on the business environment. The service provider pricing model was based on complete outsourcing of Network-as-as-Service in a BOOR model, with an option to transfer to another service provider for sustained quality of service. The new platform managed its milestones well and increased the installed capacity by about seven times and reduced response time by 40 times.

Co-location was another important aspect of the new-age market ecosystem. These machines were under the control of the trading members, running strategy-based algorithms and responding to millions of tick by tick market data. NSE started co-location facility at its BKC premises ahead of many of its international peers. Over the last couple of years, it has reached a state of maturity after integration of other new-age layers.

Rough Walls, Tough Grips

The big change came with its fair share of challenges. But Mr. Muralidaran had anticipated them well before he started the initiative. Managing high throughput, low latency, ensuring reliability and recoverability, being confident of causing minimal disruption to the market participants and putting in place a smooth replacement were some of these challenges. He and his team addressed these challenges with a combination of meticulous planning, flawless execution and impeccable team work involving both internal and external teams.

"This was a high impact change, as we were replacing a running high-velocity market step-by-step, compared to the ground-up implementation done in 1994. This was equivalent to replacing the complete market infrastructure," he says. Not only that, there was the need for co-existence with the old system involved and involved changing systems with a zero interference/disruption to market participants. As Mr. Muralidaran rightly puts it - it could be considered equal to changing the engine of an aircraft while it is flying. "The mandate for the project was to transparently change the various pieces in the trading system, which means that the new system had to ensure that all the functional pieces of the legacy system was captured and engineered correctly. All the necessary interfaces and touch points required to support the existing and new system had to be provisioned for."

Despite all the challenges, it was a learning experience for the team and they completed the mammoth task successfully. "The new trading platform positioned NSE as the No. 1 Exchange in the world in terms of the number of transactions," Mr. Muralidaran says. The trading platform was launched in the equity and currency derivatives segments. NSE witnessed a sharp increase in the order processing rate and currently processes close to half a billion messages every trading day, he adds.In spite of the jump in the number of messages, the investment in infrastructure has not gone up. In fact, it has reduced as the new platform is deployed on commodity class servers instead of proprietary mainframes.

There has been 20 times reduction in equipment cost. The new platform has also earned recognition for providing facilities which are ‘first of its kind'. "In India with bandwidth available as commodity, NSE in its N3 platform has created architecture to add bandwidth on demand and to handle ever increasing transaction volumes from the market," he says.

The Touchdown

All these strategic initiatives undertaken by NSE were ahead of its time. "It helped the team to continue our innovative leadership and introducing new products and services. It set a trend for NSE as an industry leader and others had to follow suit. The project enabled knowledge development in areas such as project management, reverse engineering, solution architecture, etc," Muralidaran says.

The project is a noteworthy attempt, not only for replacing a reliable, robust system that has been running perfectly well for more than 20 years, but also because it was done with in-house expertise. NSE's technology team collaborated with the leading IT vendors of India to achieve this remarkable feat.One of the key objectives of this project was to bring down the latency, and that has happened -by almost 20 times. It has made NSE's platform equal to some of the leading global platforms. Also, NSE's order processing rate increased 10 times, catering to each type of trading member: from small investors and member brokers to the most sophisticated high -frequency traders.

More milestones of the new ecosystem:

- 10 times increase in order processing rate wrt trading volumes

- 20 times reduction in equipment cost

- 20 times improvement in latency

- New revenue stream with introduction of co-location facility

To conclude, it may look easy as a hop but when you look inside and exert some moments, you discover the length and height that this stride was all about. Superheroic or not, it was quite a dent into old walls.

So next time when a bug bites, remember that some problems lead to undiscovered powers in disguise.

case-studies