[an error occurred while processing this directive]
Issue dated -11th October 2004


Previous Issues



Symantec Report
Security Headquarters
Openings At Jobstreet.com

 Network Sites
  IT People
  Network Magazine
  Business Traveller
  Exp. Hotelier & Caterer
  Exp. Travel & Tourism
  Exp. Pharma Pulse
  Exp. Healthcare Mgmt.
  Express Textile
 Group Sites
  Indian Express
  Financial Express

Front Page > Case Studies > Story Print this Page|  Email this page

Linux stars in Mission Critical

Linux is growing out of Unix's shadow and handling
mission-critical tasks on its own. Check out its role at the NSE

THE National Stock Exchange (NSE) was one of the first exchanges in the country to introduce instruments such as derivatives (which are similar to equity) where you trade in futures and options.
NSE had a Unix-based system in place for equity trading, but felt the need for a more cost-effective risk management solution for its derivatives segment. The team working on the project had to consider various issues. Real-time throughput was required to analyse trade risk factors. The team also had to ensure that the system wouldn't collapse under high loads. Linear scalability was another key issue.
Says G M Shenoy, senior vice president of NSE.IT, "NSE was looking at large amounts of computation. There were more than 1,000 trading members and over 200 clearing members. Initially we were looking at around 50 trades per second, which would mean 100 value at risk (VAR-a term for risk analysis that has been defined and approved by economists) computations/second. Parallelism was the only answer."

A Linux cluster

After considering various options, it was decided to go for a Linux cluster computing solution, initially on an experimental basis.
The development team decided to use a cluster of inexpensive Linux-based machines powered by Intel processors, and use MPIS (message passing interface standard) for clustering infrastructure, with a switched Ethernet 100 Mbps network. It was decided to use multiprocessor hardware to handle non-parallel components such as data sharing.
During this experimental period, R&D was conducted on the platform to ascertain the robustness of the system and to demonstrate the use of Linux for a real business application requiring high-level floating point computations at high performance levels.
The concept worked. From there on it was a matter of fine-tuning and putting it into production.

The answer is Prism

Initially NSE had deployed SuSe Linux, but later on moved to RedHat Linux 7.2 as it was more ubiquitous. Explains Shenoy, "Plenty of information was available on the Internet with reference to Red Hat Linux, which was not the case with other Linux packages. Also, it was easier to get drivers for Red Hat Linux. Another factor that went in favour of Red Hat was that it was easier to get support."
The system NSE set up for doing VAR calculations has been christened Prism (parallel risk management system). Prism consists of a centralised mother machine that is in turn connected to a number of child machines.
The mother, a dual-CPU Intel Xeon 1 GHz server, which runs Red Hat Linux 7.2, receives the trade information. The trading positions and margins of all the members are maintained in the memory of the mother machine, which is fault-tolerant. All the computations that take place on the mother machine are copied onto another machine. This ensures data recovery in case of a breakdown or disaster.
Information is distributed by the mother machine to the child machines. The child machines are single-CPU Intel PIII 800 MHz workstations. These machines do the margin computation, which is then sent back to the mother. Currently there are five child machines, but NSE plans to scale up once trade volumes go up. The Linux children need no management; only the IP address has to be configured to install a replacement child.
Two important aspects of the derivatives segment are clearing and settlements, and risk management. The risk management system at NSE enables online risk and position monitoring of members. The system, developed by NSE.IT, sends out alerts to the trading system and risk management team, and if a broker goes past his acceptable risk limits, his account is disabled in real-time. For instance, if one trade crosses the limit, the next trade a second later will be automatically rejected. Thus, NSE's risk management system plays a critical role by nipping the possibility of a default in
the bud.

Cost-effective and robust

Prism cost NSE around Rs 15-20 lakh. The savings have been substantial when compared to the cost of setting up a Unix-based system, where the cost of acquiring the mother machine itself would have been more than Rs 20 lakh.
The current system can be scaled up to 1,000 trades/second. The software development has already been done, and the hardware costs are trivial. According to Shenoy, it has been possible to keep the costs so low only because of the Linux-based systems.
Very few exchanges in the world are said to have a robust and cost-effective system like the one developed by NSE. Indeed, Prism is a system that can be replicated by other exchanges in the country, even in the capital markets, and also to drive up volumes in the derivatives segment.
An interesting sidelight in this effort is the fact that Prism was developed by NSE.IT, a spin-off company that consists of the in-house IT team that had worked on NSE's core systems. Thanks to this proactive initiative, NSE is today counted among the top five exchanges in the world in terms of volumes traded.

<Back to top>

Copyright 2003: Indian Express Group (Mumbai, India). All rights reserved throughout the world. This entire site is compiled in
Mumbai by The Business Publications Division of the Indian Express Group of Newspapers.
Please contact our Webmaster for any queries on this site.