Case Study: 300% Increase in App Performance with India Rail on Pivotal GemFire

November 17, 2014 Neela Chaudhari

featured-india-railAccording to McKinsey, India is moving from 120 million online users in 2013 to 330 million users in 2015. With the growth, India’s e-ticketing demand for rail travel has increased exponentially—from just a few hundred tickets to 500 thousand tickets per day.

CRIS, India Railway’s Centre for Railway Information Systems, worked with Pivotal to scale their system and recently presented at our EMC Forum in Delhi, India. Suneeti Goel, Chief Project Engineer, talked about the new e-ticketing system capabilities.

According to Goel, design limitations prevented the old system from scaling, and, in 2013, the Ministry of Railways decided to re-architect and implement a new system. There were several key requirements from a growth, scale, and business agility perspective. Beyond the increase in internet users, a key goal of the e-ticketing system was to reduce reliance on offline ticket counters and transactions, and this added to the volume of transaction support needed. Two, national holidays like Diwali create massive spikes in use or cause unexpected changes to peak load times, and the system needed to be able to accommodate these spikes flexibly. Three, the system needed to operate with business agility as expected from a modern web and mobile-based application, expanding on a true, scale-out data store.

Target Performance, Scalability Factors, and Data Grid Architectures

Growing over time, the environment initially needed to support the types of traffic seen by very large, consumer internet websites and e-commerce stores with:

  • 120,000+ concurrent sessions during peak time
  • 30+ million registered users
  • 700 thousand bookings per day
  • 30 million queries per day
  • 600+ million Rupees of revenue per day

The CRIS architecture team had already determined that simply adding new hardware would not solve performance issues. They designed a completely new application and incorporated a data grid architecture that would manage huge concurrent workloads, support millions of users, and provide dynamic load balancing to seamlessly manage demand at peak hours.

After evaluating a variety of options, CRIS IT leaders chose Pivotal GemFire 8, part of Pivotal Big Data Suite. GemFire’s distributed, in-memory, shared-nothing architecture manages CRIS application data across many nodes—load balancing, high availability, and scale-out capabilities are built in. Response to application workloads are extremely fast because data is not only served from memory, additional throughput can be achieved by adding more nodes. In addition, the data store can be stand-alone or, as in this case, sit as a front-end cache, reading and writing to back-end systems independently of high-volume application queries.

The Results with Pivotal GemFire

With Pivotal GemFire in place, the following performance improvements were attained:

  • 120,000 concurrent users can book e-tickets simultaneously—a 300% improvement
  • Bookings increased by 500% in the first 10 minutes of running GemFire underneath Tatkal, a short-notice booking application.
  • E-ticket sales have grown significantly overall, especially for Tatkal
  • Response time has seen a major improvement, entering the sub-second response time
  • Wide variations in user access patterns are handled without concern for SLAs or downtime
  • Per minute booking capacity increased by 5 times to 10,000 transactions per minute

In summary of what was accomplished, Goel shared, “Our next-generation e-ticketing system faces the same scale issues as the largest e-commerce sites in the world, and Pivotal GemFire allows us to support this type of highly concurrent environment, maintain high availability, and increase capacity over time. In our evaluation, we saw nothing that could give us this type of performance except GemFire.”

With a country-wide responsibility for making e-ticketing work for its citizens, India Rail relies on Pivotal GemFire to help them address the future as internet users increase, new applications connect, and functionality expands to support consumers.

Learn More


About the Author


Pair Design Rule #5: So Humility, Such Enlightened, Much Wrong
Pair Design Rule #5: So Humility, Such Enlightened, Much Wrong

Co-authored by Pam Dineva and Kim Dowd. Part of pairing is oversharing. To characterize, we externalize. An...

Cloud Foundry Development Hardware Requirements
Cloud Foundry Development Hardware Requirements

Abstract How much RAM, Disk, and CPU are needed for a Cloud Foundry deployment? We believe you need this mu...

SpringOne 2021

Register Now