The First Indian Airline Born on the AWS Cloud

The airline operates with a lean IT team thanks in part to AWS Support, and it now subscribes to the AWS Enterprise Support plan. “Even when we began as a startup, AWS Basic Support was very useful,” Mishra says. “The enterprise support from AWS is highly focused—going above and beyond to help customers to make sure their objectives are met. Being customer-focused has earned AWS our trust.”

Relying on security services—such as Amazon GuardDuty, AWS Web Application Firewall (AWS WAF), and AWS Key Management Service (AWS KMS) in integration with the SOC framework—has allowed the airline to automate threat monitoring. With AWS Direct Connect, Vistara experiences consistent performance in terms of network speed in comparison with its previous virtual private network (VPN) across several offices. Switching to AWS Direct Connect has resulted in 50 percent cost savings compared to the VPN. Vistara also uses Amazon Relational Database Service (Amazon RDS), which Mishra values for its flexibility and scalability.

With AWS, Vistara has experienced 99.99 percent service availability. “Our e-commerce website sees seamless traction of over 5.5 million hits per day with an average of 133 hits per second. Since using AWS services, we have never faced a single instance of downtime,” Mishra says. This includes promotion periods, when the website experiences heavy spikes in traffic. Mishra has established a practice using AWS Auto Scaling that when any workload hits 60 percent utilization, it is put into auto-scale mode.

The scalability of its core platform has given Vistara confidence to expand. Within 2019, the airline will open its first international routes.

In 2016, Vistara began a project to improve route profitability analysis, using Amazon Simple Storage Service (Amazon S3) as a data lake and Amazon Redshift as a data warehouse for structured data. The goal was threefold: determine which routes to start next, which are not performing, and how to make changes to improve profitability. “Route profitability is the most important parameter for any airline,” explains Sachin Madhukar Khaire, principal architect at Vistara. “Providing my management team with data at their fingertips can impact decision making about various cost/revenue data points.”

According to Khaire, cultivation of route results in the Indian airline industry is mostly a manual exercise, with decisions made on “intuition or sheer tribal wisdom.” This is partially due to complex volumetric data sets, averaging 40 GB per route for 100,000 flights, with analysis carried out in bits and pieces. This intuitive practice not only slows the analytics process, but also blurs accuracy. Khaire’s idea was to build a fully automatic big data analytics platform where all key performance indicators would be generated in real time without any manual intervention.

The solution required data to be ingested into a data lake from sources such as Vistara’s own operational data store and SAP. Then, data had to be processed and analyzed using services such as Amazon S3 and Amazon Redshift. The data pipeline—from ingestion to visualization—was fully automated using serverless tools such as AWS Lambda and AWS Glue.

Since launching its data lake on AWS, the time required to generate route profitability reports has decreased from two weeks to less than a day. Every department now uses the data lake to carry out its functions, extracting route-specific data points that are updated daily. This results in higher precision across the organization, from finance to revenue management and network planning. When beginning the project, Khaire was under a mandate to keep the long-term cost of deployment to a minimum, considering that data volumes would continue to grow with a multifold increase in fleet quantities and types, as well as new routes and countries.

The company had been storing all revenue data on a global database server. “Having our data lake in place is key to supporting international operations due to the sheer complexity of route analysis at an international scale,” says Khaire. “We will have a lot more data points to work with in the future, and I can use AWS in other ways such as machine learning, which is profoundly applicable on existing data in AWS. The scalability, feature stack, and centralized information storage are the keys for everything to work.”