success story

Migration of Real Time Bidding Platform with 10+ Millions Requests Per Minute in 10 Weeks

The client is a leading American mobile technology company offering targeted placement and tracking of advertisements on computers, mobile phones, and end-point systems such as cash registers. Their products collect data on how their clients' ads perform in terms of consumer exposure. The need to be more nimble and agile to their client's needs led them on a journey to digitally transform systems and processes. The client embarked on a project to lay down the cloud foundation on Google Cloud and migrate its real-time bidding platform to the Google Cloud from Rackspace.

 

 

The Challenge

The client operates in a very time-sensitive, competitive space. They have mapped all 150 million households of United States. Their product matches users with relevant advertisers; participates in online auctions for space, and, if the bid is successful, places the advertisement. This entire process should be completed within 100 milliseconds; otherwise, it could result in a loss in revenue. In the existing Rackspace environment, provisioning of computing resources would stretch from weeks to months. Traffic fluctuates throughout the day, high (up to 15 million requests per minute) in the morning and evening, and traffic drops by 70% in the night. The application required a considerable amount of manual intervention by administrators to stabilize the system during unpredictable traffic fluctuations.

The client wanted to move Microservices running in Docker Swarm in Rackspace to the Google Kubernetes Engine (GKE) in Google Cloud with minimal effort and achieve a flexible architecture and scalability for the future business growth.

The migration project would have to automate the deployment process for infrastructure and applications and minimize any additional effort and expense for incremental application migration.

The Solution

Virtusa, a long term Google partner, built the enterprise foundation and migrated the client's real-time bidding platform to Google Cloud. The Google Cloud foundation was built using Google Cloud stack of Cloud IAM, Cloud SQL - MySQL, VPN/VPC/Networking, Cloud Storage, Stackdriver Logging/Monitoring, Cloud NAT, Cloud DNS, Google Compute Engine, Google Kubernetes Engine, Google Cloud Pub/Sub, GLB, and NLB.

  • We wholly automated the provisioning, configuration management, and de-provisioning of Infrastructure Services to enable DevSecOps team to build and tear down environments on-demand within minutes/hours, as against the traditional ITOps environments requiring several weeks/months to setup.
  • We defined and built continuous deployment pipelines for real-time bidding platform.
  • We also managed Instance Group's Auto-scaling and Auto-healing features to drastically reduce manual intervention by administrators during traffic fluctuations.
The Solution
The Benefit
  • Simplified management of real-time bidding platform requiring less than 100 millisecond for overall response time
  • Improved reliability and stability of the platform using auto-scaling and auto-healing, making the platform more responsive to real-time traffic fluctuations
  • Reduced downtime and risk by using an automated Blue-Green deployment process
Related content