Migrating Microsoft Workloads to AWS Cloud for a Government Body – Entity Under Ministry of Railways

Challenge

Creating a time table for trains on a busy network like the Indian Railways is an extremely challenging task. Planners on the zonal railways work independently and then collaboratively with other zonal level planners to design the All India time table. The software application manages the train scheduling and network governance, this project is entrusted to enable effective trains scheduling without any hassles which require a lot of agility on the infrastructure from computing capacity stand. Applications needed infrastructure agility their application primarily targeting up-time and ability to save cost. Following challenges were posted:

1. Infrastructure should be extremely secure.

2. Running cost of the server and maintenance should be very low.

3. Local team should be able to understand and adapt to the change.

4. Infrastructure cost should not exceed the overall budget allocated.

The application is core Microsoft technologies based like Microsoft Visual Studio C# for development, Microsoft Expression Blend for designing Interfaces, WPF framework for Rich Graphical Interface and MVVM for design patterns. Oracle 11G provides the backbone for the centralized database. This application is designed based on a three-layered architecture. Terminal servers are used to host application centrally.

Solution Overview

Solution Architects at AWS and Minfy devised a strategy to re-engineer the application so that it would be smart enough to understand system idle time , that would automatically stop the system once not in use. This was only for the development and test systems. The production systems were carefully sized to meet the expectations of the client. The unique cost-effective solution was proposed leveraging AWS –Lambda- to set the policies based on instance state and user slogged in. The custom application was developed with a capability to learn whether an user is in active state or not, An activity monitoring agent was installed on every production application and database instance. This agent posted an activity timestamp into DynamoDB which acted as a “Registry” and was read by a self-triggering Lambda function at regular intervals. If the last timestamp in the registry was 15 minutes earlier than the current timestamp the Lambda would trigger a stop command. In case there was no user active the all the production instances like App Server, Database, Active Directory and terminal servers would be sent to stop state. The servers would start automatically as soon as the web link to access the portal was accessed. This became a highlight of the implementation and the customer accolades it as well. It helped to bring down the running cost of the development & test servers to effectively a few hours a week.

Benefits

- Cost of running the infrastructure was bought down to lower than the approved budget.

- Security was established by implementing the IPsec tunnel between on-premise and Active Directory.

- Effective use of ALB and VPC ensured that the servers were not exposed to the internet. Automation reduced the operative maintenance cost to almost zero.

- Faster GTM and deployment.

- Scalability - To upscale basis load tests and seasonal demand.

- A complete Opex model vis-à-vis Capex investment they had to put in for on-premises. Flexibility to check and finalize the specs for more stable environments.