GTU Windows Workload

Challenges:

  • GTU developed more than 100+ applications, all of them are on AWS
  • As such the overall students strength is 5,00,000+ and many applications require truly dynamic workload like Results, Submission, Exam Papers etc.
  • Below are the challenges
  1. Shared, stable and elastic storage availability
  2. CodeDeploy to update .Net code in windows EC2
  3. Stateless applications
  4. Active Directory for all AutoScale windows EC2
  5. Windows EC2 automatic registration with Active Directory.
  • The IT department needs to scale vertically based on their previous experience and best predictions. This many times turns into over provisioning.
  • Manual Scaling is always challenging and requires skill resources availability and some downtime.

Solution:

  • Horizontal scaling with Auto Scaling so no special skill or resources are required.
  • FSx is introduced to provide common storage for all Auto Scale instances.
  • Active Directory auto registration powershell script created.
  • New images are maintained for all code updates.

Architecture Diagram:

Outcomes

  • Auto scaling provides excellent response to dynamic workload on demand with horizontal scaling of windows EC2, so manual scaling is not required.
  • Due to Auto Scaling of windows instances, over provisioning does not occur and lots of cost can be saved with better user experience at peak load.
  • With the use of AWS FSx services common storage problems were resolved and the initial pilot run result was excellent. It is working as per expectation in Production also.
  • Special designed PowerShell script with Windows EC2 user data served the purpose and helped us achieve auto registration to Active Directory for new windows instances launched.
  • All AWS and windows security best practices adopted so the solution is stable and secured.

Conclusion

  • GTU has achieved autoscaling as per their requirement, to match their daily dynamic workloads. 
  • They are also replicating this architecture for their other existing windows workload in AWS.