IU's Hybrid Cloud Model

James Bennett
Enterprise Platforms, Indiana University
https://go.iu.edu/4QGf

Enterprise Platforms

  • Part of central IT at IU: UITS
  • Combination of several former groups:
    • Systems Integration/Middleware
    • Linux Administration
    • Database Administration
    • Cloud Support

Agenda

  • Historical problems and solutions
  • Current strategy
  • Migration towards the cloud

Historical problems and solutions

Problem 1: Connecting back to resources on-prem

  • Need access to protected resource
  • Networking group maintains Direct Connect / ExpressRoute / etc
  • Networking expertise brought in for configuration
  • Billing shared among groups utilizing this service

Problem 2: Storage

  • So many storage solutions
  • On-prem options too expensive
  • Moving administrative storage to SaaS solutions like Microsoft Teams
  • Assistance with understanding and configuring cloud storage

Problem 3: Understanding cost/funding model

  • Well-defined, fixed costs on-prem
  • Transition to OpEx model is challenging to navigate
  • Varying challenges with resellers
  • Bringing visibility with Kion

Problem 4: Availability of on-prem resources

  • Need access to assets like our IAM stack

Current strategy

Containerization

  • Utilizing containers in production since 2015
  • Robust deployment of Kubernetes utilizing
    VMware Tanzu
  • Portability and strong operating model

Types of workloads

  • Java, .NET Core, Python, Node,
    Ruby on Rails, PHP, etc.
  • Persistent workloads like databases,
    caches, queues, etc.
  • Supporting services for ingress,
    secrets, certificates, etc.

Rethinking development methodology

  • Moving towards smaller, stateless applications
  • Well defined relationships between infrastructure components
  • Templates lower barrier of entry

Migration towards the cloud

Exercising container portability

  • Workloads can be deployed to multiple clusters
  • Need to migrate to new Kubernetes distribution
  • Lessen or remove impact on development teams

Migration process

  • Heavy use of Helm
  • Argo Events
  • Argo Workflows
  • Kubernetes API
  • F5 Load Balancer AS3 API
  • Parallelization
  • Synchronization
                    {
  "release": "myapp"
  "source": {
    "cluster": "old-es-test.kube.iu.edu",
    "namespace": "ua-vpit/enterprise-systems/ep"
  },
  "target": {
    "cluster": "new-es-test.kube.iu.edu",
    "namespace": "ua-vpit/enterprise-systems/ep"
  }
}

Results

  • Migrated 163 Redis clusters in test
    (~2,000 containers)
  • Zero downtime or monitoring alerts
  • Future flexibility in maintaining our
    workloads
  • Clusters as cattle

Next Steps

  • Tanzu Mission Control for managing
    cloud-based Kubernetes clusters
  • Deploying assets to multiple clusters
  • Replicating secrets across clouds
  • Lots of fine details to figure out

Questions?

https://go.iu.edu/4QGf