Oh if the world was only Greenfields. If everything was new and we didn’t have to worry about legacy systems, what a wonderful world it would be. But that is not the way of it unfortunately. More times than not using the cloud for our applications and infrastructure means migrating some existing infrastructure from an on premises install up to the cloud. How do you do that successfully?
You have to break some eggs to make an omelet, but migrating to the cloud isn’t like whipping up breakfast. We can’t afford down time, broken links or broken apps. As more and more apps are moved to the cloud, a solid process needs to be put in place. I asked a few people I know to help me put together a checklist of things to consider and do in moving from on prem to the cloud or more likely deploying a hybrid environment with both on prem and cloud.
I asked my friend JP Morgenthal who knows a thing or two about cloud migration and DevOps and I dug in a bit with Alex Rosemblatt from Datadog. Datadog actually has quite a bit of experience in this area. Their product/service monitors both cloud and on prem servers side by side, so they deal with this very situation quite a bit. Taking what both of these gentlemen had to say, here is a checklist that should help you:
- Evaluate application for functional limitations to running in the cloud. This could include operating system dependencies, storage performance, network bandwidth/performance. Some of the key requirements to note:
CPU
Memory
Storage
Attached storage vs block shares
IOPS
DR
Performance
Underlying hardware Choose your method of migration: Lift-n-drop or re-platform. Lift-n-drop is simply porting the application to the cloud without modifying the architecture and providing all the application components. Re-platforming takes into account using cloud services where possible in place of existing application components.
Pick an IAAS vendor, some criteria to use:
Available functionality – auto-scaling, additional services etc
Instance type/AMI availability
Regional zone distribution
Support SLAs
API availability
Performance metrics/log accessPut in place a migration project management plan that includes:
Cutover date selection
Testing project plan
Acceptable performance benchmarks
Cutover day ops
Internal communication and additional training
Ops team
Dev team (s)
Changes in app dev to account for new functionality?
QA team
Management
Non-tech org teams affectedIf hybrid architecture (and it should probably be hybrid), select how the existing data center will integrate with the cloud service. This could be Layer 2 networking through telecommunications provider or Layer 3 through Internet. Now would also be a good time to think about how you will monitor both on prem and in the cloud. Again Datadog does this side by side.
Establish the network scaffolding for the application. Develop the allowed pathways for various application components to communicate. This will simplify the deployment of application components later. Test that the right ports are open for sending and receiving.
Deploy the application in a single subnet with full Internet access. Remove any barriers to ensuring the application will operate on the cloud service provider’s resources. This is simply a test environment to make sure the baseline components will operate and communicate with each other.
Deploy the application into the appropriate subnets to ensure servers are not available via the public Internet.
Port the data from the current application environment
Perform a complete regression test of the system
Add additional testing to see how the system will perform if cloud services become unavailable
This is obviously not an exhaustive list, but should give you a start in your planning. Eventually you are going to have to make a cloud migration if you have’nt already. Better to start planning for it now.
