If the first age of computing was centred around the concept of the mainframe and massive, centralised computing power and the second was all about the PC and client-server, we are now in the third major era. Cloud computing has delivered massive computing power and storage, at a relatively low cost, that has allowed all sorts of new innovations. But moving from an existing on-prem infrastructure can be challenging. While Australian cloud adoption is quite high, many businesses are still to make the move. I recently spoke with Lee Atchison from NewRelic about how to successfully move to the cloud.
Through NewRelic Atchison has seen many cloud migrations - both successful and those that have provided companies with lots of "learning opportunities". But he has three specific pieces of advice that can maximise the chance of things going well during a cloud migration.
How Do You Want To Measure The Migration?
Atchison said it's critical to set both business and technical key performance indicators (KPIs) so you can measure whether the migration actually worked. He breaks those KPIs into four groups.
- User experience: for example, shopping cart abandonment rates
- Application performance: things like latency
- Infrastructure: processor and memory utilisation, storage performance
- Business: number of transactions per minute, financials
There needs to be a focus on both the customer and system and it's important to focus on items that are both useful and easy to measure.
Have A Migration Architect
Migrating to the cloud can be a complex project with many different parts of the business impacted. By detonating one person and the migration architect, everyone has one person to go to when they have a question, problem or suggestion. The migration architect needs to have experience and be given oversight so they can keep control of the project.
With migration, Atchison said one thing that is always underestimated is the complexity of data migration. During project post-mortems, data migration almost always is identified as the "hardest and most underestimated" part of the project.
Think About Migration Models
Migrations can be executed either by using a Big Bang or phased approach. Atchison called this the master/slave switch or synchronised master model respectively.
In the synchronised master model you replicate your data between the on-rem and cloud databases. The data is kept in sync and applications are progressively moved across to the cloud one at a time. This is more complex but has the easiest migration and least downtime he said.
With the Big Bang, or master/slave, data is copied from the on-site master to the cloud and all the applications are moved in one go. This approach is often used when transition time is a factor.
What Are The Challenges?
Atchison said there were three main challenges he sees. They are:
- People get hung up on less important details such as things like multi-cloud
- Data migration is underestimated
- Migration costs are either missed or underestimated
He added that it is important to be aware of what services you are accessing from cloud providers. While many are very good at that they do, he said, it is important to ensure you balance the value of the services you receive with the longer term impact of vendor lock-in.