As a data centre consultancy, we have planned and executed hundreds of data centre migrations over our 28-year history. From a few hundred up to 12,000 operating system instances within a single client engagement. Across all of those projects there have been consistent lessons we have learnt and incorporated into our migration planning.
1. Planning is the key to success
The old adage “no one plans to fail, they fail to plan” rings true in data centre migrations. Even for what appear to be the simplest of migrations, thorough planning is crucial to a successful migration. We have learnt that planning always takes longer than people expect and so there can be pressure to shorten the duration of planning which is a mistake as it introduces risk into the project. One of the key areas of planning is application profiling and dependency analysis. Without a thorough understanding of how each application works you are likely to break it when you try and migrate it.
2. Migrations are not always successful
Even with thorough planning, migrations are not always successful. This means that planning for failure by building slack into your migration schedule to accommodate second efforts is prudent. You should also have rollback plans in place should you not be able to complete post-move validation tests within the time window required. Unfortunately, you will not always be able to fail-forward so whilst not desirable, rolling back may be your best option and therefore also needs to be planned in detail.
3. Old equipment is temperamental
Ideally replacing old equipment ahead of the migration means there is one less complication. However, that’s not always possible and the old equipment has to be lifted and shifted. In these cases, you should purchase or borrow spare parts for every component of the old equipment prior to the move as insurance. It’s common for old components to fail whether that’s a hard drive, power supply or system board. Of course, you should have also backed up and tested restoring data in case of a failed disk.
4. Migrations are complex
Move events need to be rehearsed so you can move from the theoretical in your planning to the practical in actually testing your migration processes. Dry runs, ghost moves, and contingency actions should be performed for all move types so you can see what doesn’t work and fix it before the real migrations begin. When you start your real migrations, you should begin with low risk applications in a small scale will allow you to further refine your processes to ensure success.
5. Scenario planning and risk mitigation is necessary
You need to think of the scenarios that could disrupt your migrations or cause them to fail. For example, for large migrations you should have multiple move teams for coverage and so that you don’t burn out your resources. For physical moves performing dry runs with the actual trucks to be used for the migration, and identifying and testing alternative routes should be part of your pilot migrations.
In summary migrations are complex and without thorough planning you are likely to miss something that can cause an incident and business disruption. That could be costly to your business and the impact is likely to far exceed the time and cost of a long planning period to ensure your teams are ready and confident of migration success.
CS Technology is an IT consultancy with data centre strategy and migrations at the core of its advisory services. The firm is headquartered in New York with regional offices in Sydney, Melbourne and London. For more information go to CSTechnology.com