Many organisations are in the process of migrating services to the cloud as part of their digital transformation journeys. One potential advantage is to improve scalability and availability, but for this to be achieved migrations must be carefully planned.
When is a quick win, a quick fail?
There is often a focus during these large, complex and expensive projects, to prioritise workloads based on ‘the quick wins’.
What is often overlooked is the business continuity (BC) and disaster recovery (DR) scenarios for workloads. Enterprise services are complex beasts often with a web of unmapped dependencies. If these dependencies are not properly understood and used to plan co-migration and update DR scenarios and plans, then the service is likely to end up with an unrecognised BC & DR risk, leading to a potentially big outage. These oversights could be as simple as migrating a tightly coupled service and not recognising the dependencies that exists on-premise, thereby creating additional dependencies on the connectivity between the cloud provider and the data-centre.
How to avoid a cloud disaster?
- Use your architecture team and specialised products to help build dependency maps that are accurate pre-migration, for example - AWS Application Discovery Service
- Use these maps to understand tightly coupled services and plan migration order, rather than focussing on ‘quick wins'
- Keep your dependency maps up to date as you migrate, this will be a continuously changing and complex picture
- Don’t assume that migration automatically covers backup, BC & DR risks
- Engage existing BC & DR teams, they might need bringing up to speed on the nuances of cloud, although they will be the ones who understand existing BC & DR plans
- Don’t forget to consider the newer threats to BC & DR scenarios, including ransomware attacks encrypting data
- Be aware of SLA stacking when multiple 3rd parties are involved, your downtime might be greater than you initially estimate
- Create tabletop disruption scenarios in order to understand the impact
- Test post migration with a variety of disruption scenarios before any service switch over