A database that works is invisible. One that fails takes everything with it. Many teams give database care to whoever knows it best, which is rarely a dedicated specialist. Outsourcing database administration can fix that, but only if you are clear about what you hand over and what you keep.
What a database administrator actually does
- Backups, and tested restores
- Performance tuning and slow-query analysis
- Replication and high availability
- Upgrades and migrations
- Access control and auditing
- Capacity planning
What to hand over, and what to keep
Routine operations are a good fit for a provider: backups, monitoring, tuning, patching and upgrades. You should keep ownership of the data itself, decisions about schema and application behaviour (with your developers), and approval of destructive or irreversible changes.
Before you hand anything over, build an inventory
For every database, write down:
- The engine and version
- Its size and growth rate
- Which application and which team own it
- How critical it is to the business
- Where it runs, and where its backups are stored
Two numbers to agree on
Recovery point objective (RPO) is how much data you can afford to lose, measured in time. Recovery time objective (RTO) is how long you can afford to be down. Your backup and replication design follows from these two numbers, so agree them with the business before you agree a price.
Questions to ask a provider
- Which engines and versions do you support, including the ones we run?
- How often do you test restores, and can you show the result?
- How will you access production data, and who can see it?
- How do you test an upgrade before running it on production?
- What do you do when a query suddenly slows down?
- How do you handle data location and privacy requirements that apply to us?
- What documentation do we receive, and who owns it?
- How is the work priced, and what triggers extra cost?
Red flags
- Backups with no restore tests
- Unrestricted access to production data with no audit trail
- Upgrades run directly on production
- No clear owner for each database
Databases for AI workloads
If your databases feed AI applications, such as search over your documents or analytics, the same discipline applies, with extra attention to data freshness and access. Treat those stores like any other production database.
See our database administration and supportservice, or raise a ticket if you have a database problem now.