SLAs, Explained
What is an SLA?
A Service Level Agreement (SLA) is basically a promise between you and the cloud provider about how reliable a service will be. It’s measured in uptime percentage, usually across a month. It looks something like this:
- DigitalOcean - 99.99% uptime on Droplets
- AWS - 99.99% for multi-AZ EC2 deployments
- Azure - 99.9% for Storage accounts
- Oracle Cloud - 99.95% for Autonomous Database
99.99% uptime means your service can still be down for about 4 minutes and 23 seconds per month (or roughly 52 minutes per year) and still meet the SLA (not a math whiz, I used this calculator 🤓).
Why The Downtime?
Cloud providers can’t guarantee perfection. Power outages happen, bugs creep in, and the interns.. do their thing. Instead of saying “trust us,” they give you a number - an SLA - that defines how much downtime they consider acceptable.
Even with multiple Availability Zones or regions, failures can ripple across services. SLAs exist to acknowledge these realities - they set boundaries around what level of downtime is considered acceptable, rather than promising perfection.
The Numbers Behind the Nines
Here’s what those uptime percentages mean in real terms:
- 99% (two nines) - ~7 hours downtime/month
- 99.9% (three nines) - ~43 minutes/month
- 99.99% (four nines) - ~4 minutes/month
- 99.999% (five nines) - ~26 seconds/month
Examples:
- DigitalOcean - Droplets promise 99.99%, so ~4 minutes downtime is acceptable
- AWS - S3 offers 99.99%, but EC2 in a single AZ is only 99.5%
- Azure - SQL Database has a 99.99% SLA with geo-replication
Why It Matters
The point of an SLA is to remind you that not all setups are equal. Running just one virtual machine is like balancing everything on a single chair. If it breaks, you're stuck. That's why AWS encourages using multiple zones, and Oracle offers features like Data Guard for their databases.
It's also important to set expectations. “Four nines” sounds impressive, but even with Azure’s zone-redundant storage, there is still a chance for short outages. And of course, higher reliability usually means higher cost. Whether it is paying extra for AWS multi-region setups or Oracle’s advanced redundancy, better uptime almost always comes with a bigger bill.
The TAM Lens
SLAs sound impressive on paper, but they are not magic spells. A lot of organizations assume that a “four nines” promise means smooth sailing forever. In reality, an SLA is more like a warranty - reassuring, but it won’t stop things from breaking.
Final Thoughts
SLAs aren’t magic shields - they’re more like insurance policies with fine print. The real skill is designing your systems so you’re not completely dependent on a single SLA. After all, a 99.99% sunny forecast still means there’s a 0.01% chance your weekend gets ruined 😞.
