A Service Level Agreement (SLA) is more than just fine print at the bottom of a contract. In the world of distributed systems and AI infrastructure, it is a formal, quantifiable commitment between a service provider and a customer.
The Core Components of an SLA
While every organization tweaks the language, a robust SLA typically consists of three non-negotiable pillars:
- Service Level Objectives (SLOs): The target performance level. This is the goal. For example, "99.9% availability" or "response time under 200ms for 99% of requests."
- Service Level Indicators (SLIs): The actual metric used to measure compliance with the SLO. This defines how you measure success. Is it the percentage of HTTP 200 OK responses? The latency measured at the load balancer?
- Penalties & Remedies: The "or else." If the provider misses the SLO, what happens? Usually, this involves service credits (money back) or the right to terminate the contract.
Types of SLAs
Not all SLAs are created equal. In complex organizations, you'll encounter different tiers:
Customer-Based SLA
A specific agreement with a single customer group covering all the services they use. Examples include custom enterprise contracts for large banks using a public cloud.
Service-Based SLA
A standard agreement for all customers using a specific service. For instance, the standard 99.9% uptime guarantee provided to all users of an email marketing SaaS.
Why They Matter for Engineering
For engineering teams, SLAs translate business requirements into technical constraints. They dictate your error budgets. If you have a 99.9% SLA, you know exactly how much downtime you can afford before you start burning cash in refunds. This clarity allows teams to balance feature velocity against reliability work effectively.
Recover your SLA credits
Stop leaving money on the table. Breachr automates the evidence gathering and negotiation process for you.