Composite SLA Calculation: Combining Multiple Service SLAs

What Is a Composite SLA?

When your application depends on multiple services (database, cache, message queue, CDN), the overall availability is determined by how these dependencies combine. A composite SLA is the calculated availability of the entire system based on individual component SLAs.

Serial Dependencies (AND)

When services are chained in series — meaning all must be available for the system to work — you multiply their SLAs:

Composite SLA = SLA_A x SLA_B x SLA_C

Example: Web App → API → Database

Component	SLA
Web Server (EC2)	99.99%
API Layer (ECS)	99.99%
Database (RDS Multi-AZ)	99.95%

Composite = 0.9999 x 0.9999 x 0.9995 = 0.9993 = 99.93%

Three services each at ~99.99% combine to give only 99.93% — lower than any individual component. Each additional serial dependency reduces overall availability.

Parallel Dependencies (OR)

When redundant services are in parallel — meaning the system works if at least one is available — the formula becomes:

Composite SLA = 1 - (1 - SLA_A) x (1 - SLA_B)

Example: Primary DB + Read Replica (either can serve reads)

Composite = 1 - (1 - 0.999) x (1 - 0.999)
         = 1 - 0.001 x 0.001
         = 1 - 0.000001
         = 0.999999 = 99.9999%

Two 99.9% components in parallel give 99.9999% (six nines) for that specific function.

Mixed Architecture Example

User → [CDN (99.99%)]
       → [Load Balancer (99.99%)]
       → [App Server A (99.9%) || App Server B (99.9%)]  ← parallel
       → [Primary DB (99.95%) || Replica DB (99.95%)]    ← parallel

App Servers parallel: 1 - (0.001 x 0.001) = 99.9999%
DB parallel: 1 - (0.0005 x 0.0005) = 99.999975%
Serial chain: 0.9999 x 0.9999 x 0.999999 x 0.99999975 ≈ 99.98%

Key Insight

Adding parallel redundancy improves availability dramatically, but every serial dependency degrades it. The weakest serial link dominates your composite SLA.