-
Notifications
You must be signed in to change notification settings - Fork 0
Resources
| Name | Description | Working when... |
|---|---|---|
| Service | A collection of one or more APIs | Depends on APIs, but usually based on effective request service time |
| Platform | A container for a service. Ranges from physical hosts to AWS Lambda. | ... |
| Link | A network connection | Latency within advertised capability |
| Relay | Routers, load balancers, caches, etc. | ... |
| (name needed) | A deployable platform config, including code | ... |
If an API has a "doStuff" call which has a timeout/retry mechanism then the effective service time is how long it actually takes clients to get a successful result. For example, an API for spinning up new VMs might have an SLA ensuring 99.99% of requests are satisfied within five minutes. If a service speaking this API is timing out but all retries succeed less than five minutes after the request they are repeating then the service is working. On the other hand, if 1% of requests are successful more than five minutes after their original request, the service is degraded.
An API is a contract specifying a protocol by which a client and server communicate about service requests. The contract should include service level information such as how an instance of a service indicates it's promised service levels and an escalation path for handling breaches of those thresholds. The contract may also constrain the behavior of clients and specify how non-compliant clients will be treated.