Split.io provides a suite of server-side and client-side SDKs designed to enable feature flagging, gradual rollouts, and A/B testing. The SDKs act as the local evaluation engine, fetching feature definitions from Split servers and evaluating treatments (variations) locally to minimize latency. This report outlines the core architecture, the concept of "split" synchronization, operational modes, and best practices.
android // ... splits abi reset() include "armeabi-v7a", "x86" universalApk true
| Aspect | Characteristic | | :--- | :--- | | | < 1ms (in-memory hash + rule evaluation) | | Network Dependency | Only for updates (polling/streaming). Flag evaluation works offline. | | Failure Mode | If SDK cannot sync, continues using last known splits. getTreatment returns control . | | CDN Fallback | SDKs retry with exponential backoff if primary API fails. |
Always provide a default value: