# Add Sliding-Window Get and Put Activity Metrics to a Key-Value Store Extend an in-memory key-value store with `averageGet(window)` and `averagePut(window)` activity metrics for sliding windows of one hour, one day, and one week. In this practice contract, each metric is the average operation rate over the full requested window: the number of completed operations of that type whose completion timestamps lie in `(now - window, now]`, divided by the window length in seconds. Do not measure operation latency. Avoid rescanning all history on every query. ### Constraints & Assumptions - Timestamps are monotonic and windows use operation completion time. - Supported window lengths are exactly `3600`, `86400`, and `604800` seconds. - Metrics must not change key-value correctness. ### Clarifying Questions to Ask - Is the intended meaning of “average” a full-window request rate, as defined here, or another source-provided denominator? - Must metrics survive restart? ### What a Strong Answer Covers - Timestamped get/put counts, eviction, concurrency, and memory bounds. - Clear behavior at window boundaries and under idle periods. ### Follow-up Questions - How would approximate buckets reduce memory? - How would you expose peak rates without confusing them with the requested average?