MCP Time Server Node

# NodeCache Memory Tracking - Verified Behavior **Date**: 2025-07-20 **NodeCache Version**: 5.1.2 **Node.js Version**: 18.19.0 ## Key Findings ### 1. Built-in Memory Tracking NodeCache provides `getStats()` which returns: ```javascript { hits: number, // Cache hits misses: number, // Cache misses keys: number, // Number of keys ksize: number, // Total size of all keys (in chars) vsize: number // Total size of all values (mysterious units) } ``` ### 2. vsize Calculation - **NOT** based on JSON.stringify length - vsize seems to use a fixed overhead calculation - For our typical cache entry (144 bytes JSON), vsize reports 400 bytes - Ratio is inconsistent - small objects have high overhead, large objects have low overhead ### 3. Memory Usage Observations - Empty cache: 0 bytes - Typical time entry: 400 vsize units (330 actual heap bytes) - 10MB = ~30,000 entries based on actual heap measurement - vsize updates correctly on set/delete operations ### 4. vsize Reliability Issues Testing showed vsize is NOT accurate for memory limits: - 1KB string object: vsize reported only 80 bytes increase - Actual memory usage is 3-4x higher than vsize for typical entries - vsize appears to count characters, not actual memory bytes ## Conclusion NodeCache's vsize is not suitable for enforcing a 10MB memory limit. We need to: 1. Track actual memory usage ourselves 2. Use process.memoryUsage() for verification 3. Implement our own size calculation ## Recommended Approach Create a wrapper that tracks memory using JSON.stringify length as a consistent approximation, with a safety factor to account for object overhead.