Full Page Cache or Varnish: Trade‑off for Concurrent Request Latency

Goal: lower average and 95th‑percentile response time when Magento serves many simultaneous storefront requests.

Constraint: the solution must keep dynamic blocks such as the mini‑cart, login status, and personalized recommendations up‑to‑date without serving stale content.

Uncertainty: Full Page Cache relies on backend storage and can miss when URL variations or session‑dependent blocks differ, while Varnish offloads caching to the HTTP layer but requires accurate ESI tags and purge rules to avoid stale personalized data.

Which approach yields lower latency under a steady load of 200 concurrent users while preserving cart accuracy? What are the operational trade‑offs in managing FPC filesystem/Redis storage versus Varnish ESI configuration and purge complexity? How does each method affect cache hit ratio for personalized pages when TTL values are adjusted?