Content Freshness

The primary mechanism for enforcing freshness. An absolute expiration time or relative lifetime is assigned to each cached object. The cache serves the object without revalidation until the TTL expires. Upon expiry, the cache must perform an origin revalidation using conditional requests like If-Modified-Since or If-None-Match before serving the object again. This model is foundational to HTTP caching and CDN behavior.

A Cache-Control extension that optimizes latency by allowing a cache to immediately serve a stale response to a user while asynchronously revalidating it in the background. This ensures that a user never experiences the latency penalty of a blocking origin fetch. The directive specifies a window of time, in seconds, during which this asynchronous revalidation is permitted after the initial TTL has expired.

An event-driven alternative to TTL expiry where the origin server proactively notifies the cache to purge or update a specific object when the source data changes. This is often implemented via webhooks, purge APIs, or proprietary protocols like Cache Invalidation Protocol (CIP). It guarantees immediate consistency but introduces complexity in ensuring reliable delivery of invalidation events to all distributed cache nodes.

Freshness is quantified through specific HTTP headers. The Age header indicates the time in seconds since an object was generated at the origin. The Cache-Control: max-age directive defines the freshness window. An object is considered stale when its Age exceeds its max-age. Advanced systems also track current age and freshness lifetime to make nuanced eviction and pre-fetching decisions.

When an origin server does not provide explicit freshness directives like max-age or Expires, a cache may employ a heuristic algorithm to estimate a plausible freshness lifetime. A common method, as specified in RFC 7234, is to calculate a fraction (often 10%) of the difference between the Date and Last-Modified headers. This allows for reasonable caching behavior even for legacy content without explicit controls.

For dynamically assembled pages, ESI allows different components to have independent freshness policies. A page's outer template might have a long TTL, while a personalized user greeting fragment has a short TTL or is marked non-cacheable. This fragment-level freshness is assembled at the edge, maximizing cache hit ratios for shared components while ensuring dynamic parts are always up-to-date.

The algorithm that determines which data to remove from a full cache to make space for new content. LRU-K extends standard Least Recently Used by tracking the timestamp of the last K references, preventing one-hit-wonder content from polluting the cache. Other policies include LFU (Least Frequently Used), FIFO (First-In, First-Out), and TTL-based eviction, which directly ties to content freshness by expiring items whose time-to-live has elapsed.

A Cache-Control HTTP extension that instructs a cache to immediately serve a stale (expired) response to the user while asynchronously fetching a fresh version from the origin server in the background. This strategy masks origin fetch latency, ensuring the user never waits on a cache miss. The next request for that content receives the updated version, balancing freshness guarantees with low-latency delivery.

The practice of pre-loading a cache with relevant data before it goes live or during off-peak hours to prevent the cold start problem. Without warming, a new or restarted cache node starts with a 0% hit ratio, forcing all requests to the origin. Warming scripts typically replay historical request logs or use content popularity prediction models to pre-fetch high-demand assets, ensuring high initial cache hit ratios from the moment of activation.

A key performance indicator measuring the percentage of content requests successfully served from a cache versus those requiring retrieval from the origin server. A high hit ratio (e.g., 95%+) indicates effective caching and reduced backhaul load. This metric is directly impacted by content freshness policies—overly aggressive TTL values can invalidate content prematurely, lowering the hit ratio, while overly conservative TTLs risk serving stale data.

The primary mechanism for enforcing content freshness, where each cached object carries a Time-To-Live value set by the origin server. When the TTL expires, the cached copy is considered stale and must be revalidated or evicted. TTL values can be static (e.g., 300 seconds for news articles) or dynamic, adjusted by machine learning models that predict the rate of content change based on historical update patterns.

A proactive caching strategy that uses handover prediction and user trajectory forecasting to pre-place content on the base stations a mobile user will connect to next. This requires tight coordination with content freshness—cached data must remain valid for the duration of the user's journey across cells. If a user moves through three cells in 60 seconds, the cached content's TTL must exceed that transit time to avoid a cache miss mid-session.