Akamai Web Performance flashcards
Akamai Web Performance · Build recall with a focused review session.
32 cards. Read the term and the definition here. The deck at the top of this page is optional.
- Round-trip time (RTT)
- The time for a packet to travel to a destination and back. Each new connection adds handshake RTTs before any content can be delivered.
- HTTP/2 multiplexing
- Sends many concurrent request streams over a single TCP connection, eliminating per-request connection setup but sharing TCP head-of-line blocking.
- QUIC and HTTP/3
- A UDP-based transport with independent streams so packet loss on one stream does not stall others, plus faster connection resumption.
- Anycast routing
- Advertising the same IP address from many locations so BGP routes each user to a topologically nearby point of presence.
- Cache hit ratio
- Cache hits divided by total requests. Higher ratios mean more content served from edge and less origin load.
- Time to first byte
- Elapsed time from request initiation to the first response byte, combining DNS, connection setup, server processing, and network latency.
- Property version
- An immutable snapshot of a property's rules and behaviors. It must be activated on a network before it affects live traffic.
- Rule matching order
- Rules evaluate top-down and the first matching rule wins unless child rules or fallthrough criteria are configured to continue evaluation.
- Behaviors versus match criteria
- Match criteria decide when a rule applies, such as hostname or path; behaviors decide what happens, such as caching or origin selection.
- Property Manager API
- A REST interface for creating properties, editing versions and rules, managing hostnames, and triggering asynchronous activations.
- Activation rollback
- Restoring prior behavior by activating an earlier property version on the network, since live configuration cannot be edited in place.
- Include rules
- Reusable rule sets shared across properties; the including property can override specific behaviors where it defines them.
- s-maxage directive
- A Cache-Control directive setting freshness lifetime for shared caches such as CDN edges, overriding max-age for those caches.
- Cache key composition
- The set of request attributes that identify a cached object, typically hostname, path, query string, and selected headers or cookies.
- stale-while-revalidate
- Allows a cache to serve stale content for a defined window while asynchronously fetching a fresh copy from the origin.
- stale-if-error
- Permits serving stale cached content when the origin returns an error or is unreachable, improving availability during incidents.
- Tag-based purge
- Invalidation using cache tags or CP codes rather than enumerating individual URLs, useful when content has many URL variants.
- Tiered distribution
- Inserts a parent cache tier between edge servers and origin, raising edge hit ratio and shielding the origin from repeated misses.
- Vary header impact
- Varying on high-cardinality headers like User-Agent or Cookie fragments the cache and sharply lowers hit ratio.
- Render-blocking resources
- CSS and synchronous scripts in the head delay first paint. Defer or async scripts and inline critical CSS to shorten the critical path.
- Preload versus prefetch
- Preload fetches resources needed for the current navigation at high priority; prefetch fetches likely future resources at low priority.
- Preconnect hint
- Opens DNS, TCP, and TLS to an origin ahead of need, removing setup latency from the critical request path.
- Edge compute
- Running logic such as personalization, header rewriting, or response assembly at the edge to avoid origin round trips.
- Brotli compression
- A compression algorithm that generally outperforms gzip on text assets; avoid applying it to already-compressed media files.
- Cumulative layout shift
- A Core Web Vital measuring unexpected visual movement of page elements, often caused by images without reserved dimensions.
- Synthetic versus RUM
- Synthetic tests run scripted checks from fixed locations; real user monitoring captures actual user experiences across devices and networks.
- Waterfall breakdown
- Splits request timing into DNS, connect, TLS, TTFB, and download phases to identify whether latency is network, edge, or origin.
- Staging versus production activation
- Staging uses a separate network and hostname for validation; production activation serves live traffic and should follow staging verification.
- DataStream logs
- Per-request log delivery that lets teams correlate traffic shifts with purges, activations, or origin errors.
- Hit ratio drop causes
- Common causes include cache-key changes, purge storms, query-string variance, and new cookie or header variation in requests.
- HSTS policy
- A response header instructing browsers to use HTTPS only for a domain, preventing protocol downgrade and mixed-content exposure.
- Origin shielding
- A parent or shield tier that absorbs traffic so the origin sees fewer requests and is less exposed during spikes or attacks.