Computer Networks flashcards, generated for you.

Bandwidth is the maximum theoretical data rate a link can support (capacity in bits/sec). Throughput is the actual data rate achieved in practice, always ≤ bandwidth due to overhead, congestion, and protocol inefficiencies.

Latency is total delay from source to destination. Components: (1) propagation delay (distance/speed of light), (2) transmission delay (packet size/bandwidth), (3) queueing delay (waiting in buffers).

Jitter is variation in packet arrival times. It matters because real-time apps (VoIP, video streaming) require bounded delays; excessive jitter causes poor quality even if average latency is acceptable.

Reliability (retransmissions, acknowledgments, error checking) adds latency. TCP prioritizes reliability; UDP prioritizes low latency. Choose based on app needs: financial apps need TCP; gaming/streaming often prefer UDP.

UDP: stateless, no setup/teardown, low overhead, no guarantees. TCP: establishes connection (3-way handshake), maintains state, ensures ordered delivery and retransmission. TCP adds overhead; UDP is faster but unreliable.

Packet loss occurs when packets are dropped due to congestion or errors. TCP detects loss via timeouts/ACKs and retransmits (reduces throughput, increases latency). UDP ignores loss; apps must handle it or accept degradation.

Congestion control prevents network overflow by limiting sender transmission rate when congestion is detected (via loss or high latency). Without it, networks enter congestion collapse where throughput drops dramatically.

Non-persistent: new TCP connection per request (high overhead, 3 RTTs minimum per object). Persistent: reuse one TCP connection (reduces latency and server load). Trade-off: persistent requires longer server resource allocation.

Flow control: receiver tells sender how much data it can accept (via window size), preventing receiver overflow. Buffering: storing data temporarily in memory/queues. Flow control uses buffering; good buffering enables high throughput despite latency.

Hop-by-hop (e.g., per-link retransmission) adds overhead at each router. End-to-end (TCP retransmits only at source) is simpler and scales better but cannot detect mid-path corruption unless apps add checksums. Modern networks prefer end-to-end.