L18
Common performance problems: pages 71-74
Cram this first — memory hook
Latency = late; jitter = jumpy arrival; loss = missing; throughput = what got through.
Exam trap
A 1 gigabit-per-second link does not guarantee a 1 gigabit-per-second application transfer. Protocol overhead and other bottlenecks matter.
The slide facts
Page 71 introduces traffic monitoring; page 72 names bandwidth, bottlenecks, loss, latency and jitter. Page 73 links symptoms to utilization, errors, drops, speed/duplex and link state.
If you have time — extra
Bandwidth is the link's capacity; throughput is the rate actually achieved; utilization is how much of the available capacity is being used. Latency is delay, jitter is variation in delay, and packet loss means packets fail to arrive. A bottleneck is the limiting resource along the path: it might be a slow uplink, overloaded processor or constrained service. Think road width, cars actually passing, occupancy, journey time, irregular journey time and missing deliveries. The slide's phrase describing bandwidth as how much traffic is using the link is better read as bandwidth usage/utilization, not the definition of capacity. Cyclic Redundancy Check () errors can indicate damaged frames or a physical-link problem. Drops may reflect congestion or policy, so distinguish them from corrupted frames. Check a baseline rather than declaring every high counter abnormal; counters may accumulate since boot.
Walkthrough
Define the symptom and affected users, compare current rates/counter changes with a baseline, locate the bottleneck, check physical/link and resource evidence, make one justified change, then measure again.