1.2Bandwidth and throughput

"The network is slow" is one of the most common support complaints, yet few words are vaguer. Slowness has separate faces: is the link's capacity simply insufficient, is capacity present but effective speed low, or is capacity ample but response (delay) high? This section distinguishes bandwidth (the theoretical maximum) from throughput (the actual flow), adds latency, and uses two measurement methods—speed test and iperf—to build the basics of turning a subjective "slow" into numbers you can isolate with.

1.2.1Bandwidth, throughput, and latency

• Bandwidth is the maximum transmission capacity a link can theoretically carry (e.g., 1 Gbps). Like the "width of a road," it is the catalog upper bound.
• Throughput is the effective amount of data that actually flows per unit time. Due to overhead, congestion, interference, and device performance, it is always at or below the bandwidth. On a 1 Gbps link measuring 600 Mbps, the throughput is 600 Mbps.
• Latency (delay) is the waiting time until data arrives (e.g., 20 ms round trip). Even with wide bandwidth, high latency feels "unresponsive." Capacity (bandwidth/throughput) and delay (latency) are separate metrics, and not confusing them is the key to isolation.

1.2.2Speed test vs. iperf

• A speed test measures upload/download speed and latency from a browser against a measurement server on the internet. It is handy for checking "the felt speed from home or office to the internet," but the result also mixes in the path and the far server's condition.
• iperf is a tool that measures effective throughput between two endpoints (a sender and a receiver) you set up yourself. Because it isolates the capability of a specific segment—say internal point A to point B—it suits pinpointing "which segment is slow."

Suppose a branch office complains that "copying to the file server is slow." Concluding "the link is 1 Gbps so bandwidth must be plenty—it is imagined" is a wrong call, because bandwidth (the contracted 1 Gbps) is only a theoretical ceiling and differs from the throughput actually achieved. The right approach is to capture the situation in numbers first. Running iperf from a branch PC to the head-office server measures the effective throughput of that specific segment and yields a concrete figure—say only 120 Mbps. Next, running a speed test from the same PC that shows ample speed toward the internet lets you narrow the problem to a specific internal segment, not the internet link. Further, if the copy itself is fine but interactions feel sluggish overall, the issue may be latency (delay) rather than capacity, and you should look at ping round-trip times. The key posture: decompose "slow" into "insufficient capacity (bandwidth/throughput)," "low effective flow on a specific segment (isolate with iperf)," and "slow response (latency)," and turn each into numbers with the right measurement. Rather than swapping hardware on a hunch, measure first and isolate—this is the basic discipline of support.

1.2.3Section summary

• Bandwidth is the theoretical maximum capacity; throughput is the effective amount, always at or below bandwidth—do not confuse them
• Latency measures delay, not capacity; when responses are slow, suspect latency rather than capacity
• A speed test is a quick measurement toward the internet; iperf isolates a specific segment between two of your own endpoints