The majority of Internet users think of "Speed" when it comes to bandwidth and user experience. In reality, the "Speed" users are referring to is available capacity. Popular "Speed" testing tools simply fill the connection with as much data as will fit, neglecting performance, and state the "Speed" on completion.
In reality, the user experience is defined by latency. Latency in-turn determines throughput, which is how long it takes data to complete a round trip. A "Speed" test may say the connection is 1 Gbps, but in practice the throughput will be much, much less.
Achieving throughput equilibrium for a connection with a high consistency of service is the ideal outcome.
Equilibrium is when throughput achieved reaches the maximum achievable throughput, which is defined by latency. A connection that is performing well should be able to reach equilibrium, even when multiple users (also known as sessions) are using the network.
The performance of a connection is determined by latency. It directly affects the throughput rate.
For example, if the minimum trip time for a connection is 25 milliseconds, the maximum throughput rate for one user (session) is 20.9Mbps. Even if the connection capacity is 1Gbps the throughput will only be 20.9Mbps.
The size of a data window is 524,280 bits. The minimum available data windows for a user is one. Dividing the minimum latency (25) into the window size (524280) gives you the maximum achievable throughput.
Consistency of Service refers to the consistency of throughput achieved over the duration of the test in combination with the any reported delay. When delay spikes beyond the trip time it can lead to quality errors, like retransmissions.
Hardware Network Connection Satellite Devices (NCSDs) measure low-level packet data that is used to understand the quality of a network connection.
A comprehensive list of the available quality metrics can be found here.
The hardware NCSDs are required for extended quality metrics but all tests are supported by MCS and software NCSDs. The bandwidth throughput testing abides by RFC 6349, which governs the framework for TCP throughput testing. Due to this, MCS can be relied upon to provide true insight into what the user experiences.