The Gold Standard of Wireless Communication | Separating Airtime Fairness from Throughput Fairness

Airtime fairness asks how transmission time is shared; throughput fairness asks how delivered data rates compare. Because stations can transmit at different physical rates and incur different overheads or retries, equal airtime does not imply equal throughput, and equal throughput can require unequal airtime.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

In a simplified payload-only model, a station at 12 megabits per second and one at 120 each receive 50 milliseconds. They can send about 0.6 and 6 megabits respectively. Time is equal; delivered bits are not.

The practical check is this: state the fairness objective before deciding that an allocation is fair. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Next chapter

Wireless stations contend for a channel over time. The resource is not merely a headline channel rate; contention, interframe spaces, acknowledgements, aggregation, retransmissions and interference consume airtime.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

If a 2 millisecond exchange includes only 1.2 milliseconds of payload transmission, counting payload time alone hides 0.8 milliseconds of shared-channel use.

The practical check is this: choose and document the airtime accounting boundary. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

The physical-layer rate is the instantaneous signalling rate for part of a transmission. Goodput is unique useful payload delivered per elapsed time. Protocol overhead, losses and retries separate them.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

A 120 megabit-per-second PHY can deliver 60 megabits per second of unique payload over an interval after overhead and retries. Calling 120 the application throughput doubles the useful result.

The practical check is this: use unique delivered payload and the same observation interval for comparisons. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Use a stipulated model to see the distinction before adding protocol complexity. Equal time lets the faster link move more payload, which may maximise total bits while giving unequal throughput.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

Station A transmits useful payload at 10 megabits per second for 0.4 seconds and delivers 4 megabits. Station B delivers at 50 for 0.4 seconds and delivers 20 megabits. Both receive 40 per cent airtime; their goodputs over the one-second interval are 4 and 20 megabits per second.

The practical check is this: include idle and other-use time so shares sum over a defined interval. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

To deliver equal payload to a slower and faster station, the slower station generally needs more transmission time in the simplified model. Throughput equality therefore consumes airtime asymmetrically.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

To deliver 5 megabits, a 10 megabit-per-second link needs 0.5 seconds while a 50 megabit-per-second link needs 0.1 seconds. Equal payload requires an 83.3 versus 16.7 per cent split of their combined 0.6 seconds.

The practical check is this: ask whether equal delivered service justifies the capacity cost in this context. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Contention rules that give similar transmission opportunities can let slow transmissions occupy much more time, reducing aggregate performance. The effect is a mechanism, not evidence that the slow user behaved badly.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

If each station sends one equal-size frame per turn, a frame at 6 megabits per second holds the medium roughly nine times as long as one at 54 before shared overhead. Opportunity fairness is not airtime fairness.

The practical check is this: measure duration and rate rather than counting successful frames alone. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Equal numbers of wins, frames or packets do not establish equal time because frame size, rate, aggregation and retries differ. Opportunity metrics need a conversion to the resource being governed.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

One station wins ten times and sends 1,500-byte frames slowly; another wins ten times and sends aggregated bursts quickly. The win counts match, while occupied time and payload can differ sharply.

The practical check is this: record duration, bytes and retries for each successful opportunity. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

A failed attempt consumes shared time but delivers no new payload. Excluding retry airtime makes a poor link look more efficient and shifts its cost onto others invisibly.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

A station uses 30 milliseconds for a failed attempt and 35 for the successful retry. Its service cost is 65 milliseconds for one delivered payload, not 35.

The practical check is this: count all relevant attempts within the observation boundary. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Frame aggregation can amortise fixed overhead and allow more payload per transmission opportunity. Equal airtime with different aggregation efficiency can still yield different goodput.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

If each burst has 0.2 milliseconds of fixed overhead, carrying ten frames in one burst pays it once while ten separate bursts pay it ten times. The same payload then consumes different airtime.

The practical check is this: compare aggregation settings and payload sizes before attributing the gap to fairness scheduling. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

An access point may serve many downstream flows while each client contributes one uplink contender. Per-station, per-flow and per-direction fairness can therefore produce different outcomes.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

If one access point contends as one station for five downlink users while five clients each contend uplink, equal station opportunities do not give each flow an equal share.

The practical check is this: name whether the unit of fairness is station, flow, user, service class or direction. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Averages over minutes can hide bursts of starvation that harm interactive work. Short windows are noisier but reveal delay; long windows describe accumulated share.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

Two stations can each receive 50 per cent over a minute while one waits several seconds during a burst. The minute average is fair by one definition and poor for a deadline-sensitive task.

The practical check is this: select a window connected to the application’s delay tolerance. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Driver counters, monitor captures and controller summaries may use different boundaries or omit hidden activity. Treat tool output as a defined measurement, not ground truth without metadata.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

A controller may report transmit airtime for its own frames while external interference appears only as channel busy time. The two values answer different questions about occupancy.

The practical check is this: document counter definitions, clock interval and missing traffic. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Busy time caused by other networks or undetected transmitters can reduce usable capacity without being assigned to a managed station. A per-station fairness report can look balanced while service is poor.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

If 40 per cent of the interval is externally busy, two managed stations may split the remaining 60 per cent evenly. Each gets 30 per cent of wall-clock time, not 50.

The practical check is this: keep managed allocation separate from total channel availability. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Fairness objectives depend on service need. A low-bitrate voice flow may require timely small transmissions, while a backup can absorb residual capacity. Equal throughput would be a poor objective.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

Give the voice flow enough scheduled opportunity to meet its delay and loss need, then evaluate the backup on remaining goodput. This is a policy example, not a universal configuration.

The practical check is this: state priority, deadline and minimum service without calling every unequal rate unfair. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

A distant station may use a robust lower rate and more retries. Removing it improves aggregate throughput but may violate coverage or access goals. Engineering and service policy must be joined.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

A scheduler can cap excessive airtime while preserving a minimum service, but the correct boundary depends on the intended network, user need and alternatives.

The practical check is this: test coverage, placement and interference before blaming the client. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

A fairness index can summarise equality across selected values, but it cannot choose whether those values should be throughput, airtime, utility or normalised service. A high index on the wrong quantity is misleading.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

For throughputs 5 and 5, Jain’s index is 1. For airtimes 0.8 and 0.2, the airtime index is lower. Both calculations can be correct and describe different fairness questions.

The practical check is this: label the metric vector and why equality is desired. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Allocating more time to faster links can raise aggregate goodput; allocating time equally can protect access; equalising throughput can heavily favour slower links. There is no metric-free optimum.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

Compare total payload under equal airtime with equal throughput in the two-rate example. The difference is the cost of the selected fairness goal under stipulated assumptions.

The practical check is this: show the trade-off curve or scenarios instead of hiding the policy choice. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Application delay, queueing, Internet path performance and server behaviour sit beyond Wi-Fi airtime. A fair radio schedule does not prove that a learning portal works.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

Two devices can receive equal airtime while one application waits on a remote service. Conversely, unequal airtime may still meet both application needs.

The practical check is this: test the complete task journey alongside radio counters. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Use named devices, traffic direction, offered load, distance, channel, bandwidth, rate adaptation state and observation window. Repeat scenarios and preserve raw counters or captures.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

A simple matrix varies one slow station present or absent while fixed clients transfer known payload. Compare airtime, retries, unique goodput and latency, rather than running unrelated speed tests.

The practical check is this: change one material condition at a time and record the starting state. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Equal RSSI is not equal rate; equal rate is not equal goodput; equal packets are not equal airtime; equal airtime is not equal user outcome. Each equivalence skips a mechanism.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

If two stations each deliver 1,000 packets, inspect packet size, retry count, aggregation and elapsed time before declaring fairness.

The practical check is this: ask which resource, benefit and time window the equality claim refers to. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Airtime is often closer to the scarce radio resource than throughput alone. Estimating demand in airtime can reveal why apparently modest traffic saturates a slow link.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

A 12 megabit-per-second user needing 6 megabits per second consumes about half the simplified payload airtime; a 120 megabit-per-second user needing the same payload consumes about five per cent before overhead.

The practical check is this: include realistic overhead and retry margins before deployment decisions. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Next chapter

Research on the 802.11 performance anomaly and airtime fairness shows why queueing and time allocation matter in multi-station Wi-Fi, while Airtime Fairness for IEEE 802.11 Multirate Networks analyses multirate allocation. Continue with Accounting for Retry Time in Unique Payload Goodput and Separating Signal Strength from Signal-to-Noise Ratio.

A useful way to reason through this chapter is to write the measured quantity, its unit, its boundary and the decision it is meant to support before doing any arithmetic. Then write the mechanism that connects the observation to the claim. This short discipline prevents a familiar label from silently carrying assumptions about time, place, equipment or population. It also makes a calculation reproducible: another reader can see which values were observed, which were stipulated for the teaching example and which conclusion remains conditional.

The worked equations omit many protocol details to expose the distinction. They are not configuration instructions or guarantees for a specific product.

The practical check is this: use current standards, equipment documentation and competent network measurement for operational changes. Record the result that would change your mind as well as the result you expect. A Gold Standard explanation should survive a changed context, show the limit of the model and direct a real operational decision to the relevant qualified person or current official guidance.

Contents · Previous chapter · Return to Wireless Communication