16.3 Network Performance: Bandwidth, Latency, Load, and Estimating Needs
Key Takeaways
- Bandwidth is the maximum data rate of a connection, usually in bits per second (Mbps, Gbps); throughput is the rate actually achieved.
- Latency is the delay for data to travel from sender to receiver, usually measured in milliseconds; it matters most for real-time uses such as video calls and online games.
- Load is the amount of traffic on a network or server; as load approaches capacity, queues form, so latency rises and throughput per user falls.
- To estimate bandwidth, multiply the rate per user by the number of simultaneous users: 30 students streaming at 5 Mbps need about 150 Mbps.
- Data size = bit rate × time: a 2-hour stream at 4 Mbps is 4 × 7,200 = 28,800 megabits, or 3,600 megabytes (about 3.6 GB).
What this competency asks
ETS asks you to be familiar with factors that have an impact on network functionality:
- Define basic terminology (for example, bandwidth, load, latency).
- Estimate the necessary bandwidth and data size for a given situation.
- Identify critical resources for a given situation.
Key terms
| Term | Meaning | Unit | Analogy |
|---|---|---|---|
| Bandwidth | Maximum data rate a link can carry | bits per second (kbps, Mbps, Gbps) | Number of lanes on a highway |
| Throughput | Data rate actually achieved | bits per second | Cars that actually get through per minute |
| Latency | Time for data to travel from source to destination | milliseconds (ms) | Travel time for one car |
| Round-trip time (RTT) | Time to send a request and receive the reply, often measured by ping | ms | There and back |
| Load | Amount of traffic, or demand, on a link or server | Utilization, such as 80% | How crowded the highway is |
| Jitter | Variation in latency from packet to packet | ms | Travel times that are unpredictable |
| Packet loss | Share of packets that never arrive | % | Cars that never make it |
Bandwidth and latency are different. A satellite link can have high bandwidth, so large downloads go quickly once they start, but high latency, so every click waits. A connection can also have low latency with little bandwidth.
Where latency comes from
- Propagation delay: signals travel at a finite speed. Light in fiber covers roughly 200 km per millisecond, so crossing a continent and back takes tens of milliseconds. Geostationary satellites add hundreds of milliseconds.
- Transmission delay: the time to push all of a packet's bits onto the link. It is shorter with more bandwidth.
- Queuing delay: time spent waiting in router buffers. It grows with load.
- Processing delay: time for devices to examine and forward packets.
How load affects performance
As traffic approaches a link's capacity, packets wait in queues, so latency rises. Each user's throughput falls because users share the bandwidth. If buffers overflow, packets are dropped. That is why video buffers at 8 p.m., when everyone is streaming.
Estimating bandwidth
Total bandwidth ≈ bandwidth per user × number of simultaneous users (plus headroom).
| Situation | Estimate |
|---|---|
| 30 students each stream video at 5 Mbps | 30 × 5 = 150 Mbps |
| 25 students on video calls at 2 Mbps each, both directions | About 50 Mbps down and 50 Mbps up |
| A teacher livestreams to parents at 6 Mbps | At least 6 Mbps of upload bandwidth |
Leave headroom, often 20–30%, because other traffic shares the link and real throughput is below the advertised bandwidth. If the school's connection is 100 Mbps, the first scenario will not work well. That link is the bottleneck.
Estimating data size and transfer time
Data size = bit rate × time. Transfer time = data size ÷ throughput. Convert bytes to bits by multiplying by 8.
Example 1: A 2-hour lecture streams at 4 Mbps. 4 Mb/s × 7,200 s = 28,800 megabits. Dividing by 8 gives 3,600 MB, or about 3.6 GB.
Example 2: A 500 MB software update over a 50 Mbps connection. 500 MB × 8 = 4,000 megabits. 4,000 ÷ 50 = 80 seconds at best.
Example 3: A school backs up 2 TB of data nightly over a 1 Gbps link. 2 TB = 2,000,000 MB = 16,000,000 megabits. 16,000,000 ÷ 1,000 Mbps = 16,000 seconds, about 4.4 hours, at full speed with nothing else on the link.
Identifying critical resources
Different uses depend on different resources. Find the one that limits the task.
| Use | Critical resource(s) | Why |
|---|---|---|
| Streaming high-definition video | Download bandwidth | Sustained high data rate |
| Video conferencing | Latency, jitter, and upload bandwidth | Real-time, two-way communication |
| Competitive online gaming | Low latency and low jitter | Small messages, but every millisecond of delay matters |
| Uploading a large video or nightly backups | Upload bandwidth and time | Large volume sent from the site |
| Sending an email | Little of anything | Small data; delays of seconds are acceptable |
| Remote surgery or robot control | Very low latency and high reliability | Delays or dropped packets are dangerous |
| A website expecting a traffic spike | Server capacity and bandwidth at the server | Load concentrates at one point |
| Online payments | Reliability and security | Transactions must complete correctly |
Bottleneck thinking: throughput is limited by the slowest link on the path. That might be the Wi-Fi in the room, the school's Internet connection, or the server itself. Upgrading any other part does not help.
Improving performance
- Add bandwidth where the bottleneck is, or reduce demand, for example with lower video resolution.
- Reduce latency: use wired connections, serve content from nearby content delivery networks (CDNs), and avoid satellite links for interactive uses.
- Manage load: schedule backups and updates overnight, and use quality of service (QoS) settings to prioritize video calls over downloads.
- Cache frequently used content locally.
Which term describes the delay between when data are sent and when they arrive at their destination?
Thirty students in a classroom will each stream an instructional video at 5 Mbps at the same time. Approximately how much bandwidth does the class need, before adding headroom?
For a fast-paced online multiplayer game that sends small messages many times per second, which network characteristic is most critical?
A 2-hour lecture video streams at 4 megabits per second. Approximately how much data does the stream use?