Uni Reading List
Study guides /

Computer networks: choose a text and account for every delay

Packet diagrams become useful when their time axis and accounting units are explicit. This route compares two authorized free texts with a commercial application-first survey. Choose the explanation that addresses your weak point, then calculate a small idealized exchange before interpreting a real trace.

Course topic: Computer Networks

Study route and exercise: University Reading List editorial project. How we curate · Sources checked:

Download blank worksheet (Markdown) ↓

Scope and responsibility

An introductory networking study route with basic programming, binary representation and elementary algebra. The examples use an invented lossless link, not measurements, a configuration recipe or a prediction of an actual TCP connection. Dordal’s second-edition family, a fixed CNP3 PDF and Pearson 9e are distinct resources; changing web content is not silently treated as the same printed snapshot.

Choose your starting point

I want a free mechanism-first route

Use Dordal for packets, delay and abstract windows. Its historical contrasts can clarify mechanisms without asserting how every current network operates.

I want principles followed by protocols

Use the fixed CNP3 PDF, separating Part 1 principles from Part 2 protocol descriptions and Part 3 practice. Record its release and build date before comparing another version.

My course uses the application-first survey

Use the assigned ninth edition of Kurose/Ross and its verified chapter sequence. A current print identifier is not a promise of included platform access.

Compare the three books

An Introduction to Computer Networks

Peter L. Dordal · Edition: 2 · 2020

Free packet-and-mechanism spine

Choose it when
You want detailed diagrams and explanations of delay, forwarding and transport mechanisms.
Use another route when
You need a current deployment manual or assume every historical technology remains a recommended choice.
Selected reading
Second-edition online revision 2.0.11 checked here: 1.2, 7.1–7.3 and 8.1–8.3; routing chapters are later selections.
Limits and trade-offs
The author dates the second-edition family to 2020 and supplies open reading. The noncommercial/no-derivatives notice has its own terms; free reading does not imply unrestricted adaptation.

Try before choosing

After 7.1 and 8.1, draw one full stop-and-wait cycle with separate data and ACK serialization and propagation.

Check your work: A labeled timeline ends when the complete ACK returns. A rate in bits/s cannot be added to a delay in ms.

Official free reading ↗

Book record and access options →

Computer Networking: Principles, Protocols and Practice

Olivier Bonaventure · Edition: 3; Release 2021, PDF built May 15, 2024

Free principles/protocols/practice route

Choose it when
You want one structured path from a simple communication model to concrete Internet mechanisms.
Use another route when
You need an unchanging latest-protocol reference or expect an old PDF to include every new repository section.
Selected reading
Fixed PDF: Part 1 sections 2.1 and 2.6; 2.4 for transport principles; select Part 2 protocols and Part 3 practice after the model.
Limits and trade-offs
The university PDF labels Release 2021, built 15 May 2024, 418 pages. Its CC BY 3.0 notice differs from the repository README’s CC BY-SA 3.0; follow the format actually used.

Try before choosing

Read 2.6 and create a units ledger for payload bytes, header bytes, line rate, one-way delay and ACK size.

Check your work: Convert each frame to bits before division. State which overheads are included; calling a link 2 Mb/s does not make its payload throughput 2 Mb/s.

Official free reading ↗

Book record and access options →

Computer Networking: A Top-Down Approach

James F. Kurose; Keith W. Ross · Edition: 9 · 2025

Application-first ninth-edition survey

Choose it when
Your syllabus follows Kurose/Ross or you want applications before the underlying transport/network layers.
Use another route when
A free first route meets your needs, or an eighth Global Edition is being sold as the selected ninth edition.
Selected reading
Ninth edition: 1.4 for delay/loss/throughput, 1.5 for layers, 3.4 for reliable transfer; then 4–6 for data/control/link distinctions.
Limits and trade-offs
Pearson confirms 2025 print ISBN 9780135429334, separate eText 9780135415603 and ©2026. Regional date listings differ; retain the verified year rather than invent another edition.

Try before choosing

Use 1.4 and 3.4 to compare two claims: fast physical line rate and fast application delivery. List what must be known to connect them.

Check your work: Your explanation includes framing, waiting and feedback assumptions. A measured download also needs conditions absent from this paper model.

Book record and access options →

Check your preparation

Serialization versus propagation?

Serialization places bits onto a link and depends on frame length/rate. Propagation moves a bit along the link and is supplied separately here. They can overlap for different bits.

What is being called throughput?

Define useful payload bits per elapsed time. A frame rate, raw link rate and payload rate have different numerators or denominators.

Why not call the window result TCP performance?

The abstract model omits loss, congestion control, receiver limits, processing, queues and variable routes. A mathematical bound under these assumptions is not a measurement or a complete protocol model.

Six steps with a concrete output

These are your own study marks, not an assessment. They are not sent to this site.

  1. Choose the accounting unit

    Selected reading: Dordal 1.2 and 7.1; CNP3 section 2.6.

    Make this: Make a table of every given quantity and convert bytes to bits.

    Check your work: Decimal Mb means 1,000,000 bits; distinguish the 1200-byte payload from the 1280-byte data frame.

  2. Draw causality before adding time

    Selected reading: Dordal 7.1 and 8.1.

    Make this: Mark first/last data bits and first/last ACK bits on a sender/receiver timeline.

    Check your work: The receiver starts its ACK only after the entire data frame arrives; each direction has its own supplied propagation delay.

  3. Compute the full cycle

    Selected reading: CNP3 principles and Kurose/Ross 1.4.

    Make this: Calculate data serialization, round-trip propagation and ACK serialization separately.

    Check your work: The three components sum to the start-to-complete-ACK cycle, with no unstated processing term.

  4. Report payload throughput

    Selected reading: Same selections.

    Make this: Divide delivered payload bits by the cycle and explain why raw rate differs.

    Check your work: Use bits/s and label the result as ideal stop-and-wait payload throughput, not generic Internet speed.

  5. Change the feedback window

    Selected reading: Dordal 8.2–8.3 and Kurose/Ross 3.4.

    Make this: Find the smallest ideal outstanding-frame window that removes forward-link idle time.

    Check your work: Show the ceiling calculation and still subtract header overhead when reporting useful throughput.

  6. Test what the model excludes

    Selected reading: Selected text’s delay/transport discussion.

    Make this: Change one quantity and predict which timeline segments change; list evidence needed for a real trace.

    Check your work: An explanatory model identifies assumptions. Do not treat omitted queues, loss or congestion as known to be zero outside this exercise.

Worked exercise

Original link model: a fast line can spend time waiting

A fictional full-duplex link has 2,000,000 bit/s in each direction and 12 ms one-way propagation. Each data frame carries 1200 payload bytes plus 80 header bytes. The receiver sends one 40-byte ACK immediately after the complete frame arrives. A sender permits one outstanding frame. Find the complete cycle and useful payload throughput.

Assumptions

No loss, queues or processing delay; an unlimited backlog; constant rates; no other framing overhead. The next frame starts only after the entire ACK returns. One byte is 8 bits and Mb/s is decimal. These are stipulated teaching values, not a real protocol trace.

  1. Data serialization is (1200+80)×8/2,000,000=0.00512 s=5.12 ms. ACK serialization is 40×8/2,000,000=0.16 ms.
  2. The first frame finishes arriving after 5.12+12=17.12 ms. Its ACK finishes returning 12+0.16 ms later, so the full cycle is 5.12+24+0.16=29.28 ms.
  3. Each cycle delivers 1200×8=9600 payload bits. Throughput is 9600/0.02928≈327,869 bit/s, or 0.327869 Mb/s, about 16.39% of the raw 2 Mb/s.
  4. The sender’s forward serialization occupies only 5.12 ms of the cycle. Propagation is not extra payload work, and the ACK has nonzero size even with zero processing.

Answer and interpretation

The ideal one-frame cycle is 29.28 ms and useful throughput about 0.328 Mb/s. A high link bit rate does not eliminate feedback waiting or framing overhead.

Test the result

Set propagation to zero and recover a 5.28 ms cycle, still not exactly the raw payload capacity because the ACK and data header remain. Every addition in the time sum must have units of time.

Change an assumption

Follow-up: enough in-flight frames removes only one bottleneck

Keep the same ideal link and frame sizes, but allow up to W unacknowledged data frames. A backlogged sender transmits whenever the forward link and window permit; each received frame gets an immediate ACK. Find the smallest integer W for continuous forward transmission and its payload-rate ceiling.

  1. A complete ACK for the first frame returns 29.28 ms after that frame starts. W frames keep the forward link busy for W×5.12 ms, so W≥ceil(29.28/5.12)=6.
  2. At W=6, the sixth frame runs from 25.60 to 30.72 ms; the first ACK is already back at 29.28 ms, so a seventh can start at 30.72 ms with no forward idle gap. ACKs are spaced 5.12 ms apart and take 0.16 ms each, so the reverse link is not the bottleneck here.
  3. Continuous transmission still includes 80 header bytes per 1280-byte frame. Useful rate is 2×1200/1280=1.875 Mb/s, not 2 Mb/s. With W=5, each five-frame cycle has 3.68 ms idle and the model gives about 1.63934 Mb/s.

Answer and interpretation

The minimum window is 6 frames and the ideal payload ceiling 1.875 Mb/s. Pipelining removes this feedback idle time; it does not remove header bytes or establish real TCP throughput.

Test the result

Check the first seven frame starts explicitly and state the no-loss/no-queue assumptions again. If ACK processing, loss or a second bottleneck is introduced, recompute rather than applying the ceiling as a guarantee.

Common mistakes

Payload and frame mixed

Adding header bytes twice or treating payload size as full frame length changes both timing and rate accounting.

One-way delay called RTT

Round-trip propagation is twice the supplied 12 ms. RTT can include other terms in a measured system; define it rather than guessing.

A bound called a measurement

The ideal model can explain a mechanism without demonstrating deployed protocol behavior or an achievable service rate.

Before moving on

  • I can draw the ordered events and label units.
  • I can recompute cycle and payload rate after changing a quantity.
  • I can explain which assumptions make the window ceiling valid.

Sources and edition checks

Sources checked:

These are proposed study activities, not official course assignments or a claim that every book has been read cover to cover.

How we curate →