COMP90007 Internet Technologies
COMP90007 Overview
- Graduate coursework subject
- Semester Two offering
- University of Melbourne
- Layered network systems
Internet Technologies studies why a network of independently engineered machines can still exchange data predictably. The subject begins with a disciplined vocabulary: clients request services, servers respond, packets carry messages, and protocols coordinate peer entities. A layered architecture then separates concerns.
- Architecture first Separate service, interface and protocol before tracing a packet across layers.
- Calculate with units Convert message length, rate, distance and propagation speed before comparing delays.
- Reliability is layered Distinguish detecting corrupt bits from recovering lost or damaged frames.
- Pass thresholds Plan separately for the homework aggregate, project aggregate, written examination and overall result.
How COMP90007 is assessed
| Component | Weight | Format |
|---|---|---|
| Two equally-weighted homework assignments · hurdle | 10% | Individual; the two homework assignments are weighted equally |
| Networking Project 1 | 15% | Individual networking project; the published threshold applies to the two-project aggregate |
| Networking Project 2 | 15% | Group or individual networking project; the published threshold applies to the two-project aggregate |
| Written examination · hurdle | 60% | End-of-semester written examination |
The hurdle applies to the homework aggregate, the two-project aggregate, the written examination and the overall result. The red task badges indicate that each task contributes to one of those published thresholds; the two project rows share one aggregate threshold.
Assessment structure
The subject publishes separate pass thresholds for the homework aggregate, project aggregate, written examination and overall result.
What COMP90007 covers
The guide begins with the architectural language used throughout the subject, then follows data through the physical and data-link layers before comparing the access rules that make a shared channel usable.
Network Architecture and Layered Services
Requests and replies, broadcast and point-to-point links, topology, protocol hierarchies, OSI, TCP/IP and the subject's hybrid model02Signals, Media and Channel Capacity
Transmission and propagation delay, attenuation, wired and wireless media, digital modulation, Nyquist, Shannon and multiplexing03Framing, Errors and Reliable Delivery
Frame boundaries, parity and checksums, CRC, Hamming distance, acknowledgements, timeouts, Stop-and-Wait and sliding windows04Shared Channels, Wireless and Ethernet
Static and dynamic allocation, ALOHA and CSMA, collision-free access, hidden terminals, MACA, Ethernet framing and switchingEach layer offers a service upward through an interface while using a protocol to communicate with its peer. That separation lets a browser, transport protocol, network protocol and link technology change at different rates without collapsing the whole system into one design problem. This guide follows the lower-layer sequence represented in the available Semester 2 teaching materials.
It first distinguishes broadcast, point-to-point and multicast transmission and relates topology to contention. It then compares the OSI and TCP/IP reference models with the hybrid model used in the subject. At the physical layer, bandwidth is a rate while propagation delay depends on distance and signal speed. Transmission delay depends on message length and rate, so a faster link cannot remove a long path's propagation time.
Nyquist sets a noiseless signalling ceiling; Shannon sets a noisy-channel ceiling. Both are upper bounds, not promises of achieved throughput. The data-link layer turns an unreliable bit stream into framed transfers. Character counts, byte stuffing and bit stuffing solve different boundary problems.
Parity, an Internet checksum and CRC detect altered data with different structures; Hamming distance expresses the error patterns a code can detect or correct. Acknowledgements, timeouts and sequence numbers address loss and duplication, while sliding windows trade receiver memory for better use of a long-delay link.
The shared-medium material then compares fixed allocation, contention, collision-free reservation and limited contention. Wireless links add hidden and exposed terminals, so sensing near the sender is not enough. Ethernet provides a concrete integration point: its frame, CRC, addressing, collision logic and switched form connect the chapter ideas.
Assessment consists of two equally weighted individual homework assignments worth 10% together, two networking projects worth 15% each, and an end-of-semester written examination worth 60%. The published hurdle section requires at least 50% overall and at least 50% separately in the homework aggregate, project aggregate and written examination.
Treat those as four independent checks; a strong weighted total does not replace a missed component threshold. Exact submission dates, examination arrangements and any operational changes remain controlled by Canvas and the current timetable. This graduate coursework subject carries 12.5 credit points.
Separate transmission delay from propagation delay
- 1Put the rate in bits per second: 24 Mbit/s = 24,000,000 bit/s.
- 1Transmission delay is frame length divided by rate: 12,000 / 24,000,000 = 0.0005 s = 0.5 ms.
- 1Propagation delay is distance divided by signal speed: 2,000 / 200,000 = 0.01 s = 10 ms.
- 1One-way link latency is 0.5 ms + 10 ms = 10.5 ms, before processing or queueing delay.
- 1Interpret the result: distance dominates this example, so increasing rate changes only the smaller transmission term.
Key terms
- Protocol
- An agreement between communicating peer entities about how communication proceeds, including message meaning, ordering and expected responses.
- Service
- A capability a layer provides to the layer above it; the service describes what is available without prescribing the peer protocol that implements it.
- Interface
- The boundary and operations through which one layer accesses the service of the layer below.
- Bandwidth
- In the link model, the transmission rate measured in bits per second; it controls how long the sender takes to place a frame on the link.
- Propagation delay
- The travel time of a signal across the medium, calculated from path length divided by propagation speed.
- Hamming distance
- The smallest number of bit flips that changes one valid codeword into another; it determines guaranteed detection and correction capability.
- Cyclic redundancy check
- An error-detection method that treats a bit string as a polynomial and appends a remainder calculated by modulo-two division.
- Sliding window
- A reliability and flow-control scheme in which a sender may keep several numbered frames outstanding before their acknowledgements arrive.
- Carrier sense
- Checking whether a shared channel appears active before beginning a transmission.
- Hidden terminal
- A sender that cannot sense another sender but can still collide with it at their shared receiver.
- Collision domain
- The set of interfaces whose simultaneous transmissions can interfere; a switch isolates ports into separate domains.
COMP90007 FAQ
Why does layering matter in a real network?
Layering localises design responsibilities. An application can request reliable transport without knowing whether the next link is fibre or wireless, while a link protocol can frame data without interpreting the application's meaning. The price is overhead and the possibility that a clean boundary hides useful cross-layer information.
Which delay changes when bandwidth increases?
A higher bit rate reduces transmission delay because the same frame is placed on the link faster. It does not directly reduce propagation delay, which depends on path length and signal speed. Queueing and processing are additional delays and need separate inputs before they can be calculated.
When should Nyquist and Shannon be used together?
Use Nyquist to limit a noiseless channel for a stated bandwidth and number of signal levels. Use Shannon to limit a noisy channel for bandwidth and signal-to-noise ratio. When both descriptions apply, the achievable rate cannot exceed the tighter ceiling; neither equation guarantees a practical implementation.
What is the difference between error detection and correction?
Detection establishes that a received word is inconsistent with the code, after which retransmission may be requested. Correction uses enough redundancy to identify the most plausible valid word without retransmission. Hamming distance shows why correction needs more separation between codewords than detection.
Why can carrier sense still fail on wireless links?
A sender observes activity at its own location, but interference matters at the receiver. Two hidden senders may each hear an idle channel yet collide at a receiver. Conversely, exposed senders may hear one another even though transmissions to different receivers could safely proceed concurrently.
What must be passed separately in this subject?
The published syllabus requires at least half of the available result overall and at least half in each of three component groupings: the two-homework aggregate, the two-project aggregate and the end-of-semester written examination. Check current operational details and results treatment in Canvas.
How to study for the exam
Draw the layer boundary before solving a network question: name the object at that layer, the service requested upward and the peer protocol involved. For calculations, write a unit line first, then separate transmission, propagation and any round-trip terms before substitution.
For reliability, classify the failure as corruption, loss, duplication or receiver overload; that choice determines whether the relevant mechanism is a code, acknowledgement, timeout, sequence number or window. For shared access, state load, synchronisation and carrier-sense assumptions before comparing protocols. Keep a correction log that records the first mistaken distinction rather than only the final wrong number.
Because the syllabus publishes separate component thresholds, schedule revision and assessment work against every threshold rather than relying on the weighted total alone.
Your AI Computer Science tutor for COMP90007
Stuck on a hard COMP90007 question? Sia is AskSia’s AI Computer Science tutor — ask any COMP90007 Internet Technologies question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.