22-Elec-B4 Information Technology Networks · May 2018

Question 5 of 5: Layered Architecture — The OSI Reference Model

Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)

Notes on this paper

Paper format. Professional Engineers of Ontario — National Examinations, May 2018, 16-Elec-B4 Information Technology Networks. Three hours, closed book; one Casio or Sharp approved calculator permitted. The paper prints five questions of 25 marks each and any four constitute a complete paper worth 100 marks, with the marks for every sub-part shown in the left margin. All five questions are solved here, because this set is a study resource rather than an exam attempt, and because a candidate choosing which four to answer benefits from seeing the fifth worked out.

Reference texts. A. Leon-Garcia and I. Widjaja, Communication Networks: Fundamental Concepts and Key Architectures, 2nd ed. (the syllabus reference for this code); J. F. Kurose and K. W. Ross, Computer Networking: A Top-Down Approach, 8th ed.; A. S. Tanenbaum and D. J. Wetherall, Computer Networks, 5th ed.; W. Stallings, Wireless Communications and Networks, 2nd ed.; T. S. Rappaport, Wireless Communications: Principles and Practice, 2nd ed.; S. Sesia, I. Toufik and M. Baker, LTE — The UMTS Long Term Evolution, 2nd ed.

Source reading — Question 3 figure. The printed network labels two different nodes with the letter F: one on the upper row between D and the right-hand vertex, and one at the far right. This is a typographical slip in the examination paper. To keep the working unambiguous the far-right node is written F′ throughout; every distance and path below is unaffected by the naming, and a candidate should simply state the convention adopted, exactly as the paper's own instruction on assumptions invites.

Source reading — Question 2(d). The printed text says “Repeat part b”, but part (b) is the qualitative question about congestion in wired networks and carries no window to repeat. The intended reference is part (c), whose window evolution is the thing a lost packet perturbs. Part (d) is answered on that reading, and the reading is stated in the answer rather than assumed silently.

Question 5: Layered Architecture — The OSI Reference Model (25 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

Part (a) — The seven layers, one sentence each (10 marks)

Figure 5.1 — The OSI reference model. Layers 1 to 4 move bits reliably from one end system to the other; layers 5 to 7 give that bit stream meaning for the application.

The question explicitly rewards brevity, so one sentence each is what follows, from the bottom up.

Physical. Transmits raw bits over the medium, defining the connectors, voltages or optical levels, line coding, bit timing and data rate.
Data link. Groups bits into frames and delivers them across a single link, providing framing, physical (MAC) addressing, error detection and medium-access arbitration between stations sharing the link.
Network. Delivers packets between end systems that may be many links apart, providing global addressing, routing and forwarding, and fragmentation where link sizes differ.
Transport. Provides end-to-end delivery between processes on the two end systems, with multiplexing by port, and optionally with reliability, in-order delivery, flow control and congestion control.
Session. Establishes, manages and terminates dialogues between applications, adding dialogue control, token management and checkpointing so that a long transfer can resume after a failure.
Presentation. Gives the transferred data a common syntax, handling character-set and data-representation conversion, compression, and encryption.
Application. Provides the network services the user's program actually calls — file transfer, mail, name resolution, web access — and defines the protocols that implement them.

Part (b) — Locating each item in the model (10 marks)

Mapping each item to its OSI layer
Item	Layer	Why
i. End-to-end error correction and in-order packet delivery	Transport (4)	These are end-to-end guarantees between processes; the data-link layer offers the same services but only across one link.
ii. The SMTP protocol	Application (7)	Simple Mail Transfer Protocol is a user-level service protocol, carried over TCP.
iii. TCP/IP	Transport (4) and Network (3)	The name is a pair: TCP is the transport protocol, IP the network protocol. Naming both is what earns the mark.
iv. Ethernet	Data link (2), with a physical (1) component	IEEE 802.3 specifies MAC framing and addressing at layer 2, and also its own physical-layer signalling, so the honest answer names layer 2 and notes layer 1.
v. Signal voltages	Physical (1)	Electrical levels and their timing are the definition of the physical layer.
vi. Encryption and decryption	Presentation (6)	In the OSI model, transformation of the data's representation, including ciphering, belongs to layer 6. (In practice TLS sits between layers 4 and 7, and link encryption exists at layer 2 — but the model's answer is 6.)
vii. Character display, such as ASCII	Presentation (6)	Character-set and data-representation conversion is precisely the presentation layer's function.
viii. The HTTP protocol	Application (7)	The Hypertext Transfer Protocol is a user-level service protocol, carried over TCP.

Two of these items deserve a sentence of justification rather than a bare label, because a marker is looking for the reasoning. Item (i) distinguishes the transport layer from the data-link layer, since both perform error control and ordering — the discriminator is the words end-to-end, which restrict the answer to layer 4. Item (iii) is a deliberate trap: “TCP/IP” is a compound name and an answer of “transport” alone is incomplete.

Part (c) — Why complex networks are built in layers (5 marks)

Layering is used because it is the only way to build a system of the internet's complexity that can be specified, implemented and evolved by parties who do not know one another. Each layer is defined by the service it offers to the layer above and the protocol it runs with its peer at the same level, and those two definitions are all a designer needs. The internal mechanism is hidden, so the implementation of one layer may be replaced entirely without disturbing anything above or below it: IP runs unchanged over Ethernet, over Wi-Fi, over fibre and over links that had not been invented when it was designed, and a web browser is indifferent to which is beneath it.

Four consequences follow, and they are what the marks are for. First, modularity and manageable complexity: a hard problem is decomposed into layers each of which can be reasoned about and tested in isolation, which also localises faults — a diagnosis proceeds layer by layer, from cable to routing to socket. Second, interoperability through standardisation: because the interfaces are public, equipment from different vendors interworks, and competition happens within a layer rather than over the whole stack. Third, independent evolution: new physical technologies are adopted without changing applications, and new applications are deployed without touching routers — the “narrow waist” of a single network-layer protocol is exactly what let the internet absorb both the web and the mobile radio access network. Fourth, reuse: reliability, addressing and encryption are each implemented once and shared by every application above.

An answer that also names the cost is a stronger one. Strict layering imposes overhead — a header per layer and a copy per interface — and it hides information that a lower layer could have used well, which is precisely the difficulty Question 2(e) describes: the transport layer cannot see whether a loss came from a full queue or from a fade, because the link layer's internals are invisible to it. Real stacks therefore relax the model where performance demands it, through cross-layer signalling such as explicit congestion notification, and they collapse layers that carry little weight — the internet's five-layer model folds OSI's session and presentation functions into the application. The principle survives the compromises: layering is a discipline for managing complexity, not a law of nature.

Final results — Question 5
Part	Result
(a) The seven layers, bottom up	Physical, Data link, Network, Transport, Session, Presentation, Application
(b) i. End-to-end error correction, in-order delivery	Transport (4)
(b) ii. SMTP	Application (7)
(b) iii. TCP/IP	Transport (4) and Network (3)
(b) iv. Ethernet	Data link (2), with a physical (1) component
(b) v. Signal voltages	Physical (1)
(b) vi. Encryption and decryption	Presentation (6)
(b) vii. Character display, ASCII	Presentation (6)
(b) viii. HTTP	Application (7)
(c) Reasons for layering	Modularity, interoperability through public interfaces, independent evolution, and reuse — at the cost of header overhead and hidden information

Topic. Protocol layering, the OSI reference model and its relationship to the TCP/IP architecture.

Key relations. Service versus protocol: layer $n$ offers a service to layer $n+1$ through a service access point and realises it by a protocol spoken with its peer layer $n$; encapsulation, in which each layer's protocol data unit becomes the payload of the layer beneath, so a datagram carries a stack of headers. The correspondence to the internet's five-layer model is Application (OSI 5–7), Transport (4), Network (3), Link (2), Physical (1).

Why this works. Layering is information hiding applied to a distributed system. Because the only contract between adjacent layers is the service definition, the pair of implementations on either side of an interface can be written by different organisations and replaced independently, and because the only contract between peers is the protocol, two implementations of the same layer interoperate without sharing code. That is what allows the internet to be an assembly of independently administered networks rather than one engineered system. The architectural insight worth carrying away is the “narrow waist”: the model tolerates enormous variety at layers 1–2 and at layers 5–7 precisely because everything is required to meet at a single protocol, IP, at layer 3.

Common pitfall. Reciting the layers in the wrong order or from the wrong end — state explicitly whether the list runs upward from the physical layer. Assigning encryption to the application layer because TLS lives there in practice; the question asks about the OSI model, where it is layer 6, and the practical caveat belongs in a parenthesis rather than in place of the answer. Answering “transport” alone for TCP/IP. Placing Ethernet at layer 1 because it involves cables, or at layer 3 because it involves addresses; MAC addressing is layer 2. And, in part (c), listing benefits without saying what makes them possible — the service/protocol separation is the mechanism, and naming it is what distinguishes a full answer.

Source. Leon-Garcia & Widjaja, Communication Networks, 2nd ed., Ch. 2 (applications and layered architectures) and §2.4 (the OSI reference model); Tanenbaum & Wetherall, Computer Networks, 5th ed., §1.4 (reference models); Kurose & Ross, Computer Networking, 8th ed., §1.5 (protocol layers and their service models).

← Question 4 Back to the paper →