Layered Architecture: 22-Elec-B4 Information Technology Networks, December 2016, Q4

Question 4: Layered Architecture (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 advantage of layering (5 marks). Layering decomposes an intractable problem — move arbitrary data between arbitrary machines over arbitrary media — into a stack of small problems, each solved by a layer that offers a defined service to the layer above and uses only the service of the layer below. The payoff is modularity: because a layer is specified by its service interface rather than by its internals, any implementation that honours the interface may be substituted without touching anything else. Ethernet can be replaced by Wi-Fi under an unchanged IP layer; TCP can be replaced by UDP under an unchanged application; IPv4 can become IPv6 without rewriting the web.

Three further advantages follow from the same property and are worth naming explicitly. Layering permits independent standardisation and competition, since different bodies and different vendors can own different layers — IEEE the link layer, the IETF the network and transport layers — and their products still interoperate. It permits reuse, because one implementation of a layer serves every protocol above it rather than each application re-solving reliability or addressing for itself. And it makes systems tractable to reason about, test and fault-find: a defect can be localised by asking which layer's service is not being delivered, which is exactly how network troubleshooting proceeds in practice. The cost is a real one and worth acknowledging — strict layering adds per-layer headers and can duplicate work, for example error detection performed at both the link and the transport layer — but the engineering benefit of being able to change one layer without changing the others has proved decisive.

Part (b) — the seven OSI layers, one sentence each (10 marks).

Physical. Transmits raw bits over the medium, defining voltages, modulation, timing, connectors and cabling.
Data link. Groups bits into frames between two directly connected nodes, providing local addressing, medium access and error detection over that single link.
Network. Delivers packets from source host to destination host across multiple links, providing global addressing, routing and forwarding.
Transport. Provides end-to-end delivery between processes on the two hosts, with multiplexing by port and optionally reliability, sequencing and flow and congestion control.
Session. Establishes, manages, synchronises and terminates the dialogue between the two applications, including checkpointing and resumption of long exchanges.
Presentation. Handles the syntax and semantics of the transferred data — character and data representation, compression and encryption — so that machines of different internal formats can interoperate.
Application. Provides the network service the user actually invokes, such as mail transfer, file transfer, web access or name resolution.

The OSI seven-layer reference model with a representative protocol or function at each layer. Layers 1 to 3 act hop by hop along the path; layers 4 to 7 exist only in the two end systems.

Part (c) — locating eight items in the model (10 marks).

i. Encryption and decryption — the presentation layer (layer 6). In the OSI model, transforming the representation of data, including ciphering it, is precisely the presentation layer's job. In the real TCP/IP stack the equivalent function, TLS, sits between the transport and application layers, which is usually mapped onto layers 5 and 6; link-layer ciphering such as IEEE 802.11 CCMP is a separate, layer-2 use of encryption.
ii. The TCP/IP protocol — two layers. TCP is a transport-layer (layer 4) protocol and IP is a network-layer (layer 3) protocol; the name “TCP/IP” is a pairing of the two, not a single protocol at a single layer.
iii. Ethernet — the data link layer (layer 2), and the physical layer (layer 1). IEEE 802.3 specifies the frame format, MAC addressing and medium access at layer 2, and also the signalling, encoding and cabling at layer 1, so a complete answer names both.
iv. Routing — the network layer (layer 3). Choosing an end-to-end path across intermediate nodes and forwarding packets along it is the defining function of layer 3.
v. SMTP, HTTP or FTP — the application layer (layer 7). Each is a protocol between user-facing programs and appears nowhere in the intermediate nodes.
vi. The UDP protocol — the transport layer (layer 4). UDP multiplexes by port between processes on the two end hosts, which is a layer-4 service, even though it declines to offer the reliability that TCP adds.
vii. Character display, such as ASCII — the presentation layer (layer 6). Agreeing the coded representation of characters so that both machines interpret the same octets identically is a data-representation concern.
viii. Error detection over a single link — the data link layer (layer 2). The frame check sequence, a cyclic redundancy check appended to each frame, detects corruption over one hop. (Error detection also recurs end to end at layer 4 in the TCP and UDP checksums, but the phrase “over a single link” pins this item to layer 2.)

Check: the printed paper omits the item marker before “Ethernet” in part (c), so the printed list reads i, ii, [blank], iv, v, vi, vii, viii. The missing marker is plainly iii and the list is answered as eight items i–viii on that reading, which is confirmed by the surviving markers being consecutive either side of the gap.

Topic. Protocol layering: the OSI reference model, the service-interface abstraction that makes layering worth its overhead, and how real protocols map onto it.

Key relations. Layer n in one host communicates with layer n in the peer host by a protocol; it communicates with layers n − 1 and n + 1 in its own host by a service interface. Data descending the stack is encapsulated, each layer prepending its header to the unit above, so the layer-2 frame carries a layer-3 packet carrying a layer-4 segment. Layers 1 to 3 are present in intermediate nodes (repeaters at 1, switches at 2, routers at 3); layers 4 to 7 are end-to-end and exist only in the hosts.

Why this works. The service interface is the whole mechanism. Once a layer's obligations are stated as a service — “deliver this frame to the node at the other end of this link, and tell me if it was corrupted” — the layer above can be written against that promise and nothing else, which is what allows Wi-Fi to be substituted for Ethernet, or fibre for copper, with no change whatever above layer 2. The corresponding rule for locating a function in the model is to ask over what span it acts: single link means layer 2, host to host across a path means layer 3, process to process end to end means layer 4, and anything concerning the meaning or representation of the user's data belongs at 6 or 7. This span test answers every item of part (c) without memorisation, and it is why error detection legitimately appears at two different layers — the two instances protect different spans.

Common pitfall. Answering “TCP/IP” with a single layer. It spans two, and the marker is testing precisely that the candidate separates them. The second pitfall is placing Ethernet only at layer 2 when IEEE 802.3 defines the physical layer as well, and the third is describing the OSI layers at length after the question has explicitly warned that marks may be deducted for unnecessary detail — one crisp sentence per layer is the whole requirement.

Source. Tanenbaum and Wetherall, Computer Networks, 5th ed., §1.4 (reference models, OSI and TCP/IP compared); Leon-Garcia and Widjaja, Communication Networks, 2nd ed., ch. 2; ISO/IEC 7498-1, the OSI basic reference model, for the normative layer definitions.

Question 4 of 5: Layered Architecture

Question 4: Layered Architecture (25 marks)