Question 1.1 — connecting New York, London, Tokyo and Sydney
A multinational corporation needs to connect its headquarters with
branch offices across continents, with high-speed transfer,
reliability, and secure links for financial data.
Answer: a mesh topology, carried over a WAN. For
four sites a full mesh is practical; larger deployments normally use
a partial mesh between the busiest sites.
Mesh is the right fit because each requirement maps onto one of its
properties:
-
Dedicated links. Every pair of offices gets its
own point-to-point link, so the full capacity of that link carries
only their traffic. No office competes for a shared line, which is
what keeps transfers fast.
-
Reliability. If the New York to Tokyo link fails,
traffic still reaches Tokyo through London or Sydney. One broken
link never isolates a site.
-
Security and privacy. A message travels on the
dedicated link between the two offices concerned, so no other site
sees it in transit.
-
Fault isolation. A failing link is easy to
identify and route around while it is repaired.
The cost is cabling: a full mesh of n sites needs
n(n − 1)/2 links, so four offices need
4 × 3 / 2 = 6 links, and each site needs
n − 1 = 3 interfaces. That is why real deployments
lease WAN circuits or build a partial mesh rather than cabling a
full mesh worldwide.
Question 2.1 — a 20-person single-floor office
All computers connect to one central device that manages traffic and
lets staff share files and printers. It should be easy to manage and
to expand.
Answer: a star topology.
The scenario names the deciding clue: a central device every
computer connects to. Advantages:
-
Cheap to cable. Each device needs one link and
one port, so a star needs
n links where a mesh of 20
devices would need 190.
-
Easy to expand. Adding an employee means running
one cable to a free port, with no change to the other stations.
-
Easy to manage and troubleshoot. Traffic passes
through the hub or switch, so monitoring and fault detection
happen in one place.
-
Robust against link failure. A broken cable takes
down only the station on the end of it; the rest of the network
keeps running.
The trade-off to state in an exam: the central device is a single
point of failure. If the hub or switch dies, the whole network
stops.
Question 3.1 — five research computers working together
Five powerful computers should communicate directly with each other,
without relying on a central server, for maximum reliability and
performance.
Answer: a peer-to-peer model, not a client/server
model. This is a question about the network model, not the physical
topology.
Key characteristics:
-
Each machine acts as both client and server, requesting and
providing services.
-
There is no central server, so there is no single point of
failure. Losing one machine leaves the other four working.
-
Work and resources are distributed across the peers, which suits a
shared computational task.
-
It is cheaper to build, since no dedicated server hardware is
required.
-
The weakness is administration: security, backups and access
control have to be handled on every machine, so the model does not
scale to large organisations.