Network architecture: topologies, servers, models and segmentation – IB Computer Science exam questions with mark scheme

1Standard4 marks

A school's network connects 600 student devices, 80 staff computers and 40 CCTV cameras as one single, flat LAN. The technician proposes dividing it into three segments (separate subnets or VLANs) for students, staff and CCTV.

Explain how this segmentation would improve:

(i) the network's performance [2]

(ii) the network's security. [2]

(i) Broadcast traffic is confined to each segment / each segment is a smaller broadcast domain(1)
(i) so there is less unnecessary traffic/congestion on each segment and more bandwidth available to each group (e.g. broadcasts from the 600 student devices no longer reach the staff computers or cameras; allow: CCTV traffic can be kept separate/prioritised)(1)
(ii) Traffic between segments must pass through a router/firewall/layer-3 switch where rules can restrict access (e.g. students cannot reach staff computers or the CCTV system)(1)
(ii) A breach or malware on one segment is contained / cannot easily spread to the others (allow: easier to monitor each segment)(1)
Answer: Smaller broadcast domains reduce congestion; traffic between segments can be filtered, so access is restricted and attacks are contained.
Examiner tip: Give a mechanism and its effect for each part, and apply it to the scenario (students, staff, CCTV); 'it makes it faster and safer' earns nothing.

2Standard4 marks

Compare and contrast the client–server model with the peer-to-peer (P2P) model of networking. [4]

Similarity: both models let devices share resources/data/services over a network (allow: both can operate over a LAN or the internet; allow: in both, a device can request a service from another device)(1)
Difference in roles: in client–server, dedicated central server(s) provide services and clients only request them, whereas in P2P every peer can act as both client and server, sharing resources directly(1)
Difference in management: client–server allows central administration, security, backup and updates, whereas in P2P each peer manages its own resources, so it is harder to secure/back up consistently(1)
Difference in resilience/cost: a client–server network depends on the server (single point of failure/bottleneck) and servers are costly to buy and run, whereas P2P has no single point of failure and is cheaper/scales as peers join (accept either half if clearly contrasted with the other model)(1)
Answer: Similarity: both share resources over a network. Differences: central server serving clients vs peers acting as both; central management/security vs each peer managing itself; single point of failure and higher cost vs no single point of failure and lower cost.
Examiner tip: 'Compare and contrast' needs at least one similarity as well as differences, and each difference should say what happens in both models (use 'whereas'); describing only one model scores very little.

3StandardHL only4 marks

A user types www.example.org into a browser. Describe how the Domain Name System (DNS) is used to load the web page. [4]

The browser/OS sends a request containing the domain name to a DNS server (resolver) (allow: first checks its local DNS cache)(1)
The DNS server looks up (or queries other DNS servers for) the matching IP address(1)
The IP address is returned to the browser (and may be cached)(1)
The browser uses the IP address to send an HTTP(S) request to the web server(1)
Answer: Domain name sent to DNS server, IP address returned, browser contacts server at that IP.
Examiner tip: Say what is sent and what comes back: name goes out, IP address returns.

Stuck on Network architecture: topologies, servers, models and segmentation?

Get it sorted in a free 30-minute session

We work through your weakest questions on this topic on Zoom. I reply the same day.

Free 30-min session on this topic Send me 5 questions a day

4Core4 marks

A small college has three buildings, A, B and C. The diagram shows its network. Each building has one switch with the building's PCs connected to it, and the three switches are joined to one another by fibre links.

(a) Identify the topology used inside each building. [1]

(b) Identify the topology of the college network as a whole. [1]

(c) The fibre link between Switch A and Switch B is accidentally cut. Explain why PC A1 can still communicate with PC B2. [2]

(a) Star (accept: star with a switch at the centre)(1)
(b) Hybrid (accept: hybrid of star and mesh; star-mesh hybrid; do not accept 'mesh' or 'star' alone)(1)
(c) The switches are connected as a (full) mesh, so there is more than one path between Switch A and Switch B / a redundant link exists(1)
(c) Data from A1 can be sent via Switch C (A→C→B) to reach B2 (allow: traffic is re-routed via the other building)(1)
Answer: (a) Star (b) Hybrid (star + mesh) (c) The switches form a mesh, so frames travel A→C→B instead.
Examiner tip: 'Hybrid' needs to name both parts it combines if asked to describe it; in (c) say which alternative path the data actually takes, not just 'mesh is reliable'.

5Standard5 marks

A bank wants its six regional data centres to stay connected even if several links fail, so it plans to connect them in a full mesh topology using leased lines.

(a) Calculate the number of leased lines needed to connect the six data centres in a full mesh. Show your working. [2]

(b) A seventh data centre is later added. State how many extra leased lines are needed. [1]

(c) With reference to your answers, explain why a full mesh topology is rarely used to connect the hundreds of computers inside a single office. [2]

(a) Use of $\frac{n(n-1)}{2}$ with $n = 6$, or $5+4+3+2+1$ (allow $6\times5\div2$)(1)
(a) 15 (leased lines) (15 with no working scores 1 only, as working is required)(1)
(b) 6 (new centre must link to each of the 6 existing ones; allow 21 − 15 shown)(1)
(c) The number of links (and network ports/cabling) grows rapidly / roughly with the square of the number of devices (e.g. 200 PCs would need 19 900 links), so it does not scale(1)
(c) so the cost (of cabling, ports and installation) is far too high (allow: too complex to install/manage); a star with switches gives sufficient reliability far more cheaply (ignore 'too expensive' with no link to the number of links)(1)
Answer: (a) 6×5÷2 = 15 lines (b) 6 extra (c) Links grow ~n², e.g. 19 900 for 200 PCs, so cost/complexity do not scale.
Examiner tip: In (b) many students give 21: the question asks for the extra lines, so subtract or reason that the new site links once to each existing site.

6CoreHL only4 marks

A hospital's network contains several servers. Identify the type of server that performs each of the following functions.

(a) Automatically gives each tablet an IP address, subnet mask and default gateway when it joins the ward Wi-Fi. [1]

(b) Stores patients' scanned letters in shared folders, with access permissions set for each department. [1]

(c) Receives staff browser requests for external websites, blocks banned sites and keeps copies of frequently visited pages so they load faster. [1]

(d) Stores incoming messages for each member of staff until they are collected by their email client. [1]

(a) DHCP server (accept Dynamic Host Configuration Protocol server)(1)
(b) File server (accept NAS; do not accept 'database server' or 'web server')(1)
(c) Proxy server (accept web proxy / caching proxy; do not accept firewall alone)(1)
(d) Mail server (accept email server / IMAP or POP3 server; do not accept SMTP alone)(1)
Answer: (a) DHCP (b) File (c) Proxy (d) Mail
Examiner tip: Caching plus filtering of outgoing web requests is the signature of a proxy server; 'firewall' does not cache pages.

7HardHL only6 marks

A sixth-form college uses a single DHCP server to configure about 1500 laptops and phones. Addresses are leased for 24 hours.

(a) Explain why the college uses a DHCP server rather than giving each device a static IP address configured by hand. [3]

(b) At 08:00 one morning the DHCP server crashes. Students who were using their laptops on campus the previous afternoon can still use the network for a while, but visitors arriving at 09:00 cannot connect at all.

(i) Explain why the two groups are affected differently. [2]

(ii) Suggest one change that would improve the reliability of the DHCP service. [1]

(a) Devices are configured automatically as soon as they join, so technicians/visitors do not have to set up about 1500 devices manually (saves time; scales as devices are added)(1)
(a) The server allocates addresses from a central pool and records which are leased, so no two devices are given the same address and typing errors are avoided (no address conflicts)(1)
(a) Leases expire, so addresses of devices that leave (e.g. visitors) return to the pool and are reused, meaning fewer addresses are needed than the number of different devices (accept: settings such as gateway/DNS can be changed once centrally for all devices)(1)
(b)(i) Students' devices already hold a valid lease (still within 24 hours) so they keep their existing IP address/configuration until it expires(1)
(b)(i) Visitors' devices have no lease, and with no server to answer their requests they receive no IP address (allow: they self-assign an unusable link-local address) so cannot communicate(1)
(b)(ii) Add a second/backup (failover) DHCP server (accept: split the address pool across two servers; give critical devices static addresses; ignore 'restart the server' / 'longer leases' alone)(1)
Answer: (a) Automatic configuration of ~1500 devices; central pool prevents duplicate addresses; expired leases are reused (or settings changed centrally). (b)(i) Existing leases are still valid; new devices cannot get one. (ii) Add a failover DHCP server.
Examiner tip: The lease is the key idea: explain that already-configured devices keep their address until the lease runs out, rather than saying they are 'remembered by the network'.

8Hard6 marks

An engineering firm has been allocated the private network 10.20.30.0/24. The network manager divides it into four equal subnets, one each for Design, Finance, Sales and Guests, in that order starting from 10.20.30.0.

(a) State the prefix length (CIDR suffix) of each subnet. [1]

(b) Calculate the maximum number of hosts that can be given an address in each subnet. Show your working. [2]

(i) the network address [1]

(ii) the broadcast address. [1]

(d) A Finance PC has the address 10.20.30.70. Deduce, with a reason, whether a guest device with address 10.20.30.200 is on the same subnet. [1]

(a) /26 (accept subnet mask 255.255.255.192)(1)
(b) 32 − 26 = 6 host bits, so $2^6 = 64$ addresses per subnet (allow 256 ÷ 4 = 64)(1)
(b) 64 − 2 = 62 hosts (network and broadcast addresses cannot be assigned) (62 with no working scores 1 only; 64 alone scores M1 only)(1)
(c)(i) 10.20.30.128(1)
(c)(ii) 10.20.30.191 (follow through from an incorrect network address in (c)(i) if broadcast = network + 63)(1)
(d) No: Finance is 10.20.30.64–127 (/26) but .200 lies in the Guests subnet 10.20.30.192–255, so traffic between them must pass through a router (need correct conclusion with a correct range or reason)(1)
Answer: (a) /26 (b) 2⁶ − 2 = 62 (c)(i) 10.20.30.128 (ii) 10.20.30.191 (d) No: .70 is in .64–.127, .200 is in .192–.255.
Examiner tip: Remember to subtract 2 for the network and broadcast addresses, and count subnets from 0: Sales is the third block, starting at .128, not .192.

9Standard4 marks

A charity's office uses one managed 48-port switch. Staff PCs, a finance PC holding donor bank details, and a guest Wi-Fi access point are all plugged into it. The charity cannot afford a second switch.

(a) Explain how virtual LANs (VLANs) could be used to separate the finance PC and the guest Wi-Fi from the other staff PCs on this single switch. [3]

(b) Staff on the staff VLAN still need to print to a printer that is placed on the finance VLAN. State what is needed for this to work. [1]

(a) Switch ports are assigned (configured in software) to different VLANs, e.g. staff, finance and guest (accept: devices tagged with a VLAN ID)(1)
(a) Each VLAN behaves as a separate logical network/broadcast domain even though all devices share the same physical switch (no new hardware/rewiring needed)(1)
(a) The switch will not forward frames between VLANs, so guest devices cannot reach the finance PC (donor data protected) / guest traffic is kept apart from staff traffic(1)
(b) A router / layer-3 switch (inter-VLAN routing) with a rule/ACL allowing staff traffic to reach the printer (accept router alone; do not accept 'a hub' or 'another switch')(1)
Answer: (a) Assign ports to separate VLANs; each is its own logical network/broadcast domain on the same hardware; the switch does not forward traffic between them, so guests cannot reach finance. (b) Inter-VLAN routing via a router or layer-3 switch.
Examiner tip: Say that VLANs are logical, configured on the switch, and that traffic between them must be routed; 'VLANs make it secure' alone gains nothing.

10Stretch6 marks

A games company releases a 40 GB update for an online game. In the first hour, 50 000 players start downloading it. The company's download servers have a combined upload bandwidth of 10 Gbit/s. (Take 1 GB = 8 Gbit.)

(a) Calculate the minimum time, in days, for the servers to deliver the update to all 50 000 players using a client–server model. Give your answer to 3 significant figures. [2]

(b) The company is considering distributing future updates using a peer-to-peer (P2P) model instead. Discuss whether P2P would be a better model for distributing updates. [4]

(a) Total data $= 50\,000 \times 40 \times 8 = 1.6\times10^{7}$ Gbit and time $= 1.6\times10^{7} \div 10 = 1.6\times10^{6}$ s (seen or implied)(1)
(a) 18.5 days (accept 18.52; 444 hours with no conversion to days scores the first mark only)(1)
(b) For P2P: players who already have (parts of) the update upload them to other players, so total upload capacity grows with the number of players / the load on the company's servers falls (scalability)(1)
(b) For P2P: no single point of failure / lower server and bandwidth costs for the company (reliability or cost)(1)
(b) Against P2P: less control/security: files could be tampered with or carry malware, so each piece must be verified (e.g. hashes/digital signatures); players may not seed or may object to their upload bandwidth (or data caps) being used(1)
(b) A reasoned conclusion linked to the scenario, e.g. a hybrid in which central servers track/authenticate and sign the update while P2P spreads the load (accept a justified conclusion either way)(1)
Answer: (a) 50 000×320 Gbit ÷ 10 Gbit/s = 1.6×10⁶ s ≈ 18.5 days. (b) P2P scales and cuts server cost/single point of failure, but needs integrity checks and players' bandwidth; a hybrid is a sensible conclusion.
Examiner tip: 'Discuss' needs points on both sides and a conclusion; use your calculation to show that the client–server bottleneck is the servers' fixed upload bandwidth.

Stuck on this topic?

Book a free 30-minute session on WhatsApp and we'll find exactly which step is costing you the marks. Or build a timed paper on it.

Free 30-minute session