I85PREP
⌘ K
NIELIT ACC study materials
NIELIT ACC study resources

A district transport office: re-segmenting a flat network before voice and video arrive

NIELIT ACC · Put your knowledge to work in context.

All case studies

A district transport office: re-segmenting a flat network before voice and video arrive

A district transport office has 58 desktop users on two floors. The LAN grew organically: three 24-port unmanaged switches are daisy-chained from a single router, every device shares 192.168.1.0/24, and DHCP is served by the router from a 60-address pool. Twelve network printers have hand-set static addresses. Three consumer Wi-Fi routers were plugged into wall ports and still have DHCP enabled. The office has a 100 Mbps fibre leased line. The department now wants 40 PoE IP phones and a daily 90-minute video conference with state headquarters. Since the phones arrived, complaints have risen: address conflicts every morning, video freezing whenever a large file is copied to the server, choppy phone audio, and the whole network stalling for a minute or two at random. A vendor proposes simply buying a 48-port gigabit switch and, in their words, turning on VLANs. There is no network diagram, no managed switch, and no further budget until the next financial year.

1. Map each symptom to its most probable cause and list the specific checks you would run to confirm them before any money is spent.

Read the worked answer

Each symptom belongs to a different layer, so separate them first.

  • Morning address conflicts → more than one DHCP server. The three consumer routers hand out 192.168.1.x with themselves as gateway. Confirm with ipconfig /all on a conflicted PC (look at the DHCP server address), arp -a for duplicate MACs, and a capture filtering DHCP on the router's LAN port.
  • Video freezing during file copies → uplink oversubscription and no QoS. Every inter-switch byte and every call share one 100 Mbps daisy-chain. Confirm by reading interface utilisation on the router LAN port and running iperf between the two floors at peak time; a sustained 90+ Mbps with a call running is the proof.
  • Choppy phone audio → no segmentation and no priority. On unmanaged switches the phone's RTP competes with broadcasts and SMB traffic, and DSCP EF marking is not honoured. Measure broadcast rate and check whether the phones are even on a separate subnet (they are not).
  • Random network-wide stalls → broadcast domain of 58 PCs + printers + phones + three rogue APs, plus possible duplex mismatch on any port set manually to 100/full. Check late collisions and FCS errors per port.

What I would not do: add a fourth unmanaged switch, or renumber hosts before the rogue DHCP servers are removed.

2. Propose a target design - VLAN and IP plan, switching, DHCP, QoS, routing - and justify each choice against the office's constraints.

Read the worked answer

VLAN and IP plan (private ranges, chosen not to overlap the HQ VPN pool):

  • VLAN 10 Data - 10.20.10.0/26, 62 hosts: 58 PCs, printers, file server; gateway .1 on the L3 switch.
  • VLAN 20 Voice - 10.20.20.0/26, 40 phones; DHCP option 150 points to the phone TFTP server; access ports configured voice VLAN tagged, PC untagged.
  • VLAN 30 Guest Wi-Fi - 10.20.30.0/27; internet-only ACL, client isolation on.
  • VLAN 40 Management - 10.20.40.0/28 for switch, AP and UPS management, reachable only from an admin host.

Switching: two managed gigabit L2/L3 switches with 802.1Q trunks and at least two aggregated 1 Gbps uplinks (10 G fibre if the cable run permits), PoE on access ports. 48 ports is not enough: roughly 100 access ports are needed.

Routing and policy: inter-VLAN routing on the L3 switch; a firewall between VLAN 10/20 and the WAN for stateful inspection. Trust DSCP from phones and give EF a priority queue on trunks; cap the guest VLAN at about 20% of WAN bandwidth.

DHCP: moved onto the L3 switch with per-VLAN pools, 8-day leases, exclusions for printers and servers; DHCP disabled on the router and all APs.

Why: one /24 carrying voice and data cannot be given QoS or ACLs, and 40 extra endpoints make the broadcast and conflict problem worse. Segmentation is what makes the 90-minute conference predictable.

3. The vendor's 48-port switch has a 370 W PoE budget. Evaluate whether it is adequate for the 40 IP phones, show the calculation, and state what you would specify instead.

Read the worked answer

Calculation: 802.3af Class 4 reserves up to 15.4 W per powered port at the switch. 40 × 15.4 W = 616 W of standards-based reservation against a 370 W budget. That guarantees only about 24 Class 4 ports.

Why it may still appear to work - and why that is the trap: most basic IP phones draw 4–7 W, so 40 phones might total 200–280 W on day one. But the budget is exceeded on paper, so the switch can refuse power or shed ports as phones are added, and there is zero headroom for two access points, a door controller or a future phone with a colour display.

Port count: 48 ports cannot host 58 PCs + 40 phones + APs + uplinks; about 100 access ports are required.

Specify: two 48-port managed switches with 802.3at PoE+ budgets of 740 W each (one per floor, or one plus a second switch for redundancy), both on UPS so phones survive a mains failure. Confirm Cat5e/Cat6 cable with all four pairs terminated - patch leads wired to two pairs cannot carry PoE safely, and that is a common cause of port shutdowns after a PoE rollout.