Junior network engineer interview questions, from subnetting on paper to a packet that never arrives
Junior network engineer interviews often include a subnetting question worked by hand, then follow one packet across a network: how the host decides where to send it, how the switch and the router handle it, and where you would look when it never arrives. Practise the arithmetic until it is quick, and answer every protocol question with what actually happens on the wire.

Expect a subnetting question you work out by hand, often early, because it tells the panel quickly whether you can do the arithmetic the job is built on. After that, most junior network interviews follow a single packet: how a laptop decides whether the destination is local, how it finds the next device's hardware address, what the switch does with the frame, how the router picks a path, and what you check when the packet never turns up. The questions below are grouped along that journey, with what a good answer contains, checked against the RFCs and vendor documentation linked at the end.
The National Careers Service job profile for a network engineer gives an average salary of £30,000 for a starter and £60,000 for an experienced engineer, and lists a Network Engineer Level 4 Higher Apprenticeship among the routes in, so you may be up against apprentices with real configuration experience behind them. You can close some of that gap with a lab you have actually broken and fixed (the home lab guide covers building one) and by talking through the questions here out loud before the day.
Work the subnet on paper first
"You have 192.168.123.0 with a mask of 255.255.255.192. How many subnets, and which addresses can hosts use?"
Show the working, not just the answer. A mask of 255.255.255.192 is 26 ones followed by six zeros, so two bits have been borrowed from the host part of a /24 and you get four networks: 192.168.123.0, .64, .128 and .192. Six host bits give 64 addresses in each, and the first and last of every subnet cannot go to a host, because a host part of all zeros names the network and all ones is the broadcast. That leaves 62 usable addresses each: .1 to .62, .65 to .126, .129 to .190 and .193 to .254. Microsoft's own TCP/IP addressing article uses this exact example, so it is worth working through until you can do it without a pen.
A common follow-up is whether two addresses are on the same subnet. With that /26 mask, 192.168.123.71 sits in the .64 network and 192.168.123.133 in the .128 network, so they need a router between them even though they would share a network under the old class C mask of 255.255.255.0.
"What are the private address ranges?"
RFC 1918 reserves 10.0.0.0 to 10.255.255.255 (10/8), 172.16.0.0 to 172.31.255.255 (172.16/12) and 192.168.0.0 to 192.168.255.255 (192.168/16) for private networks. It is easy to forget that the middle range runs to 172.31, not 172.16.255.255.
How a host decides where the packet goes
"A laptop wants to reach a server. What does it do before sending anything?"
It compares its own address and mask with the destination. If the destination falls inside its own subnet it sends the packet directly on the local network; if not, it sends it to its default gateway, the router configured on the laptop, and leaves the rest to that router. Saying this unprompted shows you understand why the mask and the gateway are both required settings and not decoration.
The follow-up is often a fault: what happens if the mask is wrong? Microsoft's article describes the classic symptom. A host still using the default class mask on a network that has been subnetted can talk to machines on its own segment and to distant networks, but not to nearby subnets that share its original class network, because it thinks they are local and never hands those packets to the gateway.
"What is ARP for?"
IP addresses are not what goes on an Ethernet cable; frames are delivered to 48-bit hardware addresses. The Address Resolution Protocol, defined in RFC 826, maps an IP address to that hardware address. The asking host broadcasts a request to every station on the segment, and the owner of the address replies directly. A good answer adds the consequence: for a destination on another subnet, the host resolves the gateway's address, not the server's, because the gateway is where the frame is going.
"Walk me through how a client gets an address from DHCP."
Four messages, often remembered as DORA. The client broadcasts a DHCPDISCOVER on its local subnet, servers answer with a DHCPOFFER that carries an available address, the client broadcasts a DHCPREQUEST for the offer it has chosen, and the server confirms with a DHCPACK. RFC 2131 describes each step. The detail that separates candidates is the relay: the discover is broadcast on the client's own subnet, so when the DHCP server is on another subnet a relay agent passes the messages between them. When users on a newly added VLAN get no address, the relay for that VLAN is one of the first things to check.
Switches and VLANs
"What is a VLAN, and why use one?"
A VLAN splits one switched network into separate broadcast domains. Juniper's documentation sums up the benefits: broadcasts stay inside the VLAN, devices can be grouped by function rather than by where they are plugged in, and each VLAN maps to a single IP subnet, which gives you a clean place to apply routing and security policy between them. If the panel pushes on numbers, Juniper's documentation gives the usable 802.1Q VLAN IDs as 1 to 4094, with 0 and 4095 reserved.
"What is the difference between an access port and a trunk port?"
An access port belongs to one VLAN and is where an end device plugs in; its traffic is normally untagged. A trunk carries several VLANs over one link, usually between switches, and marks each frame with an 802.1Q tag whose VLAN ID tells the far end which VLAN it belongs to. Expect the next question to be about the native VLAN, the one VLAN a trunk can be set to accept untagged frames for. Be ready to say that both ends of a trunk need to agree on it, and that you would check it on both switches when traffic from one VLAN turns up in another.
"Why does a switched network need spanning tree?"
Because redundant links between switches form loops. Spanning Tree Protocol, defined in IEEE 802.1D, builds a tree of links out from a root bridge and disables the links that would cause loops, leaving a single active link between any two devices. Juniper's documentation also mentions Rapid Spanning Tree (RSTP), notes that original STP converges more slowly, and says the two are compatible, so a network can run RSTP on some switches and STP on others. If you are asked how the root bridge is chosen, answer from your own vendor's documentation rather than guessing, and say that you would check which switch is root in a real network instead of assuming.
Routing beyond the default gateway
"A router has routes for 10.0.0.0/8 and 10.1.2.0/24. Where does it send a packet for 10.1.2.5?"
Down the /24. Forwarding on the internet works on a longest-match basis, as RFC 4632 puts it: when two routes overlap, the more specific one is used. It is a quick question that catches people who think the first route listed wins.
"What is the difference between a static route and a routing protocol such as OSPF?"
A static route is configured by hand and stays put until someone changes it. A routing protocol lets routers learn paths from one another and react when a link fails. OSPF is a link-state protocol: each router floods link-state advertisements describing its links, the routers in an area end up with identical databases of its topology, and each one runs the shortest-path-first (Dijkstra) calculation to find the least-cost route to every destination. Mention areas if you can: when an OSPF network is split into areas, each router keeps a detailed database of its own area and only summarised topologies of the others.
When the packet never arrives
"Explain the TCP three-way handshake."
The client sends a SYN with its initial sequence number, the server replies with its own SYN and an acknowledgement of the client's, and the client acknowledges the server's. RFC 9293 explains why: sequence numbers are not tied to a shared clock, so each side has to confirm the other's before data flows. The useful part for troubleshooting is what you see when it fails. Under RFC 9293 a SYN that arrives for a port with no connection is answered with a reset, so an immediate reset usually tells you the host was reached but nothing accepted the connection (a firewall set to reject with a reset looks the same). No answer at all suggests something in the path dropped the SYN or the reply, such as a firewall or a missing route back.
"How does traceroute work, and what do the asterisks mean?"
On Windows, tracert sends ICMP echo requests with a time-to-live of 1, then 2, then 3, and so on. Every router must reduce the TTL by at least one, and the router that takes it to zero sends back an ICMP Time Exceeded message, which reveals that hop. The destination answers with an echo reply. Asterisks mean no reply arrived within the time-out for that hop, which often just means that router does not return Time Exceeded messages; it is not proof that traffic stops there. By default tracert gives up after 30 hops.
"A user on one floor can't reach a server. Talk me through it."
The panel wants an order, not a list of commands. A sound one, working up from the user's machine:
- Check the machine has a sensible address, mask and gateway. An address in 169.254.0.0/16 is a link-local address the machine gave itself (RFC 3927), which usually means DHCP never answered.
- Ping the default gateway. If that fails, the problem is local: the cable, the port, the VLAN on that port, or the gateway itself.
- Ping the server by address. If the gateway answers and the server does not, run tracert to see where the path stops, remembering what asterisks do and do not mean.
- Try the name. If the address works and the name does not, it is DNS, not the network.
- Test the application port. A server that answers ping can still have a firewall or a stopped service in front of the port you need.
Then say what you would check that you cannot test from the desk: whether anyone else on that floor or VLAN is affected, and whether anything changed recently. The troubleshooting interview guide covers how to talk through a problem like this one in more depth.
Preparing in the week before
Work ten subnetting questions a day until /26, /27 and /28 come without thinking; the subnet splitter is useful for checking your answers. Build two VLANs in a lab, trunk them between two switches or virtual switches, route between them and then break the trunk on purpose, so that when you are asked about native VLANs and tagging you are describing something you have seen.
If the role asks for a certification, look at what the exam covers rather than collecting another. CompTIA's Network+ (exam N10-009, launched June 2024) has up to 90 questions in 90 minutes, and CompTIA recommends A+ and nine to twelve months in a junior network role before taking it, so it suits someone already in support. The certification paths page sets out where Network+ and the vendor routes fit.
Finally, prepare one networking problem of your own to describe: something you traced and fixed, at work or in your lab, told in a few sentences with what you checked and why. Panels for junior roles often ask for one.
Common questions
- Will I be allowed a calculator for subnetting questions?
- Assume not. Many panels ask you to work it on a whiteboard or out loud, because the working shows them more than the answer. Learn the block sizes for /25 to /30 so you can do the common cases in your head.
- Do I need a Cisco certification to get a junior network role?
- Not always. Read the advert: some ask for a named vendor certification, others for Network+ or equivalent experience. The concepts in this guide are the same on every vendor's kit, and a lab you can describe in detail counts for a lot at junior level.
- What if I am asked about a protocol I have never used?
- Say so, then explain what you would expect it to do from the layer it works at and how you would find out. Panels for junior roles usually prefer that to a confident guess.
Sources
Every figure and rule in this guide comes from these pages, read on 21 November 2026. Pay rates, exams and schemes change: check the source before you rely on one.
- Network engineer: job profile (National Careers Service)
- TCP/IP addressing and subnetting (Microsoft Learn)
- RFC 1918: Address Allocation for Private Internets (IETF)
- RFC 826: An Ethernet Address Resolution Protocol (IETF)
- RFC 2131: Dynamic Host Configuration Protocol (IETF)
- Bridging and VLANs (Juniper Networks)
- Configuring STP (Juniper Networks)
- RFC 4632: Classless Inter-domain Routing (CIDR) (IETF)
- Introduction to OSPF (Juniper Networks)
- RFC 3927: Dynamic Configuration of IPv4 Link-Local Addresses (IETF)
- RFC 9293: Transmission Control Protocol (TCP) (IETF)
- RFC 792: Internet Control Message Protocol (IETF)
- tracert (Microsoft Learn)
- Network+ (Plus) Certification (CompTIA)
Spotted something wrong or out of date? Tell us and we will correct it.


