roadmap
- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
Readies you for Network Engineer interviews — designing, securing, and operating IP networks. Spans routing protocols, wireless, network security, cloud networking, and automation.
6125 of 6125 questions in your plan: 0 known, 0 in progress, 0 due, 0 skipped, 6125 new. 0 cut, outside your plan.
The first conversation in a network engineer loop is usually a fundamentals screen that needs no whiteboard: subnet a block in your head, explain what happens between typing a URL and the page arriving. One or two technical rounds on routing, switching and security follow, and many loops build one of them around a broken network rather than a question: a lab, a whiteboard topology or a page of `show` output with a fault hidden in it. A design round joins the loop once the level reaches middle: a campus with redundant gateways, a leaf-spine data center, or an office joined to a cloud VPC over IPsec with BGP across the tunnel. Automation-leaning roles add a Python scripting round, and the behavioral close expects a story about an outage or a change window that went wrong. The troubleshooting rounds have their favorites. OSPF neighbors stalled in EXSTART, which classically points at an MTU mismatch. A BGP session stuck in Active because the TCP connection to port 179 never completes. A switch plugged into an access port that wins the spanning-tree root election. Large transfers hanging over a VPN while ping still works, because a firewall drops the ICMP that path MTU discovery needs. Seniority changes where the question starts more than what it is about. Early in a career you are asked how OSPF forms an adjacency or how BGP picks a best path, and the account has to be right. From senior up you are handed the symptom, and expected to name the layer at fault, the capture or command that proves it, and a fix that fits a change window without breaking what still works.
The plan opens on the machine you work from rather than the network: a shell on the systems most network tooling runs on, Bash for one-off jobs and [Python](/topics/lang-python) for parsing device output and calling APIs, then [algorithms](/topics/found-dsa), because automation-heavy loops often keep a standard coding round. The [protocol core](/topics/proto-networking) is the center of gravity, and it is one module because the questions cross layers constantly: Ethernet and ARP under IP, spanning tree at layer 2, [OSPF](/topics/proto-ospf) and [BGP](/topics/proto-bgp) above it, the client services, and design and high availability to assemble them into a topology. NETCONF and YANG sit here beside SNMP, so the vocabulary of automation arrives long before its tooling. Next the plan climbs to what travels over the network: [HTTP, TLS and PKI](/topics/proto-web-transport), then [SSH](/topics/proto-ssh), the way into most devices, then two older protocols that still open tickets, FTP with its [separate data channel](/topics/proto-ftp-nat-firewall) tripping over NAT, and mail with its DNS records. Wireless gets its own stretch, from [medium access](/topics/proto-wifi-medium-access) and roaming to the WPA generations and the attacks on them. Security follows wireless directly, and its tools are learned by running them: scans and captures only become familiar once you have aimed them at a network where you may run them without asking anyone. [Network defense](/topics/sec-network-defense) comes first, then Nmap, packet capture and the AAA pair of RADIUS and [TACACS+](/topics/proto-tacacs). RADIUS also finishes the Wi-Fi story, since it carries the EAP exchange WPA2-Enterprise relies on. The second half moves the network into software. [Proxies and load balancers](/topics/cloud-networking) lead into the [Kubernetes Service and CNI model](/topics/cloud-kubernetes-services-networking), then the cloud in provider-neutral terms: [regions and zones](/topics/cloud-providers-concepts-footprint) first, then [the private network](/topics/cloud-providers-concepts-network), where subnets, egress paths, peering and hybrid links are the part to know best, then account boundaries, outage behavior and what moving data costs. Git, [infrastructure as code](/topics/cloud-iac-concepts) and CI/CD turn configuration into reviewed changes, and [observability foundations](/topics/cloud-observability-foundations) cover what to measure once it runs. Its last module is about the interview rather than the network: the [behavioral conversation](/topics/found-behavioral), where the outage or botched change window gets told as a story, and the stages a loop moves through.
A screen in the next few days is mostly the protocol core and its failure cases, so spend the days there, with importance raised to must know and level held to the band the posting names; the order the page opens with needs no change. Practice subnetting on paper alongside, because reading answers does not make you fast at it. With a couple of months, pick by topics as the study order and stand up a lab of virtual routers early: configure each protocol on it and capture it with tcpdump before moving on, since lab rounds reward having seen the failure once already. Engineers who learned the trade on a console cable usually find the protocol and security modules familiar and the later ones thin. Systems and cloud engineers moving the other way have the opposite gap, and should expect spanning tree, OSPF and BGP to be probed at a depth cloud work rarely asks for. The plan stops where the network hands traffic to the services on it. Building container images is absent, and so is reliability practice such as service levels, incident response and on-call; the [DevOps / SRE Engineer roadmap](/roadmaps/devops) covers that ground. Security here is the network's own defense and visibility; a move toward penetration testing or a security operations team is what the [Cyber Security Expert roadmap](/roadmaps/cyber-security) prepares for.
The first conversation in a network engineer loop is usually a fundamentals screen that needs no whiteboard: subnet a block in your head, explain what happens between typing a URL and the page arriving. One or two technical rounds on routing, switching and security follow, and many loops build one of them around a broken network rather than a question: a lab, a whiteboard topology or a page of show output with a fault hidden in it. A design round joins the loop once the level reaches middle: a campus with redundant gateways, a leaf-spine data center, or an office joined to a cloud VPC over IPsec with BGP across the tunnel. Automation-leaning roles add a Python scripting round, and the behavioral close expects a story about an outage or a change window that went wrong.
The troubleshooting rounds have their favorites. OSPF neighbors stalled in EXSTART, which classically points at an MTU mismatch. A BGP session stuck in Active because the TCP connection to port 179 never completes. A switch plugged into an access port that wins the spanning-tree root election. Large transfers hanging over a VPN while ping still works, because a firewall drops the ICMP that path MTU discovery needs.
Seniority changes where the question starts more than what it is about. Early in a career you are asked how OSPF forms an adjacency or how BGP picks a best path, and the account has to be right. From senior up you are handed the symptom, and expected to name the layer at fault, the capture or command that proves it, and a fix that fits a change window without breaking what still works.
The plan opens on the machine you work from rather than the network: a shell on the systems most network tooling runs on, Bash for one-off jobs and Python for parsing device output and calling APIs, then algorithms, because automation-heavy loops often keep a standard coding round.
The protocol core is the center of gravity, and it is one module because the questions cross layers constantly: Ethernet and ARP under IP, spanning tree at layer 2, OSPF and BGP above it, the client services, and design and high availability to assemble them into a topology. NETCONF and YANG sit here beside SNMP, so the vocabulary of automation arrives long before its tooling.
Next the plan climbs to what travels over the network: HTTP, TLS and PKI, then SSH, the way into most devices, then two older protocols that still open tickets, FTP with its separate data channel tripping over NAT, and mail with its DNS records. Wireless gets its own stretch, from medium access and roaming to the WPA generations and the attacks on them.
Security follows wireless directly, and its tools are learned by running them: scans and captures only become familiar once you have aimed them at a network where you may run them without asking anyone. Network defense comes first, then Nmap, packet capture and the AAA pair of RADIUS and TACACS+. RADIUS also finishes the Wi-Fi story, since it carries the EAP exchange WPA2-Enterprise relies on.
The second half moves the network into software. Proxies and load balancers lead into the Kubernetes Service and CNI model, then the cloud in provider-neutral terms: regions and zones first, then the private network, where subnets, egress paths, peering and hybrid links are the part to know best, then account boundaries, outage behavior and what moving data costs. Git, infrastructure as code and CI/CD turn configuration into reviewed changes, and observability foundations cover what to measure once it runs. Its last module is about the interview rather than the network: the behavioral conversation, where the outage or botched change window gets told as a story, and the stages a loop moves through.
A screen in the next few days is mostly the protocol core and its failure cases, so spend the days there, with importance raised to must know and level held to the band the posting names; the order the page opens with needs no change. Practice subnetting on paper alongside, because reading answers does not make you fast at it.
With a couple of months, pick by topics as the study order and stand up a lab of virtual routers early: configure each protocol on it and capture it with tcpdump before moving on, since lab rounds reward having seen the failure once already.
Engineers who learned the trade on a console cable usually find the protocol and security modules familiar and the later ones thin. Systems and cloud engineers moving the other way have the opposite gap, and should expect spanning tree, OSPF and BGP to be probed at a depth cloud work rarely asks for.
The plan stops where the network hands traffic to the services on it. Building container images is absent, and so is reliability practice such as service levels, incident response and on-call; the DevOps / SRE Engineer roadmap covers that ground. Security here is the network's own defense and visibility; a move toward penetration testing or a security operations team is what the Cyber Security Expert roadmap prepares for.