This page needs a newer browser to run its tools. You can still read everything here.

Data centres, cloud and the full picture interview questions

Interview questions on data centres, cloud and the full picture: warm-ups, then what each harder question tests and how a good answer starts.

7 questions: 2 warm-ups and 5 with model answers · Levels: Fresher to senior

Warm-up questions

Short questions with short answers, each answered in full.

Fresher to mid

How does a layer-4 load balancer differ from a layer-7 one?

Answered in the lesson Load balancers and reverse proxies
Fresher to mid

What is the difference between traffic policing and traffic shaping?

Answered in the lesson QoS, queues and traffic shaping

Questions with model answers

What each question tests, the first step of a good answer and the follow-ups to expect.

Mid to senior

An API behind your load balancer allows each client 100 requests a minute, keyed by the first address in X-Forwarded-For. One client sends thousands a minute and is never limited. Explain how, and design a key the client cannot forge.
What it tests

How proxies build X-Forwarded-For, which entries a client controls, the trusted-proxy rule, and the layer-4 and layer-7 ways of passing the client's address on.

First step

Write down who wrote each entry of the header. The client may send any X-Forwarded-For it likes, and each proxy only appends the address it received the connection from. Then ask which entries your own proxies added and which arrived from outside.
Follow-ups
- The API sits behind a CDN and then your own balancer. Which entry is the client now?
- How does the answer change if the balancer works at layer 4?
- Many users behind carrier-grade NAT share one address. Is the address still a good key?
From the lesson Load balancers and reverse proxies
Model answer, Pro. Model answers open with any pass; prices are in the notice on this page.
Opening the model answers with your pass…
Model answers open with any MySmartCoPilot pass.
Pro pass · 1 week₹79$2.99
30 days₹179$5.99
1 year₹1,699$45
One payment; a pass never renews. The price shown is the total you pay: no fees or charges are added at checkout.
Passes can’t be bought right now, so the model answers can’t be opened at the moment. If you have a pass, enter your licence key.
Choose a pass I have a licence key
It also opens every other in-depth part in Learn and every Pro tool. Sales are final, except where the law gives you a right to withdraw, for example in the EU and the UK (refund policy).
In the answer:
- Model answer
- Follow-up answers
- Common mistakes
- Rubric
- Variants
- about 3 minutes
Mid to senior

Every deployment causes about ten seconds of 502 errors, although each new server passes its health checks before it takes traffic. The old servers are stopped as soon as the new ones are healthy. Find the causes and design a deployment without errors.
What it tests

Draining, the gap between a process stopping and the balancer noticing, keep-alive races, and graceful shutdown in HTTP/1.1 and HTTP/2.

First step

Draw the timeline of one old server: when it stops accepting work, when the balancer stops sending it requests, and which requests are in flight or about to be sent on idle keep-alive connections at that moment. The errors live in the gaps between those moments.
Follow-ups
- Some clients hold WebSocket connections for hours. How long do you drain?
- Why can a request sent on an idle keep-alive connection fail even with draining?
- How do the balancer's health-check settings change the plan?
From the lesson Load balancers and reverse proxies
Model answer, Pro.
In the answer:
- Model answer
- Follow-up answers
- Common mistakes
- Rubric
- Variants
- about 3 minutes
Fresher to mid

A home office has a connection of 100 Mbit/s down and 20 Mbit/s up. Video calls break up whenever a large file is uploading, although a call needs about 3 Mbit/s. Explain why, and say what you would change on the router.
What it tests

Recognising a full queue at the uplink, computing its delay, and the fix: shaping below the link's rate with a flow-aware queue manager.

First step

Measure latency while the upload runs, for example by pinging the ISP's first hop. If it rises by hundreds of milliseconds, a queue is filling in front of the 20 Mbit/s uplink: work out how long a full buffer takes to drain at that rate.
Follow-ups
- Why does a faster download plan not help?
- Why shape at 19 rather than 20 Mbit/s?
- Would marking the call's packets EF be enough on its own?
From the lesson QoS, queues and traffic shaping
Model answer, Pro.
In the answer:
- Model answer
- Follow-up answers
- Common mistakes
- Rubric
- Variants
- about 3 minutes
Mid to senior

After video conferencing was given a strict-priority queue on a 50 Mbit/s branch link, nightly backups and some daytime file transfers began to fail during company-wide meetings. Why, and how would you configure the link instead?
What it tests

Starvation under strict priority, sizing and policing a priority class, and weighted sharing for everything else.

First step

Add up what the priority queue carries during a large meeting: each participant's video stream times the number of people at the branch. Compare it with 50 Mbit/s, and remember that strict priority serves the other queues only when the priority queue is empty.
Follow-ups
- How large should the priority class be, and what happens to video above it?
- Why might a weighted queue for video be better than priority?
- How would you show from interface counters that starvation is the cause?
From the lesson QoS, queues and traffic shaping
Model answer, Pro.
In the answer:
- Model answer
- Follow-up answers
- Common mistakes
- Rubric
- Variants
- about 3 minutes
Mid to senior

Calls between two offices are marked EF on the LAN, but captures at the far office show DSCP 0, and the calls are choppy only when the WAN link is busy. Where can the marks be lost, and what would you do?
What it tests

Trust boundaries, re-marking at borders and tunnels, finding the hop that re-marks, and why the queue at the WAN edge matters most.

First step

Capture at several points along the path: leaving the phone, leaving the access switch, entering and leaving the WAN router, and at the far end. The first point where the DSCP changes is the device that re-marks the packets or does not trust their marks.
Follow-ups
- The calls cross an IPsec tunnel. Does the outer header keep the DSCP?
- The provider re-marks everything to zero. What can you still control?
- What happens to ECN if a device clears the whole byte?
From the lesson QoS, queues and traffic shaping
Model answer, Pro.
In the answer:
- Model answer
- Follow-up answers
- Common mistakes
- Rubric
- Variants
- about 3 minutes

All Computer Networks interview questions · The Computer Networks track

Warm-up questions

How does a layer-4 load balancer differ from a layer-7 one?

What is the difference between traffic policing and traffic shaping?

Questions with model answers

An API behind your load balancer allows each client 100 requests a minute, keyed by the first address in X-Forwarded-For. One client sends thousands a minute and is never limited. Explain how, and design a key the client cannot forge.

Every deployment causes about ten seconds of 502 errors, although each new server passes its health checks before it takes traffic. The old servers are stopped as soon as the new ones are healthy. Find the causes and design a deployment without errors.

A home office has a connection of 100 Mbit/s down and 20 Mbit/s up. Video calls break up whenever a large file is uploading, although a call needs about 3 Mbit/s. Explain why, and say what you would change on the router.

After video conferencing was given a strict-priority queue on a 50 Mbit/s branch link, nightly backups and some daytime file transfers began to fail during company-wide meetings. Why, and how would you configure the link instead?

Calls between two offices are marked EF on the LAN, but captures at the far office show DSCP 0, and the calls are choppy only when the WAN link is busy. Where can the marks be lost, and what would you do?