What is Netflix system design, in short?

It is two systems. A control plane on AWS handles sign-in, the home page, licences and which server should send the video. A data plane, Open Connect, is more than 14,000 Netflix servers in 142 countries. Many sit inside internet providers, the rest at internet exchanges, and they only store and send video files. Pressing play is where the two meet: AWS returns appliance URLs and the device streams from the appliance.

What is Netflix Open Connect?

Open Connect is Netflix's own content delivery network, started in 2011. Netflix places servers at internet exchanges and, free of charge, inside internet providers. The servers store encoded video and serve it over HTTPS. Netflix's 2021 briefing paper, with data as of 2020, counts more than 14,000 of them in 142 countries. More than 1,000 providers host them.

How does Netflix decide which server sends my video?

Every appliance keeps reporting its health, its network routes and the files it stores to the control plane in AWS. When you press play, the playback service picks the files for your device. The steering service picks appliances that have them, are healthy and are close to you on the network. Your device receives their URLs and streams from them.

Why does Netflix copy videos to servers at night?

Netflix can predict what members will watch and when. So it copies files during each region's quiet hours, which it calls fill windows. It does not wait for the first viewer to ask. A few servers pull from S3 first and the rest copy from them. The busy evening is then spent only on serving.

What is per-title encoding?

Netflix does not use one fixed set of bitrates for every title. It picks a set for each title, based on how complex it is. In Netflix's 2015 example, BoJack Horseman streamed in 1080p at 1540 kbps instead of 480p at 1750 kbps. Orange is the New Black's top bitrate dropped from 5800 to 4640 kbps, 20% less, with the same quality.

What database does Netflix use?

Several. Cassandra is the backbone for member data such as viewing history. EVCache, Netflix's memcached-based cache, sits in front for fast reads. Netflix's key-value data layer hides the store behind one API. A use case can be backed by Cassandra, EVCache or both. Video files live in S3 and on the Open Connect appliances.

How does Netflix survive an AWS region outage?

It runs active-active: more than one region serves members at the same time. DNS normally sends users to the nearer region. If one region fails, Netflix moves everyone to the healthy one. No user request may call across regions. Cassandra replicates data in the background, and EVCache entries are invalidated across regions by message.

How many concurrent streams can Netflix handle?

Netflix's public record is 65 million concurrent streams, for the Jake Paul vs Mike Tyson fight in November 2024. For an interview, use that as the peak and estimate bandwidth from it. At an assumed 3 Mbps average, 65 million streams is about 195 Tbps.

System design interview guide

Netflix System Design: How Playback, Open Connect and Encoding Work

Short answer

Netflix system design is two planes that meet at the play button. The control plane runs on AWS: the Zuul gateway, playback, steering, and member data in Cassandra with EVCache in front. The data plane is Open Connect, more than 14,000 Netflix servers in 142 countries, many inside internet providers. When you press play, AWS checks your account and picks the files your device can play. It returns URLs of nearby healthy servers. Your device then streams straight from them, choosing quality piece by piece. Files are copied to those servers ahead of time, during quiet hours.

Netflix in its own numbers

Every figure below is quoted from Netflix's own engineering writing. Tap a source to check it.

300M+
users at the end of 2024
Netflix TechBlog, 2025
65M
streams at the same moment, Paul vs Tyson, Nov 2024
Netflix newsroom
14,000+
Open Connect servers in 142 countries
Open Connect briefing paper, 2021
1,000+
internet providers host Netflix servers inside their networks
Open Connect briefing paper, 2021

in this guide

Design Netflix, part by part

8 parts open to everyone · 3 with a free account · 3 with the course

You can read 8 of 14 parts. A free account opens the deep dive; the course opens all of it.

Requirementsopen
Scale estimatesopen

Why this question is asked

Design Netflix checks whether you can split a system by what each part must be good at. Sign-in and recommendations need logic and data. Video delivery needs raw bandwidth close to the viewer. A candidate who draws one load balancer in front of one service has missed the point. The interviewer also wants to hear cost reasoning, because moving video bytes is the most expensive thing Netflix does.

What the interviewer asks

“Design a video streaming service like Netflix. A member opens the app, browses rows of titles, presses play and watches on any device. Focus on what happens when they press play.”

Ask these back before you draw

1.Is this on-demand video only, or live events too? (Start with on-demand; mention live at the end.)
2.How many members, and how many streams at the peak? (Use Netflix's public numbers below.)
3.Which devices? TVs, phones and browsers all decode different formats.
4.Do we own the delivery network, or rent a commercial CDN?

This page is the free part.

The course goes deeper on the Netflix design

₹1,111 in India$25 everywhere elseonce, for the whole course

The System Design course covers the Netflix design across a run of lessons, not one page. These 4 alone are about 94 minutes of step-by-step reading, every one with a quiz.

01Requirements

Requirements

Always clarify these in the first 5 minutes of the interview. Do not start drawing boxes until both lists are agreed.

Functional requirements

A member signs in and picks a profile
The home page shows personalised rows of titles

Go deeper

This guide is built on these 6 lessons. Each one teaches one piece of the Netflix design above, step by step.

Practice with 770 system design lessons

Lifetime access for ₹1,111 in India or $25 elsewhere. Interactive diagrams, quizzes, and 20 capstone projects to practise on.

Get lifetime access See all interview problems

Two courses by the author of this page:

System Design

770 lessons · 18 free to read

₹1,111 in India · $25 elsewhere, once

Get System Design

You own this course

Generative AI and Agentic AI

214 lessons · 10 free to read

₹1,333 in India · $35 elsewhere, once

Get Generative AI and Agentic AI

You own this course

Every lesson in both courses has an AI tutor beside it. It reads the same lesson you are reading and answers your questions from it. Try the tutor

Save your progress and get the interview checklist PDF

System Design Interview Cheat Sheet: the 7 interview stages, the 10 points a scorecard grades, 12 classic systems. 18 pages. Free with an account.

Netflix counted over 300 million users at the end of 2024. Its record live night peaked at 65 million streams at the same moment. The video comes from more than 14,000 Netflix servers in 142 countries. Many sit inside the viewer's own internet provider; the rest are at internet exchanges.

Designing Netflix is two designs that meet at the play button. The control plane runs in AWS. It handles sign-in, the home page, what to play and which server should send it. The data plane is Open Connect, Netflix's own content delivery network. Its servers sit inside internet providers and only store and send video files. Most of the interview is about the line between those two planes. It is also about the work done before anyone presses play. Every title is encoded into many versions, and files are copied to the right servers overnight.

Roni Das

11 June 2026 · Updated 1 November 2026 · 23 min read

Where it shows up

One of the most common system design interview questions. It also appears as design YouTube, design a video streaming service or design Hotstar. Netflix's own engineering blog is unusually detailed, so interviewers expect real numbers.

Spec sheetNetflix

01Users: 300M+
02Peak concurrent streams: 65M
03Peak egress, estimated: ~195 Tbps
04Servers needed, estimated: 2,200+

worked through below, with the maths

of encrypted video from one server

99.99%

Netflix's internal availability goal

Netflix TechBlog, 2013

Data modelfree account

Deep dives (8)free account

Unlock with the course · $25One payment, lifetime access to every lesson.

The hard parts of this design are taught step by step in the System Design course: Content Delivery Network (CDN), Adaptive Bitrate Streaming and Video Encoding.

Is recommendation quality in scope, or only the serving of rows?

Content Delivery Network (CDN)

A CDN is thousands of caches you do not operate, and the cache key decides everything about whether it helps you.

26 min · diagrams · quiz

Adaptive Bitrate Streaming

Automatically adjusting video quality based on the viewer's network speed, smooth playback across all connections.

20 min · diagrams · quiz

Video Encoding

Converting raw video into compressed formats for web delivery, codecs, containers, and quality vs size tradeoffs.

20 min · diagrams · quiz

API Gateway

Every mechanism in this chapter running in one component in front of everything, what that concentration costs, and how to keep it from becoming an application.

28 min · diagrams · quiz

all part of

System Design Masterclass

770 lessons · about 287 hours · 18 free to read

₹1,111in India, by UPI

$25everywhere else, by PayPal

Get the System Design course Read 18 lessons free first

You own this course

Continue with Content Delivery Network (CDN)

pay once,
yours for life

Live from the course bench

Before you design Netflix, try the questions it rests on.

Real interview questions, live diagrams and measured lab results, pulled at random from the lessons. A new set every time you come back.

Two engineers sketching a system design on a whiteboard

where the work happens

Stuck on a card?

Ask the course tutor. It answers with a real diagram, drawn for your question.

Ask the tutor

Laying out the cards…

Pressing play starts the video quickly on a TV, phone or browser

Quality adjusts by itself when the network gets slower or faster

Playback resumes where the member stopped, on any device

Subtitles and several audio languages per title

New titles are available everywhere at their release time

Non-functional requirements

Very high availability on the play path: Netflix's internal goal is 99.99%
Start-up delay of a few seconds at most, and as few pauses for buffering as possible
Peak load: tens of millions of streams at the same moment (65 million on the record night)
Video served from inside or near the member's internet provider, to save cost and distance
One AWS region can fail without members noticing
Viewing history and bookmarks may be a little stale; billing and licences may not

02Scale estimates

Back-of-envelope scale estimates

Show your math. Pulling numbers from thin air signals you have not thought about the load.

Users

300M+

Netflix's own figure for the end of 2024, from its 2025 recommendation post. Each account can have several profiles, and each profile has its own history and rows.

Peak concurrent streams

65M

The Paul vs Tyson fight in November 2024, from Netflix's newsroom. Size the play path for this, not for an average evening.

Peak egress, estimated

~195 Tbps

Assume an average of 3 Mbps per stream. That is our assumption. In Netflix's 2015 fixed ladder, 720p started at 2350 kbps and the top rung was 5800 kbps. 65 million x 3 Mbps is about 195 terabits per second.

Servers needed, estimated

2,200+

One Open Connect server sent 90 Gbps of encrypted video in Netflix's 2017 tests. 195 Tbps / 90 Gbps is about 2,200 servers running flat out, before any spare. Netflix runs more than 14,000.

Viewing history writes vs reads

9 : 1

Netflix's 2018 post on viewing history: about nine writes for every read. That makes it a write-heavy store, which is why it sits on Cassandra.

03Architecture

High-level architecture

Draw two planes. The control plane runs on AWS. Every API call from a device enters through Zuul, Netflix's gateway. Behind it sit services for sign-in, the home page rows, playback and steering. Member data such as viewing history lives in Cassandra, with EVCache (Netflix's memcached-based cache) in front for fast reads. The data plane is Open Connect. It is thousands of Netflix servers, called Open Connect Appliances, placed at internet exchanges and inside internet providers. An appliance stores encoded video files and sends them over HTTPS. It keeps no member data. It reports its health, its network routes and its list of stored files to the control plane. When a member presses play, the playback service checks the account and licence. Then it picks the files this device can play. The steering service then picks the best appliances for this member and hands back their URLs. From then on, the device fetches video straight from the appliance. Before any of this, an encoding pipeline turns each studio master into many files. A fill system copies them to the appliances during quiet hours.

In a real interview, sketch this on the whiteboard before diving into any single box.

Figure 1. Two planes: decisions in AWS, video bytes from Open ConnectTap the diagram to open it full size.Everything before the first video byte is a decision, and decisions run in AWS. That means Zuul, playback, steering, and member data in EVCache and Cassandra. The bytes themselves come from an Open Connect appliance near the viewer. The appliance holds no member data; it only stores files and serves them.

Share this diagramLinkedIn

.walkthrough-module__Xo_9gG__off { display: block !important; } .walkthrough-module__Xo_9gG__controls, .walkthrough-module__Xo_9gG__dots, .walkthrough-module__Xo_9gG__breakRow { display: none !important; } .walkthrough-module__Xo_9gG__steps { display: grid; gap: 14px; }

Interactive

Press play, step by step

The same two planes, one play request at a time. Each step lights up the part doing the work. Then press Break it and replay with a server failing or the network slowing down.

One Netflix play request

Decisions in AWS on top, video servers in the internet provider below.

Normal playStep 1 of 8

Normal play, step 1 of 8
Two planes on one board
The top half runs in AWS: Zuul, the gateway, then Playback and Steering, with your viewing History in Cassandra. The bottom right is Open Connect: Netflix servers, called appliances, inside your internet provider or at an internet exchange.
Normal play, step 2 of 8
Before you press anything
Every appliance keeps reporting to AWS: is it healthy, which network routes it has learned and which files it stores. This is why the control plane can choose a server in one step later. The appliance itself knows nothing about you.
Normal play, step 3 of 8
You press play
The TV sends the request to Netflix in AWS. Every API call from a device enters through Zuul, Netflix's gateway. Zuul routes it to the right service and can shed load if a service is struggling.
Normal play, step 4 of 8
Can you watch this here?
Playback checks your account and whether this title is licensed in your country. Then it picks the exact files for this TV: the codec it can decode, the resolution, and your audio and subtitle languages.
Normal play, step 5 of 8
Where did you stop?
Playback reads your viewing history to resume at the right second. Netflix keeps history in Cassandra, one partition per member, with EVCache in front so the read is fast. History is written about nine times for every read.
Normal play, step 6 of 8
Which server sends it?
Steering looks at the appliance reports and picks servers that hold the files, are healthy and are close to you on the network. It turns them into URLs.
Normal play, step 7 of 8
The TV gets URLs, not video
The answer to the play request is small: a list of appliance URLs for the chosen files. The heavy work happens next, and it does not touch AWS.
Normal play, step 8 of 8
Video comes from inside your provider
The TV fetches the video from appliance A over HTTPS, piece by piece. One appliance served 90 Gbps of encrypted video in Netflix's 2017 tests. AWS has stepped out of the video path completely.

An appliance dies, step 1 of 4
Playing as normal
Replay the same film. The TV is fetching pieces from appliance A and has a few seconds of video buffered ahead.
An appliance dies, step 2 of 4
Appliance A goes quiet
A disk fails, or the appliance loses power. The next piece never arrives. The buffer starts to drain while the TV waits.
An appliance dies, step 3 of 4
Try the next URL
In this design the TV was given more than one URL, so it asks appliance B for the next piece and keeps playing. Netflix's guidance to providers also routes traffic to a second site, then to Netflix servers at the exchange, when one site is down.
An appliance dies, step 4 of 4
Nobody new is sent to A
Appliance A no longer reports healthy, so steering stops choosing it for new viewers. Netflix's operations team monitors every appliance and replaces failed hardware at no cost to the provider.

Network slows, step 1 of 3
Pieces arrive too slowly
The TV is fetching a high rung of the ladder. Now each piece takes longer to download than it takes to play. If nothing changes, the buffer empties and the picture freezes.
Network slows, step 2 of 3
Step down a rung
The player, not the server, decides. For the next pieces it picks a lower rung of this title's ladder, a smaller file at a lower bitrate. The picture softens a little, but it keeps moving.
Network slows, step 3 of 3
Climb back carefully
When pieces arrive fast again and the buffer is healthy, the player steps up, one rung at a time. Per-title ladders help. In Netflix's 2015 example, one BoJack episode streamed in 1080p at 1540 kbps, where the fixed ladder gave 480p at 1750 kbps.

or use the ← → keys

Replay this play request with a server dying mid-film or the home network slowing down.

Based on Netflix's Open Connect overview and deployment guide. Also Netflix TechBlog posts on viewing history, serving 100 Gbps and per-title encoding. Moving to the next URL is this guide's design.

04One request

What happens when you press play, step by step

This is Netflix's own five-step description of playback, from its Open Connect overview. Encoding and fill must finish before step 2.

1
Appliances report in
All the time, every Open Connect appliance reports to the control plane in AWS. It sends its health, the network routes it has learned and the files it stores.
2
The device asks to play
The member presses play. The device calls the Netflix application in AWS, through the Zuul gateway, with the title it wants.
3
Playback checks and chooses files
The playback services check that the member may watch this title in this country. Then they pick the exact files to send, based on the device and the network.
4
Steering picks appliances
The steering service reads the appliance reports. It picks servers that have the files, are healthy and are close to the member on the network. It turns them into URLs.
5
The device streams from the appliance
The device gets the URLs and fetches the video from the appliance over HTTPS, piece by piece. AWS is not in the video path at all.

A top-down flow of one play request. 1: Open Connect appliances send reports to the steering service. 2: the device sends play to the playback services in AWS. 3: playback tells steering which files are needed. 5: playback returns appliance URLs to the device. The caption gives step 4: steering hands the URLs back to playback. Then the device fetches the files from the appliances. — Figure 2. Netflix's five playback steps, from its Open Connect overviewTap the diagram to open it full size.Step 1 happens all the time, before anyone presses play. The appliances keep telling AWS what they hold and whether they are healthy. That is why steering can answer in one call. The device ends up with URLs and talks to the appliance directly.

05Components

Core components

Walk through each service. The interviewer wants to hear what each one owns, not just the names.

Zuul (API gateway)

The front door for every API call from a device. It routes requests, sheds load when a backend is struggling, and can send traffic to another AWS region. Zuul 2 moved to non-blocking network code on Netty so it can hold many persistent connections cheaply.

Playback service

Answers the play request. It checks the account and the title's licence. Then it chooses the encoded files that fit this device: codec, resolution, audio and subtitles.

Steering service

Chooses which appliances serve this member. It reads what every appliance reports: health, network routes and stored files. It returns a short list of URLs, so the device never has to guess a server.

Open Connect appliances

Netflix's own servers, at internet exchanges and inside internet providers. They store encoded video files and send them over HTTPS. Netflix gives the hardware to the provider for free; the provider gives space, power and network.

Encoding pipeline (Cosmos)

Turns a studio master into streamable files. Services inspect the master, measure how complex it is and pick a bitrate ladder for that title. Then they encode it in chunks and score every encode with VMAF, Netflix's quality metric.

Fill system

Copies files from S3 to the appliances during each region's quiet hours. A few appliances pull from S3 first; the rest copy from them. The control plane tells each appliance what it should hold.

Member data: Cassandra and EVCache

Viewing history, bookmarks and other member state live in Cassandra, which takes heavy writes and replicates across regions. EVCache sits in front so most reads never touch Cassandra.

Personalisation

Builds the rows on the home page. Heavy models run offline in batch. Quick updates run nearline as events arrive. A thin online step adjusts the result when the page is requested.

the deep dive

Data model

title_files (catalog of encodes). One row per encoded file. The playback service filters these by what the device can decode. Written by the encoding pipeline, read on every play. Rarely changes, so it…

free with an account

Every table for Netflix, its keys, and which store fits it and why.

One tap with Google. Free, no card.

the deep dive

Deep dives

Press play: the control plane hands out appliance URLs. The key idea is that AWS decides and the appliance delivers. When the member presses play, the device does not look for a video server by itself. It asks the playback…

free with an account

8 deep dives on the parts that break at scale, with 4 diagrams: Press play: the control plane hands out appliance URLs; Open Connect: put the bytes inside the internet provider; Fill: copy tomorrow's titles tonight.

One tap with Google. Free, no card.

the deep dive

Trade-offs to discuss

Own CDN (Open Connect) vs a commercial CDN. A commercial CDN is the right start for almost everyone. Netflix's volume changes the maths: video is most of its cost, and its traffic is very predictable. Owning the…

free with an account

6 decisions an interviewer expects you to defend, each with the alternative and why it loses here.

One tap with Google. Free, no card.

for course owners

What the interviewer asks next

A new season releases worldwide at midnight. What do you do?

with the System Design course

6 follow-up questions with model answers, the interview scorecard for this question, and a mock interview on Design Netflix with the AI tutor.

Unlock with the course · $25One payment, lifetime access to every lesson.

10In production

How Netflix actually does it

Everything on this page comes from Netflix's own writing. The playback steps and the appliance design are from the Open Connect overview. The counts of servers, countries and providers are from Netflix's 2021 Open Connect briefing paper. So are the US$1 billion and US$1.2 billion figures. Fill windows, manifests and fill masters are from the 2016 post Netflix and Fill. The 90 Gbps figure is from the 2017 post on serving 100 Gbps from one appliance. Per-title encoding and its numbers are from the 2015 post. The Cosmos services are from the 2024 post on the video pipeline. Zuul 2, active-active regions, the key-value data layer, viewing history and EVCache each have their own post. Older posts describe the system at that date. Netflix has changed details since, so say the year when you quote a number in an interview.

Sources

Was this helpful?

the question most readers have next

A new season drops at midnight and millions press play together. Where do the video bytes actually come from, and why does Netflix not just rent a CDN?

Netflix runs its own CDN, Open Connect, with appliances inside ISP networks that are filled with content overnight, so most bytes travel a short distance and never cross the open internet. Each title is encoded into a ladder of bitrates, and the player picks a rung every few seconds based on its own measured bandwidth. The control plane, such as login and recommendations, is warmed up before the release.

Where this is used in industry

Netflix: Open Connect appliances are embedded in ISP networks and at internet exchanges and receive nightly content fills, so Netflix traffic is served locally.
Netflix Open Connect
Netflix: Encodes each title with its own bitrate ladder, chosen for that title's content, instead of one fixed ladder for everything.
Netflix TechBlog: Per-Title Encode Optimization
Amazon CloudFront: Its Origin Shield adds a cache layer in front of the origin and merges simultaneous requests for the same video object, for popular on-demand and live content.
AWS docs: Use Amazon CloudFront Origin Shield

Build it yourself: what you would have to solve

Peak traffic is a large share of the internet. Serving it from a few origins is impossible, so content must sit at the edge.
Content Delivery Network (CDN)System Design course26 min
One viewer is on fibre, another on a train. The player must step bitrate up and down per segment without stalling.
Adaptive Bitrate StreamingSystem Design course

Stuck on a step? Every lesson has an AI tutor beside it. Ask it your question and it answers from the lesson you are reading. Try the tutor

Get the free System Design Interview cheat sheet (PDF)

Where this is taught

System Design Masterclass

770 lessons · about 287 hours · 18 free to read

₹1,111 in India by UPI$25 elsewhere by PayPal, once, for life

See the System Design course

You own this course

11FAQ

Frequently asked questions

India: UPI, cards or netbanking through Razorpay. Elsewhere: card or PayPal. Download at once from My packs.

NEWSystem Design Hands-On

Included in the System Design course: real systems, built live

Postgres Sharding: Complete Hands-On (PgBouncer + Citus). 2 h 36 min and 20 chapters: one PostgreSQL loaded until it breaks, then four shards behind a router you write, PgBouncer and Citus, with the 139-page slide deck. The first 15 minutes are free to watch.

Watch the free 15 minutes See System Design Hands-On

Free PDF · 79 pages

25 real tech interview questions, answered in full

Five from each of our AI, GenAI, Inference, Maths for ML and Forward Deployed packs. Each with a diagram and a model answer. Free by email.

free tool

Check these numbers yourself

Put in your own guesses for Netflix and see requests per second, storage, bandwidth and servers, with every sum written out.

Open the Capacity Estimator

A cartoon and an action film need different bitrates for the same quality. A fixed ladder wastes bandwidth on one and blurs the other.

Video EncodingSystem Design course20 min

A 4K stream needs roughly 25 Mbps. You have to know what each link can carry before you promise a resolution.

Bandwidthfree lesson17 min

Also by the same author

214 lessons · about 224 hours · 10 free to read

₹1,333 in India by UPI$35 elsewhere by PayPal, once, for life

See the Generative AI and Agentic AI course

what a reader said

It is really nice and good Please join the course

Gyan Kumar, Analyst, Bangalore · System Design Masterclass · read 3 lessons · all reviews

Netflix System Design: How Playback, Open Connect and Encoding Work

Netflix in its own numbers

Design Netflix, part by part

Why this question is asked

The course goes deeper on the Netflix design

Requirements

Functional requirements

Go deeper

Practice with 770 system design lessons

Get all 200 questions in the System Design and Distributed Systems pack

Before you design Netflix, try the questions it rests on.

Non-functional requirements

Back-of-envelope scale estimates

High-level architecture

Press play, step by step

Two planes on one board

Before you press anything

You press play

Can you watch this here?

Where did you stop?

Which server sends it?

The TV gets URLs, not video

Video comes from inside your provider

Playing as normal

Appliance A goes quiet

Try the next URL

Nobody new is sent to A

Pieces arrive too slowly

Step down a rung

Climb back carefully

What happens when you press play, step by step

Appliances report in

The device asks to play

Playback checks and chooses files

Steering picks appliances

The device streams from the appliance

Core components

Zuul (API gateway)

Playback service

Steering service

Open Connect appliances

Encoding pipeline (Cosmos)

Fill system

Member data: Cassandra and EVCache

Personalisation

Data model

Deep dives

Trade-offs to discuss

What the interviewer asks next

How Netflix actually does it

Lessons to study before this interview

Related system design interview questions

A new season drops at midnight and millions press play together. Where do the video bytes actually come from, and why does Netflix not just rent a CDN?

Where this is used in industry

Build it yourself: what you would have to solve

Frequently asked questions

What is Netflix system design, in short?

What is Netflix Open Connect?

How does Netflix decide which server sends my video?

Why does Netflix copy videos to servers at night?

What is per-title encoding?

What database does Netflix use?

How does Netflix survive an AWS region outage?

How many concurrent streams can Netflix handle?

Are you ready for Design Netflix?

Included in the System Design course: real systems, built live