We use analytics cookies.We use cookies to analyze site traffic and improve your experience.
Loading...
System Design and AI Engineering, by Roni Das. 984 interactive lessons.
Start LearningFoundation · Core Fundamentals
A decentralized architecture where each node acts as both client and server. No central authority. Used by BitTorrent, blockchain, and WebRTC.
In short
Peer-to-peer (P2P) is a decentralized network architecture where every node acts as both a client and a server, sharing resources like files, bandwidth, or compute directly with other nodes instead of routing everything through a central server. Each peer can request data from others and serve data to others at the same time, which removes the single point of control found in traditional client-server systems.
In short
Peer-to-peer (P2P) is a decentralized network architecture where every node acts as both a client and a server, sharing resources like files, bandwidth, or compute directly with other nodes instead of routing everything through a central server. Each peer can request data from others and serve data to others at the same time, which removes the single point of control found in traditional client-server systems.
Learn it in the course
This page is the free part.
₹1,111 in India$25 everywhere elseonce, for the whole course
The System Design course covers Peer-to-Peer across a run of lessons, not one page. These 4 alone are about 119 minutes of step-by-step reading, 1 of them with code you run in the browser, all with a quiz.
Real-time peer-to-peer communication in the browser, the technology behind video calls, screen sharing, and live streaming.
Maximum data transfer capacity of a network connection.
What is a lakehouse, and what breaks if a table is just a folder of Parquet files? Warehouse, lake and lakehouse in plain words first. Then a lab on 7,052,769 real NYC taxi trips: while a writer added February as 8 files, a reader that listed the folder saw every half-written state, 9 different row counts; a Delta Lake reader saw only 3,475,226 rows or 7,052,769, never anything between. During an overwrite, the folder reader counted 6,806,334 trips for one month that held 3,331,108. In a real race, 5 runs per case, the folder reader saw a half-written table 7 to 9 times per run and failed 955 to 5,003 reads; the Delta reader saw 0 half-written tables and 0 failures.
Defining servers, networks, databases, and cloud resources in declarative code files instead of clicking through consoles.
all part of
System Design Masterclass
770 lessons · about 287 hours · 18 free to read
₹1,111in India, by UPI
$25everywhere else, by PayPal
It is really nice and good Please join the course
You own this course
Continue with WebRTCpay once,
yours for life
In a client-server system, clients only ask for things and servers only answer. A browser talks to a web server, a phone app talks to an API. The server is special: it holds the data, does the work, and everyone depends on it.
Peer-to-peer flips that. There is no special machine. Every node, called a peer, runs the same software and plays both roles. When you download a movie over BitTorrent, your machine is downloading pieces from dozens of strangers and uploading pieces to dozens of others at the same instant. You are a client and a server in the same breath.
Because no single node is in charge, P2P systems are decentralized. There is no central authority that must stay online for the network to work. If half the peers disappear, the rest keep going. This is the property that makes P2P resilient and hard to shut down, and it is also what makes it harder to coordinate and secure.
The hard problem in P2P is discovery: how does a peer find the other peers that have what it wants? There are two common answers. Some systems use a structured approach based on a distributed hash table (DHT), where every piece of content has a key and the network agrees on which peers are responsible for which range of keys. Kademlia, the DHT used by BitTorrent and IPFS, lets any peer locate the owner of a key in about log(N) hops, so finding data among a million peers takes roughly 20 lookups.
Other systems use a tracker or a set of bootstrap nodes just to help newcomers join, then run fully peer-to-peer after that. A BitTorrent tracker hands you an initial list of peers, but the actual file transfer never touches the tracker.
Once peers know each other, they exchange data directly. They split content into chunks so a file can be assembled from many sources in parallel, and they verify each chunk with a cryptographic hash so a malicious peer cannot feed you garbage. Blockchains add a consensus layer on top, where peers must agree on a single shared history through rules like proof of work or proof of stake, because money is at stake and disagreement cannot be tolerated.
P2P shines when you want to scale capacity with the crowd. In client-server, every new user adds load to your servers. In P2P, every new peer brings its own upload bandwidth and storage, so the network often gets faster as it grows. This is why P2P is a natural fit for large file distribution, live video at scale, and systems that need to survive without a trusted operator.
The costs are real. Discovery and consistency are much harder without a central coordinator, so engineering effort goes up. Peers come and go constantly, a behavior called churn, so the system must constantly heal. Security is trickier because you are trusting unknown machines, which is why hashing, encryption, and consensus are not optional. Latency can be unpredictable because the peer holding your data might be on the other side of the planet on a slow connection.
A practical middle ground is the hybrid model. Spotify originally streamed music partly over P2P to cut bandwidth costs while keeping central servers for login and search. WebRTC sets up direct peer connections for video calls but still uses central STUN and TURN servers to punch through firewalls. Pure P2P is rare in production; most real systems keep a thin central layer for the parts that genuinely need coordination.
Say a Linux distribution releases a 4 GB ISO and a million people want it on launch day. Serving that from one set of servers would cost a fortune in bandwidth and likely fall over.
With BitTorrent, the publisher creates a small torrent file that lists the chunk hashes and a tracker URL. The publisher starts as the first seeder, a peer that already has the whole file. As new peers join, they download chunks from whoever has them, and the moment a peer finishes a chunk it starts uploading that chunk to others. A peer with the complete file is a seeder; one still downloading is a leecher.
The result is that the popular ISO actually downloads faster, because a hot file means more peers sharing it. The original publisher might only ever upload one or two full copies, and the crowd does the rest. That inversion of cost, where popularity reduces load on the origin instead of increasing it, is the core reason P2P keeps getting reinvented for new problems.
BitTorrent
Splits files into hash-verified chunks so peers download from and upload to each other in parallel, making popular files faster as more peers join.
Bitcoin and Ethereum
Run as P2P networks where thousands of nodes each hold the full ledger and reach consensus on a shared history with no central server.
WebRTC
Powers browser video calls by setting up direct peer connections, using central STUN/TURN servers only to traverse NATs and firewalls.
IPFS
A content-addressed P2P file system that locates data by its hash through a Kademlia distributed hash table instead of by location or URL.
In client-server, dedicated servers hold the data and clients only request it, so everything depends on the server staying up. In peer-to-peer, every node is both client and server, sharing resources directly, so there is no central machine the network depends on.
They overlap but are not identical. Decentralization means no single point of control, and P2P is one common way to achieve it. You can have decentralized systems that are not strictly P2P, and some P2P systems still rely on a few central bootstrap or tracker nodes to help peers find each other.
Most modern P2P systems use a distributed hash table such as Kademlia, where content is mapped to keys and the network collectively decides which peers are responsible for which keys, letting any peer locate data in about log(N) hops. Others use lightweight trackers or bootstrap nodes only for the initial join.
It can be, but security has to be built in because you are connecting to untrusted machines. P2P systems verify every chunk of data with cryptographic hashes, encrypt connections, and blockchains add consensus rules so no single peer can rewrite history. Without these protections, a malicious peer could serve corrupted or fake data.
Each peer that joins brings its own upload bandwidth, so a popular file has many peers sharing chunks in parallel. Unlike a single server that slows down under load, P2P capacity grows with the crowd, so a hot torrent can download faster than a rare one.
Client-Server Model
The foundational architecture of the web: clients (browsers, apps) send requests and servers process them and return responses. Every web interaction follows this pattern.
Gossip Protocol
A peer-to-peer communication protocol where nodes share information with random neighbors, spreading it like gossip. Used for cluster membership and failure detection.
Distributed Lock
A lock that coordinates access to a shared resource across multiple machines. Implemented via Redis (Redlock), ZooKeeper, or etcd. Much harder than local locks.
Latency
The time delay between sending a request and getting a response. Amazon found every 100ms of extra latency costs 1% in sales.
Throughput
The number of operations a system can handle per unit of time. Think of it as how many cars a highway can move per hour.
Bandwidth
The maximum amount of data that can be transferred over a network in a given time. It's the width of the pipe, not how fast the water flows.
This page explains the idea. The course covers it step by step, with diagrams. It is part of the System Design Masterclass course, which has 770 lessons. 18 of them are free to read. You pay once and keep it for life, with no subscription.
See the System Design Masterclass coursecourse 1
From absolute beginner to principal engineer, drawn step by step.
Asked about Peer-to-Peer in an interview?
Each question has why it is asked, a full answer, what the interviewer listens for, the follow-ups, the common mistakes and a diagram (268 diagrams in all). A PDF you keep, junior to staff level.
India: UPI, cards or netbanking through Razorpay. Elsewhere: card or PayPal. Download at once from My packs.
