Consistent hashing implementation for distributed systems. Routes keys to nodes with minimal redistribution when the cluster changes.

Traditional modulo hashing (hash(key) % N) reassigns most keys when nodes are added or removed. Consistent hashing with virtual nodes minimizes this to ~1/N keys, making it ideal for:

• Cache sharding (Redis, Memcached)
• Database partitioning
• Load balancing with sticky sessions
• Distributed storage replication

Pure algorithm implementation with no opinions about your infrastructure. Compose it with whatever caching, database, or networking library you use.

Use hash-orbit when you need consistent key-to-node routing in a distributed system and want a pure algorithm without infrastructure opinions.

Choose hash-orbit over hashring when:

• You want a focused, dependency-free implementation (hashring pulls in 5+ dependencies)
• You need getN() for replication (route a key to multiple nodes)
• You need serialization/deserialization (toJSON/fromJSON) for ring state transfer

Choose hashring instead when:

• You need weighted nodes (hash-orbit treats all nodes equally)
• You need built-in server failover and connection management

Choose a managed solution (Redis Cluster, DynamoDB) when:

• You don't want to manage the routing layer yourself
• You need automatic rebalancing and failure detection

npm install hash-orbit

import { HashOrbit } from 'hash-orbit';

const ring = new HashOrbit({ replicas: 150 });

// Add nodes
ring.add('cache-1');
ring.add('cache-2');
ring.add('cache-3');

// Route a key to a node (deterministic)
const node = ring.get('user:123'); // => 'cache-2'

// Get multiple nodes for replication
const nodes = ring.getN('user:123', 2); // => ['cache-2', 'cache-1']

// Remove a node (only ~1/3 of keys will move)
ring.remove('cache-2');
ring.get('user:123'); // => 'cache-1' (new assignment)

new HashOrbit(options?: { replicas?: number })

Creates a hash ring. replicas controls the number of virtual nodes per physical node (default: 150). Higher values improve distribution but use more memory.

add(node: string): void Add a node to the ring.

remove(node: string): void Remove a node from the ring.

get(key: string): string | undefined Get the node responsible for a key. Returns undefined if ring is empty.

getN(key: string, count: number): string[] Get N unique nodes for a key (useful for replication).

toJSON(): { nodes: string[], replicas: number } Serialize ring state for persistence or transfer between processes.

static fromJSON(json): HashOrbit Restore a ring from serialized state.

size: number - Number of nodes in the ring nodes: string[] - List of all nodes

const ring = new HashOrbit();
ring.add('redis-1:6379');
ring.add('redis-2:6379');
ring.add('redis-3:6379');

async function get(key: string) {
  const server = ring.get(key);
  return redisClients[server].get(key);
}

async function set(key: string, value: string) {
  const server = ring.get(key);
  return redisClients[server].set(key, value);
}

const ring = new HashOrbit({ replicas: 200 });
ring.add('storage-1');
ring.add('storage-2');
ring.add('storage-3');

async function writeWithReplication(key: string, data: Buffer) {
  const targets = ring.getN(key, 2); // Write to 2 nodes
  await Promise.all(targets.map((node) => storage[node].write(key, data)));
}

// Save ring state
const state = ring.toJSON();
await fs.writeFile('ring.json', JSON.stringify(state));

// Restore later
const restored = HashOrbit.fromJSON(JSON.parse(await fs.readFile('ring.json')));
restored.get('user:123'); // Routes to same node as original

Virtual Nodes: Each physical node gets replicas positions on the ring (default 150). This ensures even distribution and minimal disruption when nodes change.

Binary Search: Keys are hashed to a position, then binary search finds the next node clockwise on the ring in O(log n) time.

Minimal Redistribution: When adding/removing nodes, only ~1/N keys need to move (where N is the number of nodes).

Memory: ~64 bytes per virtual node position (30KB for 450 positions).

Fully typed with strict mode. Ships with .d.ts files.

import { HashOrbit, hash32 } from 'hash-orbit';

const ring: HashOrbit = new HashOrbit({ replicas: 100 });
const node: string | undefined = ring.get('key');
const nodes: string[] = ring.getN('key', 3);

npm install      # Install dependencies
npm test         # Run tests (fast: ~300ms)
npm run coverage # Generate coverage report
npm run build    # Build for production

100% test coverage. All tests run in <500ms.

See examples/ for complete working examples:

• Cache replication with failover
• Database sharding
• Session-based load balancing

MIT © 1mb-dev

• Consistent Hashing and Random Trees (Karger et al.)
• Amazon Dynamo Paper