My Algorithm Summary
  • Introduction
  • Data Structure
    • Linked List
    • Stack
      • Monotone Stack
        • 42 Trapping Rain Water
        • 84 Largest Rectangle in Histogram
        • 85 Maximal Rectangle
        • 255 Verify Preorder Sequence in Binary Search Tree
        • 316 Remove Duplicate Characters
        • 402 Remove K Digits
        • 456 132 Pattern
        • 496 Next Greater Element I
        • 503 Next Greater Element II
      • 20 Valid Parentheses
      • 71 Simplify Path
      • 150 Evaluate Reverse Polish Notation
      • 155 Min Stack
      • 173 Binary Search Tree Iterator
      • 224 Basic Calculator
      • 227 Basic Calculator II
      • 232 Implement Queue using Stacks
      • 341 Flatten Nested List Iterator
      • 394 Decode String
      • 439 Ternary Expression Parser
      • 636 Exclusive Time of Functions
    • Heap
    • Trie
    • Segment Tree
    • Tree
      • 94 Binary Tree Inorder Traversal
      • 104 Maximum Depth of Binary Tree
      • 144 Binary Tree Preorder Traversal
      • 145 Binary Tree Postorder Traversal
      • 199 Binary Tree Right Side View
      • 226 Invert Binary Tree
      • 272 Closest Binary Search Tree Value II
      • 508 Most Frequent Subtree Sum
      • 513 Find Bottom Left Tree Value
      • 515 Find Largest Value in Each Tree Row
      • 617 Merge Two Binary Trees
      • 637 Average of Levels in Binary Tree
      • 653 Two Sum IV - Input is a BST
      • 654 Maximum Binary Tree
      • 669 Trim a Binary Search Tree
      • 666 Path Sum IV
      • 230 Kth Smallest Element in a BST
      • 250 Count Univalue Subtrees
      • 538 Convert BST to Greater Tree
      • 404 Sum of Left Leaves
      • 582 Kill Process
      • 112 Path Sum
      • 108 Convert Sorted Array to Binary Search Tree
      • 111 Minimum Depth of Binary Tree
      • 501 Find Mode in Binary Search Tree
      • 102 Binary Tree Level Order Traversal
      • 107 Binary Tree Level Order Traversal II
      • 103 Binary Tree Zigzag Level Order Traversal
      • 113 Path Sum II
      • 437 Path Sum III
      • 99 Recover Binary Search Tree
      • 687 Longest Univalue Path
      • 285 Inorder Successor in BST
      • 101 Symmetric Tree
      • 129 Sum Root to Leaf Numbers
      • 298 Binary Tree Longest Consecutive Sequence
      • 270 Closest Binary Search Tree Value
      • 549 Binary Tree Longest Consecutive Sequence II
      • 98 Validate Binary Search Tree
      • 652 Find Duplicate Subtrees
      • 314 Binary Tree Vertical Order Traversal
      • 333 Largest BST Subtree
      • 563 Binary Tree Tilt
      • 110 Balanced Binary Tree
    • Graph
      • Detect Cycle
  • Algorithms
    • Union Find
      • 695 Max Area of Island
      • 684 Redundant Connection
    • Binary Search
    • Topological Sorting
    • Breadth-First Search
      • 694 Number of Distinct Islands
    • Depth-First Search
    • Two Pointers
    • Sorting
    • Backtacking
    • Dynamic Programming
      • Interval DP
        • Matrix Chain Multiplication
        • Merge Stone
      • KnapSack Problem
        • 0-1 KnapSack
        • Unbounded KnapSack
      • Longest Increasing Subsequence
      • Longest Common Subsequence
    • Reservior Sampling
    • Bipartite Graph
      • Check Bipartite Graph
      • Maximal Matching - Hungarian Algorithm
    • String Pattern Matching
      • KMP Algorithm
      • Rabin Karp Algorithm
  • System Design
    • Consistent Hashing
    • Bloom Filter
    • Caching
      • LRU
      • LFU
    • Mini Twitter
    • Tiny Url
Powered by GitBook
On this page
  • 1. Introduction
  • 2.Simple Version
  • 3. Better version

Was this helpful?

  1. System Design

Consistent Hashing

PreviousSystem DesignNextBloom Filter

Last updated 5 years ago

Was this helpful?

1. Introduction

Consistent Hashing is used for horizontal sharding, check for mode details

2.Simple Version

  • Idea: Check the description in LintCode

  • Implementation:

    public List<List<Integer>> consistentHashing(int n) {
        List<Integer> machine = new ArrayList<>();
        machine.add(0);
        machine.add(359);
        machine.add(1);
        res.add(machine);

        for (int i = 1; i < n; i++) {
            List<Integer> nextMachine = new ArrayList<>();
            int index = 0;

            // Look for the largest interval
            for (int j = 1; j < i; j++) {
                int nextInterval = res.get(j).get(1) - res.get(j).get(0) + 1;
                int curInterval = res.get(index).get(1) - res.get(index).get(0) + 1;

                if (curInterval < nextInterval) {
                    index = j;
                }
            }

            // Divide the largest interval into havles
            int start = res.get(index).get(0);
            int end = res.get(index).get(1);
            res.get(index).set(1, (start + end)/2);

            // The other half will be assigned to the new machine
            nextMachine.add((start + end)/2 + 1);
            nextMachine.add(end);
            nextMachine.add(i + 1);
            res.add(nextMachine);
        }
        return res;
    }

3. Better version

  • Idea: Check the description in LintCode

import java.util.*;

public class ConsistentHashing {
    // Number of interval
    private int n;
    // Number of virtual nodes for each machine
    private int k;
    // Set to store k distinct virtual nodes for a new mahine 
    private Set<Integer> ids = null;
    // Map to store a list of virtual nodes for a machine
    private Map<Integer, List<Integer>> machines = null;

    public ConsistentHashing(int n, int k) {
        this.n = n;
        this.k = k;
        ids = new HashSet<>();
        machines = new HashMap<>();
    }

    public List<Integer> addMachine(int machineId) {
        Random rand = new Random();
        List<Integer> virtualNodes = new ArrayList<>();
        int nodeId = 0;

        //Randomly create k distinct virtual nodes
        for (int i = 0; i < k; i++) {
            nodeId = rand.nextInt(n);

            while(ids.contains(nodeId)) {
                nodeId = rand.nextInt(n);
            }
            ids.add(nodeId);
            virtualNodes.add(nodeId);
        }

        Collections.sort(virtualNodes);
        machines.put(machineId, virtualNodes);

        System.out.print("virtual nodes for machine " + machineId + " : ");
        for (int node : virtualNodes) {
             System.out.print(node + " ");
        }
        System.out.println();
        return virtualNodes;
    }

    public int getMachineIdByHashCode(int hashCode) {
        // Initialize distance with max possible value (the distance on the circle will be no more than n)
        int distance = n + 1;
        int diff = 0;
        int machineId = 0;

        for (Map.Entry<Integer, List<Integer>> entry : machines.entrySet()) {
            int key = entry.getKey();
            List<Integer> virtualNodes = entry.getValue();

            for (int nodeId : virtualNodes) {
                diff = nodeId - hashCode;

                // Offset diff by n since we will route data to closest machine clockwise
                if (diff < 0) {
                    diff += n;
                }

                // Update the distance and the closet clockwise machine in that distance
                if (diff < distance) {
                    distance = diff;
                    machineId = key;
                }
            }
        }
        System.out.println("Data is routed to machine " + machineId);
        // Return the closest clockwise machine given the hashCode
        return machineId;
    }


    public static void main(String[] args) {
        ConsistentHashing hash = new ConsistentHashing(100, 3);
        hash.addMachine(1);
        hash.addMachine(2);
        hash.addMachine(3);
        hash.getMachineIdByHashCode(60);
        hash.getMachineIdByHashCode(95);
    }
}

Implementation:

here
Lintcode Consistent Hashing
Lintcode Consistent Hashing II