/bitcask-rb

📁 Build your own fast, persistent KV store(Based on bitcask paper)

Primary LanguageRubyMIT LicenseMIT

bitcask-rb: A Log-Structured Hash Table for Fast Key/Value Data

Ruby BDD Ruby Style Guide CircleCI codecov DeepSource

Fast, Persistent key/value store based on bitcask paper written in Ruby. An attempt to understand and build our persistent key/value store capable of storing data enormous than the RAM. This, in any way, is not intended for production. For simplicity, implementation has ignored a few specifications from the paper.

Prerequisites

  • Ruby
  • bundler

Setup

bundle install

Usage

db_store = Bitcask::DiskStore.new('bitcask.db')

# Setting values in store using put or hash style
db_store.put("Anime", "One Piece")
db_store["Luffy"] = "Straw Hat"
db_store[2020] = "Leap Year"
db_store.put("pie", 3.1415)

# Getting values from store using get or hash style
db_store["Anime"]
db_store.get("Luffy")
db_store["pie"]
db_store.get(2020)

# Listing keys
db_store.keys

# Size of the store
db_store.store

Tests

bundle exec rspec

Advantages

  • Simple yet powerful
  • Easy to use
  • Store data enormous than RAM
  • Data file is OS agnostic
  • Portable
  • Faster read/writes
  • Minimal dependecies

Features

Feature Support
Persisted in disk ✅
Get API ✅
Put API ✅
Int, Float and String for k/v ✅
CRC ✅
Directory Support ❌
Delete API ❌
Files Merge and LSM Trees ❌

Benchmarks

CPU Cycles

Benchmarked with value_size of 78 bytes
                                                         user     system      total        real
DiskStore.put : 10k records                          0.026963   0.018983   0.045946 (  0.046078)
DiskStore.get : 10k records                          0.031211   0.009727   0.040938 (  0.041154)
DiskStore.put : 100k records                         0.367399   0.196536   0.563935 (  0.566659)
DiskStore.get : 100k records                         0.313556   0.102338   0.415894 (  0.416280)
DiskStore.put : 1M records                           4.649307   2.209731   6.859038 (  6.887667)
DiskStore.get : 1M records                           3.357120   1.047637   4.404757 (  4.409732)
avg_put:                                             0.000005   0.000002   0.000007 (  0.000007)
avg_get:                                             0.000003   0.000001   0.000004 (  0.000004)

Iterations Per Second

Warming up --------------------------------------
DiskStore.put : 100 records with data size: 702 Bytes
                       149.000  i/100ms
DiskStore.get : 100 records, value size: 702 Bytes
                         2.389k i/100ms
DiskStore.put : 100 records, value size: 6 Kb
                        22.000  i/100ms
DiskStore.get : 100 records, value size: 6 Kb
                         2.098k i/100ms
DiskStore.put : 100 records, value size: 660 Kb
                         1.000  i/100ms
DiskStore.get : 100 records, value size: 660 Kb
                       148.000  i/100ms
Calculating -------------------------------------
DiskStore.put : 100 records with data size: 702 Bytes
                          1.552k (±15.7%) i/s -      7.450k in   5.008519s
DiskStore.get : 100 records, value size: 702 Bytes
                         24.100k (± 3.7%) i/s -    121.839k in   5.062195s
DiskStore.put : 100 records, value size: 6 Kb
                        655.280  (±53.4%) i/s -      1.716k in   5.272456s
DiskStore.get : 100 records, value size: 6 Kb
                         19.883k (± 7.1%) i/s -    100.704k in   5.090910s
DiskStore.put : 100 records, value size: 660 Kb
                          4.479  (± 0.0%) i/s -     23.000  in   5.166651s
DiskStore.get : 100 records, value size: 660 Kb
                          3.286k (±39.4%) i/s -      8.140k in   5.272424s

Contributing

If you wish to contribute in any way, please fork the repo and raise a PR.

CircleCI