baradb

baradb is a Key/Value storage engine based on Bitcask model, which is a log-structured storage.

bara: バラ, Japanese name of rose

Design Overview

WRITE:

Append a log record to the current active log file
Update the index

READ:

Find the position in the index by the given key
If the position is found, then find the value of the given key in a log file by the position

References

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

Install

$ go get -u github.com/saint-yellow/baradb

Usage

Basic Operations

// launch a DB engine 
opts := baradb.DefaultDBOptions
db, err := baradb.Launch(opts)
if err != nil {
    panic(err)
}

// put a key/value pair to the DB engine
err = db.Put([]byte("114514"), []byte("1919810"))
if err != nil {
    panic(err)
}

// get vlue of a key from the DB engine  
value, err := db.Get([]byte("114514"))
if err != nil {
    panic(err)
}
fmt.Println(string(value))

// delete a key/value pair from the DB engine
err = db.Delete([]byte("114514"))
if err != nil {
    panic(err)
}

Batch Operations

// initialze a write batch
wb := db.NewWriteBatch(baradb.DefaultWriteBatchOptions)

// put a key/value pair to the write batch 
err = wb.Put([]byte("1919"), []byte("1919"))
if err != nil {
    panic(err)
}

// the DB engine will not store data in the write batch 
// since the write batch is not committed
value, err = db.Get([]byte("1919"))
if err != nil {
    panic(err)
}
fmt.Println(string(value))

// delete a key/value pair in the write batch 
err = wb.Delete([]byte("114514"))
if err != nil {
    panic(err)
}

// commit the write batch 
err = wb.Commit()
if err != nil {
    panic(err)
}

// the DB engine can get data in the write batch 
// that is successfully committed
value, err = db.Get([]byte("1919"))
if err != nil {
    panic(err)
}
fmt.Println(string(value))

Features

Strengths

Low latency per item read or written
High throughput, especially when writing an incoming stream of random items
Ability to handle datasets larger than RAM without degradation
Single seek to retrieve any value
Predictable lookup and insert performance
Easy backup
Batch options which guarantee atomicity, consistency, and durability
Support iterator for forward and backward

Weaknesses

Keys must fit in memory
Startup speed is affected by the amount of data

Applications

Integrate HTTP into the K/V storage engine

Case: baradb-http

Implement Redis on the top of the K/V storage engine

Case: baradb-redis

And... More!

Contributing

PRs accepted.