genesis

genesis is a disk-based, log-structured hash table merge tree KV store originally built upon the Bitcask research paper. Built purely for educational purposes.

Features in the works:

• Serialize/Deserialize header + key, value
• Store data on disk
• Support Put(Key, Value)
• Support Get(Key)
• Support Delete(Key)
• Crash safety (CRC)
• Convert implementation to Log-Structured Merge Tree
  • Swap out keydir with red-black tree memtable
    • Implement red-black tree
  • Write-ahead-logging (WAL)
    • Create WAL file and write to it after Put(k, v) operations
    • Reconstruct memtable with WAL in case of crash
  • Implement SSTables
    • Flush memtable to data file in sorted order
      • Conditional flushing (size threshold)
    • Index file
    • Bloom filter
    • Multiple levels
    • Get(key) operation on tables
    • Compaction
• Generic key/value support (currently limited to strings)
• Make this distributed