liveBook · Manning

Write amplification

As explained in ScyllaDB in Action

Write amplification is the amount of data written by a storage system relative to the original data size.

Write amplification measures how much data a storage system writes compared with the original data size. Compaction strategies can increase it when they rewrite existing data. section 8.4.1

Compaction strategy changes the expected amplification and its operational trade-offs:

  • STCS typically produces about 4× amplification and is the default choice. section 8.4.1
  • LCS typically produces about 10× amplification because compaction may rewrite an entire level, consuming more disk bandwidth. section 8.4.1
  • Choose LCS when compacting space is a concern, while recognizing its higher bandwidth demand. section 8.4.1

Comparing compaction strategies

If an application initially writes 1 GB of data, roughly 4 GB may be written with STCS and roughly 10 GB with LCS over subsequent compactions. section 8.4.1

How ScyllaDB in Action covers it

How deep it goes
Defines the metric and compares typical STCS and LCS values, but stops short of calculating workload-specific amplification.
Examples
Compares approximately 4× amplification for STCS with approximately 10× for LCS.
What you can do after
Evaluate compaction strategies by balancing write bandwidth against compacting disk-space requirements.
How it connects
Connects compaction, SSTables, STCS, LCS, disk space, and disk bandwidth.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of ScyllaDB in Action. The section links point to the parts of the book it draws on.