TLDR: The RUM Conjecture - Storage System Trade-offs
Date: 2020-05-31 Source: https://arpitbhayani.me/blogs/rum
Overview
Explore the RUM Conjecture - How Read, Update, and Memory overheads impact storage system design and performance trade-offs. Read Overhead occur when the storage engine performs reads on auxiliary data to fetch the required intended main data.
Key Points
- Read Overhead: Read Overhead occur when the storage engine performs reads on auxiliary data to fetch the required intended main data.
- Read Optimised: Read Optimised storage systems offer very low read overhead but require some extra auxiliary space to gain necessary performance that again comes at a cost of updates required to keep auxiliary data in sync with main data which adds to update overheads.
- Update Optimised: Update Optimised storage systems offer very low Update Overhead by usually using an auxiliary space holding differential data (delta) and flushing them over main data in a bulk operation.