How Database Storage Engines Have Evolved for Internet Scale
SMRTR summary
Database storage engines have evolved from early relational systems using in-place updates to modern distributed SQL and NoSQL databases employing out-of-place updates. This shift was driven by the need to handle heavy write workloads from internet-scale events and IoT sensors. Out-of-place updates offer excellent write performance but sacrifice some read performance.
Modern storage engines typically use log-structured merge-trees (LSM-trees), accumulating writes in memory before flushing sorted runs to disk. This approach amortizes I/O costs and exploits sequential I/O. While LSM-trees excel at write-heavy workloads, they may require examining multiple sorted runs for reads, potentially impacting read performance
SMRTR provides this summary for quick context. The original article belongs to The New Stack.
Read the original article