Abstract: Adaptive database compaction technology automatically identifies cost-effective opportunities for database compaction. The adaptive compaction technology may maintain a baseline sleeve representing a performance indicator's normal range, track the current performance indicator values, and initiate compaction of a database when a compaction trigger based on at least the performance indicator occurs. The performance indicator may be a ratio of logical size to physical size, and may be based on samples from a proper subset of the database. Kernel overhead may be recognized. A low-fragmentation secondary replica may be selected, compacted, and promoted to replace the prior primary replica. Secure cloud blob storage may be used. A compaction decision may allow, delay, or even prevent compaction after the trigger is met. An automatic balance between computational costs of compaction and costs of continued database use without compaction is kept, and may be tunable by an administrator.

Abstract: Adaptive database compaction technology automatically identifies cost-effective opportunities for database compaction. The adaptive compaction technology may maintain a baseline sleeve representing a performance indicator's normal range, track the current performance indicator values, and initiate compaction of a database when a compaction trigger based on at least the performance indicator occurs. The performance indicator may be a ratio of logical size to physical size, and may be based on samples from a proper subset of the database. Kernel overhead may be recognized. A low-fragmentation secondary replica may be selected, compacted, and promoted to replace the prior primary replica. Secure cloud blob storage may be used. A compaction decision may allow, delay, or even prevent compaction after the trigger is met. An automatic balance between computational costs of compaction and costs of continued database use without compaction is kept, and may be tunable by an administrator.