immudb v1.9.3, 60% improved cache efficiency, significantly reduced memory consumption
Changes
- Replaced LRU (least recently used) caching policy with SIEVE algorithm, resulting in improvement of caching efficiency by up to 60%;
- Significantly reduced memory consumption of indexes by using pooling and by bounding the byte size of the cache, by up to 48%;
- Support compaction of indexes when S3 storage is enabled;
- Way more instrumentation reported by immuadmin database and status subcommands;
- Relaxed password hassling of user;
- Significantly improved coverage of SQL GROUP BY and ORDER BY clauses;
- Indexes are no more required on the GROUP BY/ORDER BY fields:
- Arbitrary number of fields are now supported;
- Index selection has been optimised to reduce query times;
- Support streaming in SQLQuery/TxSQL rpcs to allow fetching arbitrary number of query results from:
- client SDKs;
- PG compatible clients;
- Golang database/sql std package;
Fixes
- Fix off by one error when LIMIT clause is used;
- Add missing constraint checks that prevent indexer from getting stuck;
- Prevent replication between different immudb versions;
Full Changelog: v1.9DOM.2...v1.9.3
Download: https://github.com/codenotary/immudb/tree/v1.9.3