Dqlite 1.0 is available, Canonical's distributed version of SQLite

Canonical Company опубликовала significant project release Dqlite 1.0 (Distributed SQLite), which develops a SQLite-compatible embedded SQL engine that supports data replication, automatic failover, and fault tolerance through splitting handlers across multiple nodes. The DBMS is implemented in the form of a C library attached to applications and spreads licensed under Apache 2.0 (the original SQLite is in the public domain). Language bindings available Go.

The library is an add-on to the existing SQLite codebase that adds network protocol support for linking multiple application instances running on different hosts. An application built with Dqlite can function as a self-sufficient failover cluster that does not depend on external DBMS. In practice, Dqlite is used by Canonical in its container management system. Lxd. Among the areas of application of the library, the creation of fault-tolerant IoT devices and handlers in systems is also mentioned.
Edge-computing.

To ensure consistency in data replication, a consensus method is used based on the algorithm Raft, which is used by projects such as etcd, RethinkDB, CockroachDB, and OpenDaylight. Dqlite uses its own asynchronous implementation C-raftwritten in C language. Ready-made libraries are used to multiplex the processing of connections and organize the launch of coroutines. libuv и libco.

Compared to a similar project rqlite, Dqlite provides full transaction support, can link to any C project, allows you to use the time() function, and uses frame-based replication instead of SQL translation-based replication.

Features of Dqlite:

• Performing all disk and network operations in asynchronous mode;
• Availability of a test set to confirm the correctness of the data;
• Low memory consumption and efficient network communication;
• Permanent storage of the database and transaction log on disk (with the possibility of caching in memory);
• Fast recovery after failures;
• A stable Go CLI client that can be used to initialize the database, set up replication, and connect/disconnect nodes;
• Support for ARM, X86, POWER and IBM Z architectures;
• The implementation of the Raft algorithm is optimized to minimize delays when committing transactions.

Source: opennet.ru