Abase2 Multi‑Region Deployment and CRDT Support: Architecture, Challenges, and Redis Command Compatibility

The presentation introduces Abase2's practice of supporting CRDT in cross‑region deployments, first outlining the challenges of multi‑region NoSQL databases such as latency, disaster recovery, and resource constraints, and then describing the CRDT‑based solutions.

It reviews the evolution from Abase 1.0’s dual‑cluster asynchronous log replication—prone to high sync latency, consistency issues, and extra costs—to the modern Abase2 architecture that integrates a native multi‑active design with CRDT to eliminate conflicts and achieve RPC‑level sync delays.

Abase2’s architecture consists of a Global Abase Cluster divided into Regions, AZs, and PODs, with DataNodes layered into framework, consistency protocol, and engine tiers; the engine includes a log layer, KV layer, and cache layer that together enable fast, conflict‑free reads and writes.

The CRDT fundamentals are explained, covering state‑based and operation‑based CRDTs, their mathematical properties (commutativity, associativity, idempotence), and common types such as Register, Counter, and Set, with a focus on Counter‑type CRDT for distributed counters.

Implementation details show how mixed logical clocks provide globally unique, causally ordered timestamps, allowing the system to merge operations without full log replay, and how the cache updates handle concurrent operations efficiently.

Redis command support is categorized into unordered commands (String, Set, Hash) and ordered commands (Sorted Set, List). For unordered commands, the cache stores both the merged value and metadata to resolve conflicts; for ordered commands, operations like ZremRangeByRank are transformed into member‑based deletions to maintain consistency.

Finally, the article summarizes Abase2’s advantages—low sync latency, strong eventual consistency, reduced cost—and outlines ongoing optimizations such as improving cache hit rates, pruning cache‑miss paths, and refining RCU evaluation for complex data structures.