Cache Synchronization in High‑Concurrency Environments: Problems and JD's CDC‑Based Solution

Existing Technologies – Heterogeneous data systems often use Cache‑Aside, Read‑Through/Write‑Through, Write‑Back, and Write‑Around patterns to keep cache and database synchronized. Cache‑Aside reads from cache first and falls back to the database; writes either update the DB then the cache or delete the cache entry. Read‑Through/Write‑Through makes the cache the primary data source, delegating DB reads/writes to it. Write‑Back writes only to the cache and updates the DB asynchronously. Write‑Around writes directly to the DB and updates the cache only on a miss.

Drawbacks of Existing Techniques – Cache‑Aside can produce stale data under high concurrency because multiple update requests may reorder DB and cache updates, leading to dirty cache entries. Read‑Through/Write‑Through, Write‑Back, and Write‑Around rely heavily on the cache; if the cache fails, captured data may be permanently lost. (Image illustrating Cache‑Aside inconsistency)

JD Solution Overview – JD proposes a CDC (Change Data Capture)‑based approach: CDC captures database changes, which are published to a message‑queue (MQ) for delayed consumption. This decouples cache updates from immediate DB writes and removes strong cache dependency; when the cache is unavailable or empty, the system directly loads data from the DB.

Detailed Implementation – Reads continue to use Cache‑Aside: the application checks the cache first, falls back to the DB if needed, and then populates the cache. Writes update the relational database (Oracle/MySQL/SQL Server). CDC captures the change and stores it in a binlog‑style change table. The change is sent to MQ; consumers (cache‑updating services) consume the messages after a configurable delay, applying version‑controlled updates to the cache. Each table includes a version field; only messages with the next sequential version are applied, ensuring ordered, consistent cache updates even under high concurrency. (Images showing read flow and write flow diagrams)

Benefits – The combined Cache‑Aside + CDC + MQ + delayed consumption + versioning approach guarantees eventual consistency between cache and database in high‑concurrency scenarios, improves throughput by asynchronously decoupling components, and reduces response latency while preventing dirty cache writes.