Recently, we’ve started to dig into the internals of ClustrixDB, specifically how ClustrixDB accomplishes horizontal scaling of both writes and reads without sharding. Next, we dug into the details of the multi-patented ClustrixDB Rebalancer. This time, we will discuss the ClustrixDB Query Evaluation Model.
ClustrixDB is a MySQL-compatible distributed RDBMS that provides linear scale out of both writes and reads, while maintaining relational semantics, including ACID transactionality and referential integrity. Typically, MySQL workloads are only able to scale out both writes and reads if sharding is used. Sharding is the strategy of partitioning your MySQL application workload across multiple separate MySQL database servers, allowing queries and data CRUD operations to fan-out. This means multiple separate MySQL physical servers must be deployed, the workload data needs to be partitioned across them, and the application needs to be rewritten to manage any ACID transactionality needed between those servers. ClustrixDB is able to provide a similar linear scale out of sharding, but the data distribution is automatically handled via the multi-patented ClustrixDB Rebalancer behind the scenes. The application doesn’t require rewrites and sees only a single logical RDBMS while all cross-node ACID transactionality is handled automatically.
No comments:
Post a Comment