MySQL readies database clustering improvements

SANTA CLARA, Calif. -- MySQL plans improvements in areas such as tables, replication, and partitioning with upcoming versions of the its database clustering product, a company consultant said at the MySQL Users Conference 2005 event here on Tuesday.

The MySQL Cluster product allows the database to be run across multiple systems. The product features the database as well as a storage engine developed by Ericsson. Having recently added functions such as BLOBs (binary large objects) support to the database, the upcoming 5.0 release, now in beta, removes the 1,500-table limit per cluster. "That limit is gone in 5.0," said John David Duncan, a consultant and sales engineer for MySQL. Version 5.0 is due for general availability this summer.

The 5.1 release of MySQL Cluster, due in 2006, will integrate clustering with replication capabilities, Duncan said. This enables redundant clusters to be maintained in two different geographical limitations.

Partitioning of tables also will be enhanced. "You will be able to have clustered tables or logical tables made of logical tables and many ISAM [Indexed Sequential Access Method] tables, for instance," Duncan said.

To increase database size, indexed columns are to be kept in memory while non-indexed columns are moved out to disk. Variable length record operations also will be improved, and administrators will be able to make more on-line changes to the topology of their clusters with Version 5.1, according to Duncan.

At some point, in a release following Version 5.1, MySQL may support having transactions span both the NDB (network database) and InnoDB transactional storage engine in MySQL. "What will happen when a [transaction] spans InnoDB and NDB is that XA [distributed transactions] support in MySQL will come into play, where the MySQL server is being a transaction manager over a distributed transaction between the two storage engines," Duncan said.