Tuesday, February 14, 2006

5.1.6 release crammed with goodies

The new 5.1.6 release comes crammed with goodies for MySQL Cluster:

1) It is now possible to replicate between clusters.
2) Non-indexed attributes can now be stored on disk.
3) Disk-stored attributes uses a new page cache, data is stored in
tablespaces and UNDO log records for disk-stored attributes are
stored in LOGFILE GROUP's.
4) The default partitioning scheme has been changed to even distribution
and the previous linear hashing scheme can still be used by using
PARTITION BY LINEAR KEY().
5) Automatic partitioning can still be used, but NDB tables can also be
partitioned by the user. All the partitioning defined in the MySQL manual can
also be used to partition MySQL Cluster tables.
6) With user defined partitioning it is possible to use heterogenous machines
by using NODEGROUP when defining a partition. This defines in which
NODEGROUP a partition is to be stored.
7) ndb_restore can handle restoring backup's created in one cluster with
user-defined partition and later restored in a cluster with different set of
nodegroups.

2 comments:

Anonymous said...

Hi Mikael,

how can I store a non-indexed attributes to disk? I am already running 5.1.6 on a test machine

Thanks

Tom

Anonymous said...

Mikael - in one of your postings on MySQL you mention an approach to query parallelization. Can you forward me the thesis? My email is ben_armstrong@mckinsey.com
Thanks