Oracle 5.0 Reference Manual Download

Page: 1542 / 3232

Known Limitations of MySQL Cluster

1522

In this section, we discuss changes in the implementation of MySQL Cluster in MySQL 5.0 as
compared to MySQL 4.1.

There are relatively few changes between the NDB storage engine implementations in MySQL 4.1 and
in 5.0, so the upgrade path should be relatively quick and painless.

All significantly new features being developed for MySQL Cluster are going into the MySQL Cluster
NDB 7.x trees. For information on changes in the Cluster implementations in MySQL versions 5.1 and
later, see

Section 17.1.4, “MySQL Cluster Development History”

17.1.4.1. MySQL Cluster Developement in MySQL 5.0

MySQL Cluster in MySQL 5.0 contains a number of features added since MySQL 4.1 that are likely to
be of interest:

• Condition pushdown.

Consider the following query:

SELECT * FROM t1 WHERE non_indexed_attribute = 1;

This query uses a full table scan and the condition is evaluated in the cluster's data nodes. Thus, it
is not necessary to send the records across the network for evaluation. (That is, function transport
is used, rather than data transport.) Please note that this feature is currently disabled by default
(pending more thorough testing), but it should work in most cases. This feature can be enabled
through the use of the

SET engine_condition_pushdown = On

statement. Alternatively, you

can run

mysqld

with the this feature enabled by starting the MySQL server with the

--engine-

condition-pushdown

option.

A major benefit of this change is that queries can be executed in parallel. This means that queries
against nonindexed columns can run faster than previously by a factor of as much as 5 to 10 times,
times the number of data nodes, because multiple CPUs can work on the query in parallel.

You can use

EXPLAIN

to determine when condition pushdown is being used. See

Section 13.8.2,

“

EXPLAIN

Syntax”

• Decreased

IndexMemory

Usage: In MySQL 5.0, each record consumes approximately 25 bytes

of index memory, and every unique index uses 25 bytes per record of index memory (in addition to
some data memory because these are stored in a separate table). This is because the primary key is
not stored in the index memory anymore.

• Query Cache Enabled for MySQL Cluster: See

Section 8.6.3, “The MySQL Query Cache”

, for

information on configuring and using the query cache.

• New optimizations.

One optimization that merits particular attention is that a batched read

interface is now used in some queries. For example, consider the following query:

SELECT * FROM t1 WHERE

primary_key

IN (1,2,3,4,5,6,7,8,9,10);

This query will be executed 2 to 3 times more quickly than in previous MySQL Cluster versions due
to the fact that all 10 key lookups are sent in a single batch rather than one at a time.

• Limit On Number of Metadata Objects: Beginning with MySQL 5.0.6, each Cluster database

may contain a maximum of 20320 metadata objects—this includes database tables, system tables,
indexes and

BLOB

values. (Previously, this number was 1600.)

17.1.5. Known Limitations of MySQL Cluster

In the sections that follow, we discuss known limitations of MySQL Cluster in MySQL 5.0 as compared
with the features available when using the

MyISAM

and

InnoDB

storage engines. Currently, there are

no plans to address these in coming releases of MySQL 5.0; however, we will attempt to supply fixes
for these issues in subsequent release series. If you check the “Cluster” category in the MySQL bugs

Oracle 5.0, Reference Manual

Search results

Summary of Contents for 5.0

Reviews:

Related manuals for 5.0

Brands by name

Popular brands