Server User Manual

When specifying the page size, carefully consider the following point:
· If the table does not contain a BLOB-type column, a row of data in the table must fit within in one page.
· If a table including a BLOB-type column is defined or a BLOB-type column is added during the change of a
table definition, a row of data in the table may exceed one page. However, the total size of the data of the
columns other than the BLOB-type column must not exceed one page.
When an index must be added to a SEQUENTIAL structure
The user must add an index for a column of the corresponding table that has a unique constraint. If no index is added
for such a column, the table cannot be accessed.
4.1.2 RANDOM structure
In a RANDOM structure, collections of storage pages (called buckets) are calculated from the values of the group of
columns defined as a key for the data. The data is stored in pages within those buckets. If the data cannot fit in a
bucket, the SymfoWARE/RDB system automatically creates an overflow part bucket and stores the data in that bucket.
The collection of pages that belong to the original bucket are called the prime part for the overflow part.
A group of columns that determines the page for storing data is called a cluster key. A cluster key is determined by the
CLUSTER option of the table DSO definition. If the CLUSTER option is omitted, the cluster key becomes the primary
key of the corresponding table definition. In a RANDOM structure, data having an equal cluster key is stored in the
same packet.
The hash function is used in calculations to determine a bucket from the cluster key value. From the hash function,
SymfoWARE/RDB automatically determines a bucket for storing data. If RULE is specified in the storage option of the
table DSO definition statement, the data storing bucket is determined from the result of calculating the formula
specified at RULE.
Figure: Overview of RANDOM structure shows an overview of the RANDOM structure, using the STOCK table as an
example.
141