Our self service customer portal is currently under maintenance and will be back soon! Sorry for any inconvenience this may have caused.
Partner is an easy-to-use command line tool that analyzes the distribution of data values in Enscribe files, NonStop SQL/MP tables, and indexes.
On HPE NonStop servers, a database file is typically partitioned over multiple disk volumes so that each disk stores a subset of the records in the file. This enhances application performance by allowing parallel access to the file. Multiple CPUS, disk IO controllers, and physical disks are utilized when accessing the file. This increases the maximum IO rate that can be achieved.
The distribution of the records within the file is determined by the so-called “first key” value of each partition. Each partition stores records with key values that lie within its first key and the first key of the next partition.
As a result of normal application processing, the partitions in a file can become unbalanced. Some partitions may experience higher IO rates than others, or may use more disk space than others.
Part of the database administrator’s job is to regularly rebalance the partitions in a file. To do this he must determine which first key values to use in order to redistribute the records between the partitions. This is where Partner comes in.
Partner analyses the distribution of key values within a file (or partition) and calculates the new first key values required to achieve the required distribution of records. It then generates a FUP or SQLCI script template which can be used to implement the new partitioning scheme.
Partner is a program that analyzes the distribution of keys in key sequenced Enscribe files, SQL/MP tables and indexes. You can use it to determine when and where to split partitions, or to repartition an entire file. Partner can produce a FUP or SQLCI script.
Depending on the size and type of file, Partner offers several analysis options — full scan for smaller files and random sample methods that estimate key distribution — both without incurring the I/O costs of reading all the records.