MVSFORUMS.com

[schintala]

Hi Folks,
We have large table in our database which has more than 50 millions of rows and would like to partiton the table. We don't have any index or data column to set the range partiton. What would be best approach to make this table as partiton table. We are in V8 on z/OS.

One idea is to alter the table to have partition value as one column. Is it feasible? We are still in doubt how to make entries into this column with the exisiting data. Any suggestions would be a great help. Thank you for your time.

[CZerfas]

Well, the question is: What is the data citeria which should be used for building the partitions?

If you have no index on the table (strange, since you cannot garantee uniqueness of the rows), you probably have a timestamp when the row was inserted. This could be used nicely.

If this is not true, just look for a column which values could be divided into the number of partitions you want to create. The best would be something numeric, where the used values are evenly distributed.

regards
Christian

[rkinfy]

schintala,

When you do partitioning, the data is physically stored in that order. If you can identify a column based on which you normally search the data in that table, and if that column is not frequently changed, that would be the right column for partitioning. That way it'll increase the performance of the queries which are hitting that table.

If you are just interested in splitting the number of rows in to different partitions, as CZerfas suggested, you can look for some numeric columns and can even do the partitioning based on last digit or last two digits of that column which will distribute the load.

You can always do an alter table to add a partition key in the table but identifying a column which is already present would be easier.
_________________
RK.