Minimizing Analytic Workspace Growth

Ways to minimize analytic workspace growth

This guide presents some very basic and simple information about ways in which you can minimize analytic workspace growth.

You can minimize analytic workspace growth through the judicious use of NA stored pages and by frequently updating the analytic workspace when you are attached exclusively. You can completely reorganize the analytic workspace by exporting and importing all of the analytic workspace files.

Definition: NA pages

An NA page is an analytic workspace page that contains only NA values for a variable. Depending on the status of the variable at the time the NA values are assigned, NA pages are either unstored or stored:

• NA unstored pages -- In most cases, if an analytic workspace page would contain only NA values for a variable, then the page is not actually stored. Instead a marker is stored. This marker indicates that the page would contain only NA values. These virtual analytic workspace pages are called NA unstored pages.

• NA stored pages -- An NA stored page is an actual analytic workspace page that contains only NA values. Typically, you want NA stored pages to be created only if you want to reserve space for values of an in-place variable.

When are NA stored pages created?

NA stored pages are created only if all of the following conditions are true:

• The NA values are explicitly assigned.

• The variable is defined as an in-place variable when the NA values are assigned.

• The analytic workspace that contains the variable is attached in read/write, exclusive mode when the NA values are assigned.

• The variable contains, or has previously contained, at least one non-NA value.

• The NA values are distributed so that some analytic workspace pages contain only NA values.

Retrieving the number of NA pages

You can use the OBJ function with the NAPAGES keyword to retrieve the number of NA pages (either stored and unstored) in an analytic workspace.

You can also run DBREPORT to retrieve information about the number of NA pages (both stored and unstored) in an analytic workspace.

For more information on the OBJ function and the DBREPORT program, see the topic for the function or the program in the OLAP DML Reference.

Releasing NA stored pages

You can use the NAPAGEFREE command to release any NA stored pages that have been created for a variable. Once these pages are released, they can be used to store new data. NAPAGEFREE loops through all allocated pages for the variable and converts any NA stored pages into NA unstored pages. When it is finished, it reports the number of pages that have been freed.

For more information on the NAPAGEFREE command, see the topic for the function in the OLAP DML Reference.

Example: Releasing NA stored pages

The following example uses OBJ(DISKSIZE) to query the variable SALES before and after NAPAGEFREE is issued to show its reduction by the number of pages that are freed by NAPAGEFREE (three in this example).

The first OBJ function shows that 35 pages are being used to store the SALES variable.

show obj(disksize 'sales')
35

Now the NAPAGEFREE command frees three pages that contained only NA values.

napagefree sales
3 pages freed for SALES.

When the OBJ function is reissued, it shows that only 33 pages are now being used to store the SALES variable.

show obj(disksize 'sales')
32

When are unused pages released?

When many users are attached to an analytic workspace, unused pages are not actually released when the analytic workspace is updated. Instead, an erase list is created. The erase list identifies the pages that it can release later when only one user is attached to the analytic workspace.

Updating an analytic workspace when you are attached exclusively

When you update an analytic workspace and no other users are attached to the analytic workspace, the erase list is flushed and all unused pages are released. This creates more space in the analytic workspace files for new data. Consequently, to minimize analytic workspace growth, you want to update the analytic workspace frequently when you have exclusive use it.

Using EXPORT and IMPORT to minimize analytic workspace size

You can reorganize your analytic workspace files by exporting all of the objects in your analytic workspace and then importing them into a new analytic workspace. This procedure removes extra space. The new files may be substantially smaller.

To reorganize your analytic workspace by exporting and importing OLAP DML objects, follow the procedure outlined below.

1. Issue an ALLSTAT command against the original analytic workspace.

2. Use the EXPORT command with the ALL keyword to put all of the data in the original analytic workspace into an EIF file.

3. Create a new analytic workspace with a different name than the original analytic workspace.

4. Use the IMPORT command to import the EIF file into the new analytic workspace.

5. Use the UPDATE command to update the new analytic workspace.

6. After checking that the objects were successfully moved into the new analytic workspace, delete the original analytic workspace.

7. Rename the new analytic workspace with the original name.

For more information on importing and exporting analytic workspace files, see the topics for the EXPORT and IMPORT commands in the OLAP DML Reference.

You cannot export and import SEGWIDTH specifications

If you use CHGDFN SEGWIDTH to specify the segment size of any variable, you should be aware that this information cannot be exported and imported. If you export any variable, when that variable is imported, it will use the default segment size.