The percentage performance gain we type into this box option represents the targeted percentage improvement between the maximum and minimum query times. Twenty is a good starting target, and can be expected, generally, to result in adequate aggregation to ensure a significant increase in performance. Diminishing returns can result from setting the percentage unnecessarily high; the idea here is to attain a good level of balance between the increased disk space required by new aggregations and the level of overall performance.
A third radio button represents another option for getting to the best aggregation mix. With the Until I click stop setting, we can attempt to manually determine the best balance in conjunction with keeping an eye on the Performance vs. Size graph that appears to the right of the dialog. We would, ideally, determine the point at which the increase in performance begins to level off while storage continues to increase, and then stop the process.
As we progress through this and other series, we will focus, from time to time, on the use of the parameters found within the Set aggregate options dialog, along with numerous others, to meet specific tuning objectives, or to offer options for overall improvement in a certain aspect of query or processing performance. For now, let's get a grasp of the operation of the tool.
The Set Aggregation Options dialog, with our settings, appears as depicted in Illustration 9.
Illustration 9: Set Aggregation Options Dialog
The remaining buttons include the following:
- Start: Kicks off the aggregation design process, based upon our settings;
- Continue: Activates once we click Stop, or when the Performance vs. Size graph indicates we have met our storage or performance gain targets, as a means of resuming the design process;
- Stop: Allows us to manually stop the design process;
- Reset: Enables us to delete aggregations added and restart the design process.
To conclude this section, let's proceed with the following steps.
The Next button will activate as soon as the design process finishes. The Set aggregation options dialog, after the process is complete, appears as shown in Illustration 10.
Illustration 10: Set Aggregation Options Dialog, Results Displayed at Lower Right
We see that the Wizard has produced 12 aggregations, and reached the 20% estimated performance gain level, as indicated underneath the Performance vs. Size graph. Pressing Continue (which has become enabled) at this point will result in intermittent incremental increases above 20%, with a leveling off of the curve, so a degree of manual tweaking can be had for minimal additional effort. We will leave the results as they are, however and move ahead.
9. Click Next, to arrive at the Finish the Storage Design Wizard dialog, as shown in Illustration 11.
Illustration 11: Finish the Storage Design Wizard Dialog
We can either save our new definition at this point, or process the cube to create the new aggregations. The definition remains stored until the design is enacted, via the processing cycle.
Let's process the cube, and pay attention to the effects in the resulting Process Log.
10. Leaving the radio button (underneath What do you want to do now?) at its default of Process now, click Next.
Cube processing begins, and runs its course quickly, as witnessed by the Process Log window that next appears. Once processing is finished, we notice the green Processing completed successfully message appears at the bottom of the log window, as depicted in Illustration 12.
Illustration 12: Processing Completed Successfully Message
(Window Partially Collapsed)
We note that processing duration time, together with the various steps of the process, is detailed in the log window. This presents an opportunity, in tuning evolutions, to compare process times between the current log and previous logs (all logs are captured in a database, which we have explored in other articles, and will again examine in subsequent articles). We note, too, that the log states, "cube needs to be processed" on the fourth line from the top of the entries, an indication that the condition was noted as soon as we began processing. Keep in mind that changes planned via the Storage Design Wizard (among numerous other structural changes) require a processing run to be consummated.
11. Click Close to close the log window, once you have examined it.
We leave the Process Log window and the Cube Editor behind, and arrive in Analysis Manager, once more.
12. Delete the DBJ_STORDESIGN cube, if desired, by right-clicking and selecting Delete from the context menu that appears.
13. Select File --> Exit from the Main Menu in the Management Console to close Analysis Manager.
We have seen operation of the Storage Design Wizard from start to finish, for a simple cube. We will revisit the Storage Design Wizard from time to time, specifically within the context of our work with partitions, which allow us to design aggregations differently for separate "sections" of a cube, and within other articles where we can employ it to help us meet various maintenance and optimization objectives.
In this lesson, we introduced the Storage Design Wizard, and emphasized its value as a tool within the important, and often complex, context of storage configuration for MSAS. Our objective was to expose the tool's role in storage and aggregation design, as well as the basic concepts of MSAS storage design in general.
We explored some of the scenarios where we might use the Storage Design Wizard, and discussed the storage types that are available to us in our cube designs. We then practiced using the Storage Design Wizard within the context of a simple cube with no pre-existing aggregations, and exposed the steps involved in increasing performance through basic storage design. We processed the cube to put our changes into effect, discussing processing considerations within the scope of storage design. Finally, we discussed how the Processing Log can be used as a means of ascertaining the effects of our storage and aggregation design upon cube processing performance.
» See All Articles by Columnist William E. Pearson, III
Discuss this article in the MSSQL Server 2000 Analysis Services and MDX Topics Forum.