Database Reference
In-Depth Information
You should tell users and testers that their ideas will be reviewed and that they may or may not be included
with the next version of the software. Creating expectations that every suggestion will be included in the solution
will ultimately ruin the feedback process.
From our experience, everyone may have suggestions about how to improve the process, but some will be
more vocal about it than others. Anonymity can encourage the more timid users to participate.
Make sure to keep the revision process as transparent as you can without becoming enslaved to the whims
of the more vocal users. While there is no guaranteed way of quieting those who must critique every solution
created by someone other than themselves, you can at least mitigate the damage they can do by letting everyone
know you are not ignoring their ideas out of hand.
This brings us back to even more documentation and management coordination. By now we're sure you're
seeing a pattern! In short, consider these suggestions:
•
Create a way to collect feedback from testers and users.
•
Create a way to let testers and users know they have been heard.
•
Keep the revision process as transparent as you can.
Tuning the BI Solution
Tuning a BI solution means improving its ability to perform all of its features in a minimal amount of time.
Ideally, each process, from running the ETL process to filling a cube to rendering the reports, should be
performed as quickly as possible so that anyone requesting information will not have to wait. These three actions
can be categorized as ETL performance, processing performance, and reporting performance.
ETL Performance
The ETL operation extracts data from one location, transforms it, and loads it in another location. You can think
of the ETL process as selecting data from the source and inserting data into the destination (Figure
18-3
)
. In
general, tuning ETL performance revolves around making these select and insert statements more efficient. Keep
in mind that if your ETL processing includes incremental loading, instead of the flush and fill technique we have
used in this topic, then it may also involve updating and deleting records. In other words, you will use SQL
UPDATE
and
DELETE
statements in the destination database. When you are trying to tune ETL performance, think of which
underlying statements are in use and tune the system accordingly.
Figure 18-3.
Actions performed during ETL operations
To tune the ETL system, start by increasing select performance. Select performance can be improved in a
number of ways, but most commonly improvement comes from providing indexes, adding hardware resources,
or reducing the amount of resources needed to perform the same action.
