|
Hi,
to get a little more theoretically:
There are two kinds of indexing strategies:
1. Proactive Approach - Create the indexes you already know you'll need
You need indexes created over the fields/columns needed to join with
other files/tables.
You also need indexes over fields/columns which will used in the where
conditions with comparison = or IN-Predicate or Like without leading %.
Other access path can be determined with the reactive indexing strategy.
You may also create indexes over the fields/columns you'll use in Group
By and Order By.
But Group By and Order By are not as important as Join and Where
conditions.
The main goal is to get access to the requested data as fast as possible.
In this way the optimizer may use an index with other key fields than
specified in the order by clause.
Because it is sometimes faster to store the data in a temporary object
and sort the information stored in the temporary object,
than using an index with the requested order by columns.
Example: You specified 3 fields in the where clause (for example delivery
date and special delivery terms),
but want to have your result ordered after Order No.
In this situation it may be faster to use an index with delivery
date and delivery terms instead of using an index with OrderNo.
When Creating Indexes specify the most selective columns first.
2. Reactive Approach - Analyze your SQL environments
2.1. Look at the system wide index advisor, which is a physical
file/table SYSIXADV which exists since V5R4.
In the SYSIXADV all index advices made by both query engines (CQE
and SQE) are stored.
You'll also get information how often the indexes are advised and
when the last advice occurred.
You simply can use SQL or query to access this file, better is to
use the tools integrated into the iSeries Navigator
2.2. Run your SQL Statement with iSeries Navigator's Run an SQL Script
and use Visual Explain for explanation and to look for advises.
You may also use a STRDBG (without any parameters) in Green Screen
before executing STRSQL or calling a program with embedded SQL.
With STRDBG information about your SQL statements are written into
the joblog, which can be analyzed.
... but STRDBG is no longer a strategic product for SQL information
and does not contain all information.
2.3. Use STRDBMON or iSeries Navigator's Performance Monitors to collect
information about SQL statements
To analyze these information use the tools integrated into the
iSeries Navigator database.
There is a very good white paper about indexing strategies written by Mike
Cain. Even though it is written several years ago (and may not have the
newest features) it excellently covers the basics.
http://www-03.ibm.com/servers/enable/site/bi/strategy/strategy.pdf
DDS described logical files and SQL-Indexes:
Forget about logical files, only create SQL Indexes. Even though an SQL
index cannot be specified directly in an SQL statement it can be used with
native I/O like any keyed logical file. Beginning with release 6.1 it is
also possible to create additional columns (for example quantity*price) and
use them as key fields. Further it is also possible to specify where
conditions in an index (which is even more powerful then
SELECT/OMIT-Clauses
in DDS).
You only need DDS described logical files if you need to access a keyed
join
file with native I/O.
Mit freundlichen Grüßen / Best regards
Birgitta Hauser
"Shoot for the moon, even if you miss, you'll land among the stars." (Les
Brown)
"If you think education is expensive, try ignorance." (Derek Bok)
"What is worse than training your staff and losing them? Not training them
and keeping them!"
-----Ursprüngliche Nachricht-----
Von: midrange-l-bounces@xxxxxxxxxxxx
[mailto:midrange-l-bounces@xxxxxxxxxxxx] Im Auftrag von
Michael_Schutte@xxxxxxxxxxxx
Gesendet: Thursday, 19. November 2009 22:34
An: Midrange Systems Technical Discussion
Betreff: Ideas for logicals (index).
I'm looking for some suggestions. We're going to be creating a large
summary file and I'd like to get some ideas on what logical files I should
create. Normally, we fly by the seat of our pants and create them when
needed. But since this file is going to be large, I'd like to get them
created beforehand.
Anyway, the file unique key is by Company, Restaurant, Date, Item, Meal
Period. I believe SQL will mainly be used to retrieve the data from this
file.
I believe one logical should be created by Company and Meal Period, then
just throw in restaurant, date and item behind it. I say Company and meal
period first because currently there are only 2 possible values for company
and 4 possible values for Meal Period.
But then my concern comes in when join in other files.
For example, I may want to get region 1 restaurants only. So I'm going to
need a logical by company and restaurant, but should I throw in the other
fields too?
Then we might need to join in the Item Master file to select only
beverages.
So a logical by company and item would be needed.
Finally, I think a locigal by Company, date would be good to have.
I guess I'm all over the place here, not really sure how to explain myself.
Does anyone have any suggestions? I've read for SQL indexes that you don't
want to create indexes with the same fields (but in different order)
because
it doesn't really buy you anything (take longer for record to be inserted).
Would this be true for logicals too? What's the harm if we add the other
fields anyway, just in case someone in the future wants to do I/O reads on
the file or for the group by clause?
I think I've told before that you want to try to create you logicals where
the key fields with the least amount of unique values are first.
(Please no DDL solutions, boss doesn't want to start using them yet because
not too many of us have experience in them).
--
Michael Schutte
Admin Professional
Bob Evans Holiday Farmhouse Feast, Serves 6-8 l $74.99 A complete homestyle
meal TO GO, ready to heat at home, serve & enjoy!
Perfect for Thanksgiving, Christmas or holiday entertaining.
For more information, visit www.FarmhouseFeast.com<http://www.farmhousefeast.com/>
--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing
list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at http://archive.midrange.com/midrange-l.
As an Amazon Associate we earn from qualifying purchases.
This mailing list archive is Copyright 1997-2024 by midrange.com and David Gibbs as a compilation work. Use of the archive is restricted to research of a business or technical nature. Any other uses are prohibited. Full details are available on our policy page. If you have questions about this, please contact [javascript protected email address].
Operating expenses for this site are earned using the Amazon Associate program and Google Adsense.
midrange.com
