Oracle Query Performance Tuning:

A 12Step Program
REVIEW FILTER AND
EXAMINE THE Examine cost, ACCESS PREDICATES
Dont use
EXECUTION PLAN row count and explain plans
Know how query
time is interpreted (e.g.,
bind variables)
2
Know when
1 Get actual plan
Get actual
row counts
predicate is applied
(should be earlier
Find the (use DBMS_XPLAN. for each step rather than later)
expensive DISPLAY_CURSOR)
operators
Review data types
to avoid implicit
EVALUATE
Review table conversions
EXISTING
definitions
INDEXES Get table sizes
ANALYZE 3
COLUMNS IN Get index and row count GATHER TABLE
WHERE CLAUSE
definition INFORMATION

Look for select * or

4
If its a view, get
scalar functions (some Review current the underlying table
are less optimal)
5 statistics definitions
Know order of
columns and their
selectivity
Know if histograms or
Foreign key constraints
Know table baselines are being
can help the optimizer create
relationships (avoid used; get sample bind
Know the selectivity and REVIEW EXISTING better execution plans
duplicating efforts variable values
cardinality of the columns KEYS AND later on)
(is there data skew?)
CONSTRAINTS
TIP

6 What is primary
key definition? RUN THE QUERY AND
RECORD BASELINE
Avoid sub-queries SARG-able METRICS
(need index?)
7
TUNE
Gather average
Look for high-cost steps THE QUERY execution times
and steps with largest
RE-RUN Nested loop/ number of rows
THE QUERY merge/hash joins
8 Focus on logical
I/Os (number of
Capture wait logical reads)
TIP Focus on events
Seeks vs scanswhich most expensive
Record the 9 is more expensive in this operations first
Out-of-date
time, logical I/O scenario?
statistics can impact
and wait events, performance
and compare RE-RUN
with initial
Consider using a function-
THE QUERY
baseline
CONSIDER ADJUSTING INDEXES based index to exclude rows LOOK FOR
in very large tables PERFORMANCE
INHIBITORS
11
Make small
changes 10
Reduce
logical I/O 12
Abuse of wildcards (*)
pulling back too many
Nested views that go rows
Cursors and row-by-row
across database links
processing
Code-based SQL
generators (e.g. EMF,
LNO, Hibernate)
Join/query/table hints Scalar functions

Find and fix query performance faster with SolarWinds Database Performance Analyzer. Free 14-day trial at: www.solarwinds.com/dpa-oracle-download
2015 SolarWinds, Inc. All rights reserved.