Beruflich Dokumente
Kultur Dokumente
E : info@rittmanmead.com
W : www.rittmanmead.com
The Requirement
E : info@rittmanmead.com
W : www.rittmanmead.com
The Problem
Lots of moving parts and complexity
E : info@rittmanmead.com
W : www.rittmanmead.com
E : info@rittmanmead.com
W : www.rittmanmead.com
E : info@rittmanmead.com
W : www.rittmanmead.com
E : info@rittmanmead.com
W : www.rittmanmead.com
Two-pronged approach
Do It Right, First Time
What Makes A Performant OBIEE System?
!
If Its Not Done Right, Know How To Figure Out What Is Broke
Practical elements of diagnosing the causes of performance issues
Methodical analysis - nose to tail
E : info@rittmanmead.com
W : www.rittmanmead.com
Response time
E : info@rittmanmead.com
W : www.rittmanmead.com
BI Plug-in
Presentation
Services
[ ... ]
BI Server
WebLogic Server
Managed Server
Web Browser
BI Plug-in
Network
User
Network
[ ... ]
Data Source(s)
DWH
Presentation
Services
JavaHost
BI Server
[ ... ]
[ ... ]
Response
Not all components listed, eg cluster controller
E : info@rittmanmead.com
W : www.rittmanmead.com
Time Profile
To understand why is it slow, we first must understand where is it slow
!
E : info@rittmanmead.com
W : www.rittmanmead.com
DB Query 1
25 seconds
10 seconds
5 seconds
DB Query 2
5 seconds
Component
User
WebLogic Server
Managed Server
Web Browser
BI Plug-in
Presentation
Services
BI Server
Response
E : info@rittmanmead.com
W : www.rittmanmead.com
Data Source(s)
DWH
DWH
Time, seconds
DB Query 1
DB Query 2
Data Processing
Page Generation
Step
Action
Response
Time (s)
25.00
62.5
[5.00]
10.00
25.0
5.00
12.5
E : info@rittmanmead.com
W : www.rittmanmead.com
Total 40.00
100.0
Response
DB Connect
Compilation
Physical Query 1
BI Server
processing
(federation,
calculation,
aggregation)
E : info@rittmanmead.com
W : www.rittmanmead.com
r - d - c - max(p)
e - r
Step
Logical Query Compilation
Database connection
Physical Query 1 response
Physical Query 2 response
Physical Query n response
Logical Query Response
Logical Query Elapsed
BI Server processing (federation,
calculation, aggregation)
Send to client, wait for
acknowledgement (eg sawserver
generating pivot tables etc)
nqquery.log source
Usage Tracking
c
d
p
p
p
r
e
r-d-c-max(p)
S_NQ_ACCT.COMPILE_TIME_SEC
n/a
S_NQ_DB_ACCT.TIME_SEC
S_NQ_DB_ACCT.TIME_SEC
S_NQ_DB_ACCT.TIME_SEC
S_NQ_ACCT.TOTAL_TIME_SEC
n/a
[derived]
[derived]
e-r
[derived]
n/a
E : info@rittmanmead.com
W : www.rittmanmead.com
Elapsed
nqquery.log:
Logical Query Summary Stats: Elapsed time e
Usage Tracking: [Not Available]
Response
nqquery.log:
Logical Query Summary Stats: [...] Response time r
Usage Tracking: S_NQ_ACCT.TOTAL_TIME_SEC
uh oh!
E : info@rittmanmead.com
W : www.rittmanmead.com
Iterative Approach
Be sure to account for all time, end-to-end (nose to tail)
Expand time profile to focus on where the time is going in particular
In reality, time profile based on BI Server alone will help in a lot of cases
Browser tools
WLS logs
nqquery.log
sawlog.log
Request
OBIEE system components
Managed Server
BI Plug-in
Presentation
Services
BI Server
Response
E : info@rittmanmead.com
W : www.rittmanmead.com
Data Source(s)
Network
Web Browser
WebLogic Server
Network
User
DWH
DWH
Iterative Approach
Most components have extended logging
levels
Presentation Services
WebLogic Server
Browser Tools
E : info@rittmanmead.com
W : www.rittmanmead.com
E : info@rittmanmead.com
W : www.rittmanmead.com
http://ritt.md/oma-intro
T : +44 (0) 1273 911 268 (UK) or (888) 631-1410 (USA) or
+61 3 9596 7186 (Australia & New Zealand) or +91 997 256 7970 (India)
E : info@rittmanmead.com
W : www.rittmanmead.com
Step
nqquery.log source
Logical Query Summary Stats: Elapsed time
BI Server processing
r-d-c-max(p) [derived]
[derived]
Send to client, wait for cursor close
e-r
1
E : info@rittmanmead.com
W : www.rittmanmead.com
Time %
14
13
0
0%
0
0%
4
29%
9
64%
1
7%
E : info@rittmanmead.com
W : www.rittmanmead.com
Performance
Build a time profile
Dont know where to start? Try Usage Tracking or nqquery.log
- Long running queries
- Logical queries returning lots of data to the client
- Logical queries returning lots of data from the DB but small % to the client
- Logical queries requiring lots of physical queries
E : info@rittmanmead.com
W : www.rittmanmead.com
E : info@rittmanmead.com
W : www.rittmanmead.com
Depends
c.f. ExtremeBI
E : info@rittmanmead.com
W : www.rittmanmead.com
Dont Do It
Do It Only Once
Do It Less Often
Do It More Efficiently
T : +44 (0) 1273 911 268 (UK) or (888) 631-1410 (USA) or
+61 3 9596 7186 (Australia & New Zealand) or +91 997 256 7970 (India)
E : info@rittmanmead.com
W : www.rittmanmead.com
BI Plug-in
Presentation Services
E : info@rittmanmead.com
W : www.rittmanmead.com
BI Server
Data Source(s)
Network
Web Browser
WebLogic Server
Managed Server
Network
User
DWH
[ ... ]
nqquery.log source
Logical Query Summary Stats: Elapsed time
e
Logical Query Summary Stats: [...] Response time
r
Logical Query Summary Stats: [...] Compilation time
c
Physical Query Summary Stats: [...] DB-connect time
d
Physical query response time
p1
r-d-c-max(p) [derived]
[derived]
e-r
E : info@rittmanmead.com
W : www.rittmanmead.com
E : info@rittmanmead.com
W : www.rittmanmead.com
Yes, please
do.
Be smart about Dashboard and Analysis design - dont cram everything into one page
Less clutter : Better user experience & Better performance
E : info@rittmanmead.com
W : www.rittmanmead.com
Not forgetting
Do It More Efficiently
Aggregation
Do It Less Often
BI Server caching
!
E : info@rittmanmead.com
W : www.rittmanmead.com
If Its Not Done Right, Know How To Figure Out What Is Broke
Methodical analysis - nose to tail
E : info@rittmanmead.com
W : www.rittmanmead.com
#EOF
email:
robin.moffatt@rittmanmead.com
web:
http://www.rittmanmead.com/author/robin-moffatt/
twitter: @rmoff
E : info@rittmanmead.com
W : www.rittmanmead.com
E : info@rittmanmead.com
W : www.rittmanmead.com