Advanced ASE Performance Tuning Tips
6 likes6,035 views
In this presentation, attendees will learn how to determine the best approach for tuning SQL statements and other issues by identifying specific causes of slow performance. Real-life case studies will be used to demonstrate: 1.How to use Response Time Analysis (RTA) to quickly identify the biggest problems in a database 2.How to utilize the MDA tables to understand what resources are being used or waited on 3.Tips and techniques on examining and further fine-tuning the execution plan
![37
© 2015 SOLARWINDS
SQL Diagramming
o .02
.005
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM Customer) FROM Customer
WHERE LastName LIKE 'Griffin%'
0.005
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [SalesOrder]) FROM
[SalesOrder] WHERE OrderDate >= DATEADD(day, -30, GETDATE())
.02
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [Shipment]) FROM [Shipment]
WHERE ShipStatus <> 'C'
.04
i c
ps
a
.
04](https://image.slidesharecdn.com/sapadvancedaseperformancetuningtips15-150611172551-lva1-app6891/85/Advanced-ASE-Performance-Tuning-Tips-37-320.jpg)
![39
© 2015 SOLARWINDS
SQL Diagramming
o .02
.005
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM Customer) FROM Customer
WHERE LastName LIKE 'Griffin%'
0.005
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [SalesOrder]) FROM
[SalesOrder] WHERE OrderDate >= DATEADD(day, -30, GETDATE())
.02
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [Shipment]) FROM [Shipment]
WHERE ShipStatus <> 'C'
.04
i c
ps
a
.
04](https://image.slidesharecdn.com/sapadvancedaseperformancetuningtips15-150611172551-lva1-app6891/85/Advanced-ASE-Performance-Tuning-Tips-39-320.jpg)
![41
© 2015 SOLARWINDS
SQL Diagramming
o .02
.005
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM Customer) FROM Customer
WHERE LastName LIKE 'Griffin%'
0.005
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [SalesOrder]) FROM
[SalesOrder] WHERE OrderDate >= DATEADD(day, -30, GETDATE())
.02
SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [Shipment]) FROM [Shipment]
WHERE ShipStatus <> 'C'
.04
i c
ps
a
.
04](https://image.slidesharecdn.com/sapadvancedaseperformancetuningtips15-150611172551-lva1-app6891/85/Advanced-ASE-Performance-Tuning-Tips-41-320.jpg)
![42
© 2015 SOLARWINDS
Data Skew Problems
» Only 4% of rows are <> ‘C’
» How about changing the query?
§ AND s.ShipStatus = 'I'
» Add Histogram on ShipStatus
SELECT ShipStatus, COUNT(1)Cnt
FROM [Shipment]
GROUP BY ShipStatus](https://image.slidesharecdn.com/sapadvancedaseperformancetuningtips15-150611172551-lva1-app6891/85/Advanced-ASE-Performance-Tuning-Tips-42-320.jpg)
Ad
More Related Content
What's hot (20)
Viewers also liked (17)
Ad
Similar to Advanced ASE Performance Tuning Tips (20)
Ad
More from SAP Technology (20)
Recently uploaded (20)
Advanced ASE Performance Tuning Tips
- 1. 1 Advanced ASE Performance Tuning Tips Janis Griffin Senior DBA / Performance Evangelist
- 2. 2 © 2015 SOLARWINDS Who Am I? » Senior DBA / Performance Evangelist for Solarwinds § [email protected] § Twitter - @DoBoutAnything § Current – 25+ Years in Oracle, Sybase, SQL Server § DBA and Developer » Specialize in Performance Tuning » Review Database Performance for Customers and Prospects » Common Thread – Paralyzed by Tuning
- 3. 3 © 2015 SOLARWINDS Before we start, do you know… » The most important problem in your database instance? » » If that new program is hurting database performance? » » Did that recent fix really solved the database issue? » » Which bottlenecks in your database directly impacted end- user service? »
- 4. 4 © 2015 SOLARWINDS Agenda - 4 Tips » Focus on Tuning Queries » Utilize Response Time Analysis (RTA) » Examine Query Execution Plans » Use SQL or Query Diagramming § Who registered yesterday for Tuning Class § Check order status § Current paychecks for specific employees §
- 5. 5 © 2015 SOLARWINDS Why Focus on Queries » Most Applications § Read and Write data to/from database § Simple manipulation of data § Deal with smaller amounts of data » Most Databases § Examine larger amounts of data, return a little § Inefficiencies quickly become bottleneck » Why do SQL tuning? § “Gives the biggest bang for your buck” § Changes to SQL are Safer § Most of performance issues are SQL related
- 6. 6 © 2015 SOLARWINDS Challenges Of Tuning » SQL Tuning is Hard § Who should tune – DBA or Developer § Which SQL to tune § » Requires Expertise in Many Areas § Technical – Plan, Data Access, SQL Design § Business – What is the Purpose of SQL? § » Tuning Takes Time § Large Number of SQL Statements § Each Statement is Different § » Low Priority in Some Companies § Vendor Applications § Focus on Hardware or System Issues § » Never Ending
- 7. 7 © 2015 SOLARWINDS Which SQL to Tune Methods for Identifying » User / Batch Job Complaints § Known Poorly Performing SQL § Trace Session/Process § » Queries Performing Most I/O (LIO / PIO) § Table or Index Scans » Queries Consuming CPU » Highest Response Times - DPA (formally Ignite)
- 8. 8 © 2015 SOLARWINDS Response Time Analysis (RTA) §Understand the total time a Query spends in Database §Measure time while Query executes §SAP ASE helps by providing Wait Events Focus on Response Time
- 9. 9 © 2015 SOLARWINDS Wait Time Tables (ASE 12.5.0.3+) http://froebe.net/blog/wp-content/uploads/2013/09/ASE-15.7-Monitoring-Tables-Diagram-Sybase-Inc_.pdf
- 10. 10 © 2015 SOLARWINDS Response Time Query » MDA Query Example SELECT mP.Login, mP.DBName, mPL.ClientHost, RTRIM(mPL.Application), mPP.ObjectName, mP.WaitEventID, mPS.SQLText, mPS.BatchID, mPS.LineNumber, mP.LineNumber CurrentLineNumber FROM monProcessSQLText mPS, monProcessLookup mPL, (monProcess mPLEFT OUTER JOIN monProcessProcedures mPP ON mP.SPID=mPP.SPID AND mP.KPID=mPP.KPID) WHERE mP.SPID=mPL.SPID AND mP.KPID=mPL.KPID AND mP.SPID=mPS.SPID AND mP.KPID=mPS.KPID AND coalesce(mP.FamilyID,0) = 0 AND mP.BatchID > 0
- 11. 11 © 2015 SOLARWINDS Top Wait Events At Server Level select WaitEventID, convert(numeric(16,0),Waits) as "Waits", convert(numeric(16,0),WaitTime) as "WaitTime" into #waits1 from monSysWaits go select Description, convert(int,sum(w.Waits)) as "Count", convert(int,sum(w.WaitTime)/1000) as "Seconds" from #waits1 w, monWaitEventInfo ei where w.WaitEventID = ei.WaitEventID group by Description order by 3 desc
- 12. 12 © 2015 SOLARWINDS RTA – Proactive Refreshed on :08/12/14 02:35:18 PM
- 13. 13 © 2015 SOLARWINDS Users Complain: 10am – 12pm Slow Top 15 SQL Statements | GIBSON:5000 | August 8, 2014 to August 12, 2014 Daily Time Range: 10:00 AM-12:00 PM
- 14. 14 © 2015 SOLARWINDS RTA – Firefighting Day: Tuesday – August 10,2014 Refreshed on: 08/12/2014 05:09:57 PM
- 15. 15 © 2015 SOLARWINDS RTA - Blocking Issue Day: Tuesday – August 10,2014 Refreshed on: 08/12/2014 05:02:02 SQL Statements Executed by Blocking Session 125 while Holding the Lock | GIBSON:5000 August 10, 2014 – 12:00PM to 4:00PM
- 16. 16 © 2015 SOLARWINDS RTA – Long Term Trends Top 15 SQL Statements | GIBSON:5000 | July 31,2014 to August 24, 2014
- 17. 17 © 2015 SOLARWINDS RTA – Current Problem Refreshed on: 08/12/2014 05:34:42 PM
- 18. 18 © 2015 SOLARWINDS RTA – Current Problem Refreshed on: 08/12/2014 05:39:06 PM
- 19. 19 © 2015 SOLARWINDS Identify End-to-END Time Accurate End-to-End Response Time Analysis
- 20. 20 © 2015 SOLARWINDS Get Table & COLUMN Info » Understand objects in execution plans § Table Definitions & Segment sizes • Is it a View – get underlying definition • Number of Rows / Partitioning § Examine Columns in Where Clause • Cardinality of columns • Data Skew / Histograms § § Statistic Gathering • Tip: Out-of-date statistics can impact performance » Understand expensive data access targets § TableScans versus IndexScans § » Know the indexes – if multi-column, order is important
- 21. 21 © 2015 SOLARWINDS Get The Execution Plan » SET SHOWPLAN ON § set statistics io, time on simulate, subquerycache, plancost § set noexec on http://help.sap.com/Download/Multimedia/ASE_16.0/ptstats.pdf » Interactive SQL (Sybase iAnyWhere or SAP dbisql) § Plan Viewer– must execute query § Details, XML & Advanced (abstract & missing) » Abstract Plans-sysqueryplans /sysquerymetrics § sp_configure'enable metrics capture', 1 » Application Tracing (v15.0.2+) § set tracefile "<file-path>" for <spid> § Works when you know a problem will occur
- 22. 22 © 2015 SOLARWINDS Get The Execution Plan - tracing set statistics plancost on set tracefile ’/db1/trc/trc32.tx’ for 32
- 23. © 2015 SOLARWINDS Examine the Execution Plan » Find Expensive Operators § Examine cost of each step § Look for full table or index scans » Review the ‘restrict’ & ‘sort’ operators § Are the ‘restrict’ steps early in plan? § Large sorts that don’t fit into memory § Look for Implicit conversions when using parmeters » Review JOIN operators § Nested loop • effective when there is a useful index available § Merge Join • good when most of the rows must be processed and they are already sorted by join keys § Hash Join • good when most of the rows from source sets are processed § Nary Nested Loop • More efficient when you have more than 2 nested loops 23
- 24. © 2015 SOLARWINDS Engineer out the Stupid » Look for Performance Inhibitors § Cursor or row by row processing § Parallel processing - OLAP vs OLTP § § Hard-coded Hints § § Nested views that use link servers § Abuse of Wild Cards (*) or No Where Clause § § Code-based SQL Generators (e.g. Hibernate) § § Non-SARG-able / Scalar Functions • Select… where upper(first_name) = ‘JANIS’ 24
- 25. 25 © 2015 SOLARWINDS Execution Plan Example
- 26. 26 © 2015 SOLARWINDS » Who registered yesterday for SQL Tuning SELECT s.FNAME, s.LNAME, r.SIGNUP_DATE FROM STUDENT s INNER JOIN REGISTRATION r ON s.STUDENT_ID = r.STUDENT_ID INNER JOIN CLASS c ON r.CLASS_ID = c.CLASS_ID WHERE c.NAME = 'SQL TUNING' AND r.SIGNUP_DATE BETWEEN @BegDate AND @EndDate AND r.CANCELLED = 'N‘ Case Study 1
- 27. 27 © 2015 SOLARWINDS Relationship Diagram for Tables in Query
- 28. 28 © 2015 SOLARWINDS Execution Plan
- 29. 29 © 2015 SOLARWINDS SQL Diagramming » Great Book “SQL Tuning” by Dan Tow § Great book that teaches SQL Diagramming § http://www.singingsql.com REGISTRATION STUDENT CLASS 5 1 3 0 1 . 05 . 002 select count(1) from registration where cancelled = 'N' and signup_date between '2015-01-29 00:00' and '2015-01-30 00:00' 4,141 / 79,999 = 0.05 select count(1) from class where name = 'SQL TUNING' 2 / 1,000 = .002
- 30. 30 © 2015 SOLARWINDS create index cl_name ON CLASS(NAME) New Execution Plan Why would an TableScan still occur on REGISTRATION?
- 31. 31 © 2015 SOLARWINDS Database Diagram
- 32. 32 © 2015 SOLARWINDS create index reg_alt on REGISTRATION(CLASS_ID) New Execution Plan
- 33. 33 © 2015 SOLARWINDS create index reg_alt ON REGISTRATION (CLASS_ID, SIGNUP_DATE,CANCELLED) Better Execution Plan
- 34. 34 © 2015 SOLARWINDS » Lookup Shipment Status for caller SELECT o.OrderID, c.FirstName, c.LastName, p.ProductID, p.Description, s.ActualShipDate, s.ShipStatus, s.ExpectedShipDate FROM SalesOrder o INNER JOIN OrderItem i ON i.OrderID = o.OrderID INNER JOIN Customer c ON c.CustomerID = o.CustomerID INNER JOIN Shipment s ON s.ShipmentID = i.ShipmentID LEFT OUTER JOIN Product p ON p.ProductID = i.ProductID LEFT OUTER JOIN Address a ON a.AddressID = s.AddressID WHERE c.LastName LIKE ISNULL('Griffin','') + '%' AND o.OrderDate >= DATEADD(day, -30,getdate()) AND s.ShipStatus <> 'C' order by LastName Case Study 2
- 35. 35 © 2015 SOLARWINDS Database Diagram
- 36. 36 © 2015 SOLARWINDS Execution Plan
- 37. 37 © 2015 SOLARWINDS SQL Diagramming o .02 .005 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM Customer) FROM Customer WHERE LastName LIKE 'Griffin%' 0.005 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [SalesOrder]) FROM [SalesOrder] WHERE OrderDate >= DATEADD(day, -30, GETDATE()) .02 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [Shipment]) FROM [Shipment] WHERE ShipStatus <> 'C' .04 i c ps a . 04
- 38. 38 © 2015 SOLARWINDS New Execution Plan » CREATE INDEX CUST_LASTNAME ON Customer(LastName) »
- 39. 39 © 2015 SOLARWINDS SQL Diagramming o .02 .005 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM Customer) FROM Customer WHERE LastName LIKE 'Griffin%' 0.005 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [SalesOrder]) FROM [SalesOrder] WHERE OrderDate >= DATEADD(day, -30, GETDATE()) .02 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [Shipment]) FROM [Shipment] WHERE ShipStatus <> 'C' .04 i c ps a . 04
- 40. 40 © 2015 SOLARWINDS CREATE INDEX SALESORDER_ODATE ON SalesOrder (OrderDate) New Execution Plan
- 41. 41 © 2015 SOLARWINDS SQL Diagramming o .02 .005 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM Customer) FROM Customer WHERE LastName LIKE 'Griffin%' 0.005 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [SalesOrder]) FROM [SalesOrder] WHERE OrderDate >= DATEADD(day, -30, GETDATE()) .02 SELECT COUNT(1)*1.0/(SELECT COUNT(1) FROM [Shipment]) FROM [Shipment] WHERE ShipStatus <> 'C' .04 i c ps a . 04
- 42. 42 © 2015 SOLARWINDS Data Skew Problems » Only 4% of rows are <> ‘C’ » How about changing the query? § AND s.ShipStatus = 'I' » Add Histogram on ShipStatus SELECT ShipStatus, COUNT(1)Cnt FROM [Shipment] GROUP BY ShipStatus
- 43. 43 © 2015 SOLARWINDS UPDATE STATISTICS example.dbo.Shipment (ShipStatus) go New Execution Plan
- 44. 44 © 2015 SOLARWINDS Case Study 3 » Current paychecks for specific employees SELECT e.empno, e.first_name, e.last_name, d.dname,l.street_address,l.city, l.state_province, l.postal_code FROM emp e INNER JOIN dept d ON e.deptno = d.deptno INNER JOIN loc l ON d.location_id = l.location_id WHERE (e.first_name = @first or e.last_name = @last) AND EXISTS ( SELECT 1 FROM wage_pmt w WHERE w.empno = e.empno AND w.pay_date>= DATEADD(day,-31,GETDATE()) ) » Execution Stats – 6681 Logical I/O » Average Execution - 15.012 seconds » Resource - 99% CPU
- 45. 45 © 2015 SOLARWINDS Execution Plan
- 46. 46 © 2015 SOLARWINDS SQL Diagramming select count(1) from wage_pmt where pay_date >= DATEADD(day,-31,GETDATE()) 109568 / 1027192 = .106 select max(cnt), min(cnt) from (select last_name, count(1) cnt from emp group by last_name) 128 / 6848 = .018 – max 64 /= 6848 = .009 – min emp dept wage_pmt 1000 1 150 1 .02 .10 loc 1 9 .009
- 47. 47 © 2015 SOLARWINDS New Execution Plan » create index IX_EMP_LN on emp(last_name) » Create index IX_EMP_FN on emp(first_name
- 48. 48 © 2015 SOLARWINDS Wage_Pmt table – no index on empno
- 49. 49 © 2015 SOLARWINDS Better Execution Plan » create index IX_WAGE_EMPNO on wage_pmt(empno) » » » » » » » » » » » Execution time = 0.546 seconds
- 50. 50 © 2015 SOLARWINDS Best execution Plan Covered Index: create index IX_EMP on emp(empno, last_name, first_name, deptno) Time: 0.066 secs
- 51. 51 © 2015 SOLARWINDS Monitor » Monitor the improvement § Be able to prove that tuning made a difference § Take new metric measurements § Compare them to initial readings § Brag about the improvements – no one else will » Monitor for next tuning opportunity § Tuning is iterative § There is always room for improvement § Make sure you tune things that make a difference » Shameless Product Pitch - DPA
- 52. 52 © 2015 SOLARWINDS Improved Performance Average Wait Time per Execution for SQL Statement Current Paychecks by Name | GIBSON:5000 March 15, 2015 Daily Time Range: 5:00 PM-12:00 AM Logical Reads = 1198 Execution Time = .066 seconds
- 53. © 2015 SOLARWINDS Takeaway Points » Tuning Queries gives more “bang for the buck” » Make sure you are tuning the correct query » Use Wait Events and Response Time Analysis § Locking problems may not be a Query Tuning issue § Wait Events tell you where to start » Use “Execution or Query Plans” » SQL Diagramming - “Get it right the First Time” § Query Tuner Tools can mislead you » Monitor for next tuning opportunity
- 54. www.solarwinds.com/dpa-download/ Resolve performance issues quickly—free trial Try Database Performance Analyzer FREE for 14 days Improve root cause of slow performance Quickly identify root cause of issues that impact end-user response time See historical trends over days, months, and years Understand impact of VMware® performance Agentless architecture with no dependence on Oracle Packs, installs in minutes
- 55. © 2014 SOLARWINDS WORLDWIDE, LLC. ALL RIGHTS RESERVED. The SOLARWINDS and SOLARWINDS & Design marks are the exclusive property of SolarWinds Worldwide, LLC, are registered with the U.S. Patent and Trademark Office, and may be registered or pending registration in other countries. All other SolarWinds trademarks, service marks, and logos may be common law marks, registered or pending registration in the United States or in other countries. All other trademarks mentioned herein are used for identification purposes only and may be or are trademarks or registered trademarks of their respective companies. Thank You! Q & A