Application Acceleration: Faster Performance for End Users

Preview:

Citation preview

Grab some

coffee and enjoy the pre-show banter

before the top of the

hour!

H T Technologiesof 2016

HOST:EricKavanagh

THISYEARis…

THELINEUP

ANALYST:

DezBlanchfieldDataScientist,TheBloorGroup

GUEST:

BillEllisSalesEngineer,IDERA

ANALYST:

RobinBloorChiefAnalyst,TheBloorGroup

INTRODUCING

RobinBloor

Robin Bloor, PhD

Application

Performance

& Service

Levels?

Performance Management

Performance can be one-dimensional but service levels are MULTI-DIMENSIONAL

The Diagram of Everything

All IT Assets and Relationships

TIME

Performance & SLA Complexity

Complexity is about variety, structure and constraints – and the numbers

The numbers always increase

The constraints are determined by the business processes and business goals

A Business Perspective

  In some systems, the optimal response time is one tenth of a second the human response time.

  Software can have higher expectations and may require a faster response time.

  In some situations – market situations - being first is where the business value is. This may apply to the business process rather than an app.

The Big Picture

SystemManagement

ServiceManagement

BusinessPerformanceManagement

OptimizedServices

IntegratedServices

StandardizedServices

FragmentedServices

Service Management Automation

ServiceManagem

entS

trategy

Instrumentation

The sub-optimization problem…

Clearly nothing much can be achieved without sophisticated

tools?

INTRODUCING

DezBlanchfield

@dez_blanchfield

PERFORMANCE&SLA’S

OURNEW“ALWAYSON”WORLDNOWDEMANDS100%UPTIME,

BUTBEINGUPDOESN’TALWAYSMEANYOUAREMEETINGSLA’S

@dez_blanchfield

THENEEDTOMANAGE

APPLICATIONPERFORMANCE

ANDAVAILABILITYSLA’SHAS

UNDERGONEAFUNDAMENTAL

SHIFTINTHELASTDECADE

@dez_blanchfield

Onceupona2mewe’dhaveasingleserverrunningmul2pleservicesandmonitoringlifewasrela2velystraightforward..

§  Istheserverupandresponding

§  IstheoperaMngsystemonline

§  AretheapplicaMonsupandrunning

§  Aretheappservicesresponding

§  Canusersconnecttotheservices

§  Isthehelpdeskphoneringing

§  Who’sturnisittogetdoughnuts

WHENTHINGSWEREALOTSIMPLER

@dez_blanchfield

The2mewhenafewgoodapplica2on,systems,networkanddatabaseadminscouldmeetperformanceSLA’sislonggone..

§  By1996databasedrivenappstransformedwiththeInternetboom

§  Intheearly2000’swehadtocopewiththeDotComboom

§  ThenWeb2.0causedyetanotherupwardspikewithsmartphones

§  In2016we’renowfacedwithaquagmireofCloud,BigData&Mobility

§  Nextweneedtoaddressreal-MmeM2M,SmartBuildings&SmartCiMes

§  ThenthenextbigwaveistheheadychallengeoftheInternetofThings

§  Andwe’realsonowbuildingA.I.&CogniMveCompuMngintoeverything

WECAN’TDOITBYHANDANYMORE

@dez_blanchfield

THECOMPLEXITYOFTODAY’S

STANDARDAPPLICATION

STACKISFARBEYOND

HUMANCAPABILITIES

@dez_blanchfield

Applica2ons&databases,havebecomesolarge,andsocomplex,thatmanagingperformanceisascienceinitself..

§  On-site/Off-site/DataCenters

§  Physicalservers/Virtualservers

§  Cloud/IaaS/PaaS/SaaS

§  SDN/NFV/MicroServicesvsApps

§  Kubernetes/Docker/Swarms

§  TradiMonalDB/HadoopBigData

§  Mobility/BYOD/M2M/IoT/I.A.

HYPERSCALE,COMPLEXITY&SPEED

@dez_blanchfield

MANAGINGPERFORMANCE

SLA’SATTODAY’SSPEEDAND

SCALEREQUIRESTOOLSAND

SYSTEMS,BYDEFAULT!!

@dez_blanchfield

Thisisahighlevelapplica2ondesigndiagramfortheOpenStackopensourcesoGwaredefinedCloudstack..Yes..Highlevel!!

DONOTTRYTHISATHOMEKIDS!!

@dez_blanchfield

BIGDATATECHNOLOGYLANDSCAPE

@dez_blanchfield

EXPLODINGVERTICALLANDSCAPES-1/2

@dez_blanchfield

EXPLODINGVERTICALLANDSCAPES-2/2

@dez_blanchfield

NOHUMANCANENSURESLA’SHERE

INTRODUCING

BillEllis

© 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

INTRODUCTION TO PRECISE APPLICATION PERFORMANCE PLATFORM Every important transaction interacts with the database

2 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

Capacity Planning / Processing Efficiency

Meeting SLAs

Precise Application Performance Platform

Industry First PMDB

3 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

WE WILL COVER

§  Process of managing performance §  Context – Precise’s secret sauce §  Useful visibility for all stakeholders §  Customer Case Studies

© 2015 IDERA, Inc. All rights reserved. Proprietary and confidential. 4

PROCESS OF MANAGING PERFORMANCE §  Reduce Time & Frustration

TRADITIONALMONITORING

DB Metrics

OS/Net Metrics

App Metrics

OS/Network support

Operations Application

Support

Database Administration

Service Desk

Storage Management

Application Development

Client & Web

Metrics

Storage Metrics

Client

THEREHASGOTTOBEABETTERWAY

ABETTERWAY…

- Acorrelatedendtoendviewofhistoricalperformance

-  Shareinforma6onamongdiversetechnicalteams

- Automa6ciden6fica6onofproblemareas

-  Easyrepor6ngonbusinessimpactandpriority

- Andaknowledgebasetoprovideac6onableadvice

OS/Network support

Operations Application

Support

Database Administration

Service Desk

Storage Management

Application Development

Client

DB Metrics

OS/Net Metrics

App Metrics

Client & Web

Metrics

Storage Metrics

Storage Management

Service Desk

Database Administration

Application Development

OS/Network support

Application Support

Operations

Executive (CIO/CEO/LOB)

Executive (CIO/CEO/LOB)

Precise PMDB

THEPRECISEWAY

Business

Applications

Experts

•  CIO •  Line of Business •  Business Analyst

•  Help Desk •  Performance Team •  Service Delivery

•  DBA, Storage •  SAP, Oracle Ebiz •  J2EE, .NET

THERIGHTVIEWOFPERFORMANCEANALYSIS

Precise PMDB

Storage Management Database

Administration

Application Support

Performance Team

LOB

Help Desk

Executive (CIO/CEO)

Analyst

APM–TRANSACTIONPERFORMANCE

Business

Applications

Experts

• CIO • Line of Business • Business Analyst

• Help Desk • Performance Team • Service Delivery •  SAP, Oracle Ebiz,

PeopleSoft •  J2EE, .NET •  Oracle, SQL Server,

Sybase, and DB2

Investing Transaction

Storage Management

Application Development

Performance Team

LOB

Help Desk

Executive (CIO/CEO)

Analyst

Precise PMDB

Application Support

Database Administration

© 2015 IDERA, Inc. All rights reserved. Proprietary and confidential. 9

CUSTOMER CASE STUDIES §  Getting back to business quickly

CASESTUDIESNew Storage Unit 1

Steelcase Inc. 2

Global Application going in Production 3

Pack and Ship not meeting business SLA 4

SAP transaction breaking SLA 5

PeopleSoft Optimization 6

11 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential. 11

Database Intelligent

§  Auto capture real end-user transaction flows §  Minimal configuration §  Flow across VMware environment; auto detection of contention

§  Every important business transaction interacts with the database; the database is the foundation of performance §  Precise will help you get the very best performance §  Upgrade from sub-second to sub-millisecond

Auto-discover Transactions

Production monitoring

§  All users transactions identified by name & location §  Translates URL/URI to transaction name for SAP, PeopleSoft, and eBiz so IT can talk to the business §  Low resource overhead

Technology - what makes Precise different?

PRIVATE CLOUD

© 2015 IDERA, Inc. All rights reserved. Proprietary and confidential. 12

FINAL POINTS §  We want your business

VALIDATINGDATAWAREHOUSEUPGRADECase Study – An Automobile Manufacturer Problem Background An Automobile Manufacturer invested in a new VMAX SAN to improve the performance of its Data Warehouse application. After installation, An Automobile Manufacturer grew concerned over significant increased disk activity. Existing storage monitoring tools showed an increase in disk contention, prompting An Automobile Manufacturer to begin planning an expensive SAN roll back. Investigation Precise was purchased to manage transaction performance across all tiers of the supporting infrastructure. An Automobile Manufacturer used Precise to compare end-to-end transactions and specific supporting application, database and storage performance details before and after the VMAX upgrade. Findings While contention on the SAN was up, looking at the complete transaction - peak throughput increased from 400MB/sec to 700MB/sec and average I/O wait time decreased from 38% to 17% of overall database time. Summary Precise was able to validate the success of the VMAX upgrade by focusing on transaction performance and proved that the increased disk activity was a direct result of increased throughput – not a transaction performance issue. An Automobile Manufacturer additionally found that isolated SAN (or tier) monitoring can be misleading and not reflective of the true transaction performance. An Automobile Manufacturer was able to mitigate the risk and the cost of a more than $500,000 roll back.

ANAUTOMOBILEMANUFACTURER

DRIVINGDOWNSAPOPERATINGCOSTSCase Study – Steelcase Inc.Problem Background A key end user (CHOUSE) complained about poor response time while running a custom transaction (ZPP_CM25). The problem surfaced after a major upgrade of SAP and directly impacted overall production planning and inventory control. Investigation Steelcase uses Precise to manage the performance of SAP. After the end user complained about performance, Steelcase was able to quickly filter and focus on the end user transaction.

Findings Steelcase was able to isolate the bottleneck to the database tier. A close look at the database query identify a high number of executions with an extremely efficient response time (.002 seconds). This pointed to a coding issue, not a database issue. A quick look at the application code identified that a program loop generated the excessive calls. A code change to issue one query and loop thru the result cursor saw a tremendous improvement in end user response time (average backend database time decreased from over 53 seconds to 4.5 seconds per transaction).

Summary Steelcase was able to quickly isolate the root cause and verified that the code change had a significant impact. End user productivity increased significantly. CM25 is an SAP transaction responsible for capacity leveling of production planning and inventory control. The improvement resulted in better management of inventory and more accurate production planning, enabling Steelcase to significantly drive down operating costs.

STEELCASEINC.

17 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

Steelcase Inc.

18 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

Steelcase Inc.

Back to case study page

ANELECTRICUTILITYCase Study – An Electric Utility Problem Background An Electric Utility in the west is Precise for SAP customer. It is a privately held energy company with a portfolio of competitive and regulated energy companies. Its portfolio of competitive businesses consists primarily of An Electric Utility, a retail electricity provider with more than 2 million customers in Texas, and Luminant, which is engaged largely in power generation and related mining activities, wholesale power marketing and energy trading. An Electric Utility was experiencing regular disruptions to transaction performance across their SAP application stack which had a direct impact on their customer’s satisfaction and a costly impact to the business. Investigation An Electric Utility purchased purchased Precise to manage the performance of SAP. Within hours of deployment, Precise quickly identified the root cause of several key performance bottlenecks. Findings Over the next several months, Precise was able to reduce the amount of SEV1 support calls by 25%. Summary Precise is now an integral part of problem resolution at An Electric Utility. Prorated to a full calendar year, An Electric Utility will save more than 1.9 million dollars with a 25% reduction in SEV1 support cases.

20 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

An Electric Utility / An Electric Utility

Increased DB time leadto high SAP queue time

Increased DB time is waiting on a table lock (BALDAT)

Log Purger holding lock

Web

J2EE Presentation

J2EE Logic

Database

Oracle

Un-monitored-1 Un-monitored-2

27 Executions 50 Executions

Average Time: 4.4 sec

Average Time: 5.7 sec Average Time: 4.2 sec

Average Time: 0.7 sec Average Time: 0.4 sec

Average Time: 3.3 sec

Average Time: 5.7 sec

Critical Alerts: 4

Critical Alerts: 3 Critical Alerts: 27

Critical Alerts: 0 Critical Alerts: 4

Critical Alerts: 5

Critical Alerts: 18

Average Time � Show Size by:

Layers þ Average Time

¨ SLA

¨ Critical Alerts

¨ Memory

¨ CPU

þ

Web Metrics

Java Metrics

Database Metrics

Storage Metrics

Web

J2EE Presentation J2EE

Logic

Database

Oracle

Un-monitored-1 Un-monitored-2

SLA

Availability

Monitoring Status

Alerts

Configure Health

Size by: Average Time �

Virtualization event May 1,11 10:37 vm-sap-portal on oryesx04.precise.com has moved to oryesx05.precise.com

49 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

PACK&SHIPNOTMEETINGBUSINESSSLACase Study – Frustrated Users Problem Background Warehouse operations occasionally had screen freezes that would not allow the shipping department to meet the business SLA of shipping all orders received that day by 3 PM. Investigation Precise pinpointed the underlying SQL statements. The detailed history showed very different execution times. Tracked to flip-flopping execution plans. Findings One plan is fast; the other is slow. Summary Lock in the fast execution plan to deliver consistent performance to the application users.

51 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

52 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

53 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

54 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

PRECISEFINDINGSCase Study – Global Application going in Production Problem Background New version of application with worldwide users going into production. Web servers maxed out at 100% CPU. Investigation Most transactions and code executed to expectations. However, outliers could prevent the application from scaling. Findings Precise Findings automatically pinpointed bottlenecks. Summary Precise was instrumental in isolating and resolving performance bottlenecks so that Go-Live date could be met.

56 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

57 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

58 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

BEFORE&AFTERMEASUREMENT

60 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

61 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

62 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

63 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

Storage Devices implicated in performance contention

64 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

Precise Findings automatically identify performance bottlenecks and Precise makes recommendations.

Tempdb is a type of I/O wait as SQL statements require additional space to process an interim result set.

65 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

Recommendation – investigate storage devices.

66 © 2016 IDERA, Inc. All rights reserved. Proprietary and confidential.

Tempdb lives on device 32A8. It is 100% SATA.

SM51 has the highest average response time. Click here to drill into this transaction.

XSJONNAL ran this transaction

Click here to analyze the transactions database time

Most of the time is spent in the backend database and the transaction has mostly failed its SLA

Top statements for XSJONNAL execution of SM51

Click here to tune the top statement

Precise recommends an index to improve SM51

Click here to simulate the impact of adding this index

Significant performance improvements for SM51 and many other statements

PRECISEFINDINGSPeopleSoft Optimization in Production Problem Background Performance is good; could it be better? Investigation Review environment for optimization opportunities and question about AJAX. Findings AJAX is not a significant contributor to performance, opportunities in Tuxedo and the database. Summary Many shops get super interested in performance when there is a problem. Another approach is to treat like a practice.

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

PEOPLESOFT HR SCREENS

TheArchiveTrifecta:•  InsideAnalysiswww.insideanalysis.com•  SlideSharewww.slideshare.net/InsideAnalysis•  YouTubewww.youtube.com/user/BloorGroup

THANKYOU!

Recommended