5 IT Trends That Reduce Cost And Improve Web Performance - A Forrester and Gomez Webinar

5 IT Trends That Reduce Cost and Improve Web Performance

Jean Pierre Garbani - VP, Principal Analyst Forrester Research

Imad Mouline - CTO, Gomez

IT Business and Management Trends

JP Garbani

Vice President and Principal Analyst

Forrester Research, Inc.

Agenda

1) Key trends: the role of IT in the Enterprise

2) Th t IT l d2) The current IT landscape

3) Solutions for a leaner future

4) Performance issues

5) Recommendation and conclusion

4 Entire contents © 2008 Forrester Research, Inc. All rights reserved.

IT Role As Defined By Business Users


IT Key Priorities


Productivity Equals Application Performance

The key to productivity in interactive applications is

response time:response time:

• .1 second is instantaneous

• 1 second is when a response time is perceived1 second is when a response time is perceived

• 10 seconds is when users lose focus on the interaction

Basic rules

• Simple and frequent tasks: 1 second or less

• Common tasks: 2 to 4 seconds

• Complex tasks: 8 to 12 seconds

Source: Usability Engineering by Jacob Nielsen (Morgan Kaufman San Francisco)


Source: Usability Engineering by Jacob Nielsen (Morgan Kaufman, San Francisco)

But… Trends and Challenges

• User’s performance expectations are rising as IT budgets are squeezed

• Application business value and complexity increasesM lti ti li ti th– Multi-tier applications are the norm

• Time to value reduced through re-use, web services and SOA

• Interoperability, portability, device diversity and size leads to configuration, deployment and production issues that impact performancesissues that impact performances

• Outsourcing, hosting, cloud computing and virtualization opacity present new challenges


p y p g

Web Application Complexity and User Expectationsand User Expectations


Web Application Delivery Chain

Client andEmbedded

devices

SoftwareSolutions

e g SOAdevices e.g.SOA

Business ServicesPervasive Interactions

Business MetaData CoreProcesses, Policies. Organic ITUnified

Structure, etc. Organic ITUnifiedCommunications

DataVoiceMobile

Network InfrastructureCl d C ti

TransactionVideo

Collaboration


Infrastructure e.g. Cloud Computing Virtualization

Translates into Performance Issues

• As the size of software increases, the number of problems per KLOC increases proportionallyproblems per KLOC increases proportionally

• 74% of problems are reported by the end users through the service desk, not detected by infrastructure management

• An average of 6 service desk calls are needed to identify the problem owneridentify the problem owner

• Typical SLA: 95% of level 1 problems are answered in 2 hoursin 2 hours

IT Operations are bursting at the seams


Source: Forrester Research

Reducing Diversity For a Leaner IT

• Outsourcing legacy applicationsOutsourcing legacy applications

• Standardizing infrastructures using virtualization

• Using packaged applications and SOA derivatives• Using packaged applications and SOA derivatives

• Using SaaS for selected management functions


Application Management Pain Points

26%Poor or erratic Java EE performance

Rank the pain level caused by the following situations (high to severe)

26%Capturing and quantifying actual end user experience

30%

Di ti bl ti hit t th

Improving IT responsiveness by arming techs with datarather than relying on end user description of problems

31%

31%Lacking accurate tools for end to end transactionmanagement

Diagnostic problems across n-tier architectures than spanmainframe and distributed

31%

37%Determining root cause of poor application performance

management


Ziff Davis survey of 1190 IT professionals

Reducing MTTR and MTBF

Productivity costs

Recovery costs

Extra costs

Cost of data recovery

+Cost of external

Duration of outage

xNumber of Lost

costscosts

ProblemIdentification

external support people

affectedx

Average % lost productivity

x

Duration of outage

x

Lost revenue

ProblemIdentification Identification

+Analysis and

resolution +

Validation t ti

xAverage cost of employee

+Average

employee

Percentage of unrecoverable

businessx

Average revenue per

Identification +

Analysis and resolution

+Validation

testingp y

overtime cost revenue per hour

Costs before deployment Costs of downtime and brown-outs in production

testing


Source: Forrester Research, Inc.

Evaluating the Black Box Performance

ResponseTime

μ

Fixed Parameters(System Design)

Time

λService time (μ) is the

characteristic end result of a workload (λ) applied to

Black Box

Workload Output

( ) ppa given processing

capacity

Variable Parameters(System Configuration, Application)


What we Should do:• Validate application architecture at the development stage and

create a performance hypothesis

• Validate performance hypothesis in QA:– Using a realistic environment

Using end to end monitoring tools– Using end to end monitoring tools

• Understand the workload variations– What are the normal performance levels for a given periodWhat are the normal performance levels for a given period

– Distinguish between real problems and transient variations and anomalies

• Consider efficient alternatives such as SaaS to in house testing and monitoring


Recommendation: Towards a Leaner IT

• IT evolution drivers are economical rather than technicalImproving workforce productivity and IT efficiency are the key–Improving workforce productivity and IT efficiency are the key themes

• Simplification and standardization of infrastructures will be the dominant themes of the next five years

• Any initiative such as consolidation, standardization, IT governance improvement or service management requiresgovernance improvement, or service management requires preparation

–Understanding workloads and performance issues are key

• Productivity and efficiency objectives are only reached when we consider all alternatives to perform a given performance management function


g

Conclusion

• The major driver for IT management and IT OperationThe major driver for IT management and IT Operation evolution is to improve business productivity as efficiently as possible

• Many current management solutions are still very muchMany current management solutions are still very much technology silo oriented and ignore the end user perspective

• Most of the current tools compound the issue by adding complexity of deployment and operationcomplexity of deployment and operation

• The consensus is that IT service delivery is better served not only by improved IT management processes, but also by

it i l ti b d th S S d l th t id thmonitoring solutions based on the SaaS model that provide the right data for an optimized operational cost


Thank you

Jean-Pierre Garbani

+1 617/613-6091

jgarbani@forrester [email protected]

www.forrester.com

19 Entire contents © 2008 Forrester Research, Inc. All rights reserved. Entire contents © 2005 Forrester Research, Inc. All rights reserved.

5 IT Trends That Reduce Cost and Improve Web Performance

Jean Pierre Garbani - VP, Principal Analyst Forrester Research

Imad Mouline - CTO, Gomez

About Gomez

We help customers ensure quality experiencesWe help customers ensure quality experiences for their Web and mobile users

Optimal performance and availability

Quality operation across all browsers and mobile devices

Rapid notification of issues with actionable diagnostics

Insight into how these issues affect businessInsight into how these issues affect business

SaaSPlatform

Over 2,500 Customers

Solid Business Performance

Recognized as Industry LeaderPlatform

• No installation• Rapid startup & payback

• World’s most

Customers• 1,250+ enterprise customers- 12 of top 20 US sites

Performance• Five consecutive years double digit revenue growth

• 44% revenue

Industry Leader• Forrester: “The leader in Web Experience Management”

World s most comprehensive testing network

• 1,250+ SMB customers

growth 2008 • Gartner: “A cool company”

The Traditional View of Web Application Delivery

…user is happy

Systems management tools: “OK”

Users

Load Balancing

W b li ti

Web Servers

App Servers

Web application

Storage

DB Servers

Mobile ComponentsVirtualized Services

Traditional zone of control

The Challenge of Ensuring Quality Web Experiences

The “Web 2.0 application delivery chain”


…user is NOT happy

22 333rd Party/CloudWeb Services

Local ISP

Load Balancing

Users22 33

Web Servers

App Servers

Internet

MajorISPStorage

DB Servers

M bil

Content DeliveryNetworks

Mobile Carriers


11 Networks

Zone of customer expectationTraditional zone of control

Managing the Performance of Virtualized Applications

d / l d


…user is NOT happy

3rd Party/CloudWeb Services

Local ISP

Load Balancing

Users

CompanyWeb Servers

App Servers

• Broadcasting conglomerate delivering 400+ websites

Situation• IT/ O ti d t ti l itInternet

MajorISPStorage

DB Servers

M bil

• IT/ Operations needs to continuously monitor performance of virtualized applications on shared infrastructure

• Business owners want assurances that virtualization

Content DeliveryNetworks

Mobile Carriers


Business owners want assurances that virtualization will not negatively impact customer experience

• IT/Operations wants to validate that workload variations for virtual services do not impact end-users11 Networks

Zone of customer expectationTraditional zone of control

Validate Virtual Application Performance for End‐Users

Virtual 2

Virtual 1 • Consolidated applications to virtualized servers

• Validated performance for end-

Virtual 3p

users across multiple VM hosts

• Alerted to performance issue because of increased workload onbecause of increased workload on “Virtual 3”

• And, validated end-user performance for b siness before &First byte time are a key health

Performance for all 3 virtual resourcesMonitoring shows the

performance for business before & after launch

First byte time are a key health indicator that virtual resources are “underwater”

New & heavier site released on virtual hardware

Monitoring shows the business that the larger site is delivered at a faster speed and with improved virtual hardware p pconsistency for end-users on virtual hardware

Work Effectively with Outsourced Providers

Ensure SLA Compliance & Performance with Key Vendors

Issue #1Issue #2

• Quantify the benefits of technology and outsourcingoutsourcing decisions (such as CDNs, cloud resources, and

Vendor A’s product color swatches fail to load after trying for 30 seconds

resources, and infrastructure)

• Isolate the root cause of slow performance for key providers

Product images delivered using Vendor B, a CDN, are loading slowly and show long first byte times

• Ensure SLA compliance

g y

Managing Performance for End‐Users

Breaking open the Performance ‘Black Box’

D d P • Distinguish between real problems and anomalies

Decreased Page views AND increased object errors

• Real user monitoring captures the business impact of end-user performance issues

• Correlate transaction volume & end-user performance to understand overall system

Lower than expected page views indicates an issue for end-users

understand overall system health

Drill down diagnostics shows if theDrill down diagnostics shows if the issue is widespread across:• Geographies• Connection Speed or ISPs• Browsers• Infrastructure

Key Takeaways From Forrester

Validate performance hypothesis in QA:Using a realistic environmentUsing a realistic environment

Using end to end monitoring tools

Understand the workload variationsWhat are the normal performance levelsWhat are the normal performance levels

for a given period

Distinguish between real problems and transient anomalies

Many current management solutions are still very much technology silo

oriented and ignore the end user perspective

Most of the current tools compound the issue by adding complexity of p y g p y

deployment and operation

Consider efficient alternatives such as SaaS to in house testing and

it imonitoring

The Gomez Solution: Web Application Experience Management

The Gomez PlatformThe Gomez PlatformSelf‐Service Unified SaaS Portal

Web Performance

Business Analysis

Web Performance

Business Analysis• Compatibility

Cross‐Browser and Device Testing

Cross‐Browser and Device Testing

• High‐volume

Load and Performance

Testing

Load and Performance

Testing• Real User monitoring

Web Performance Management

Web Performance Management

• Web performance p y• Functionality• Performance

gbackbone load

• Real‐world Last Mile load

g• Last Mile monitoring• Backbone monitoring

panalytics

• Benchmarks• Business dashboard

HTML, Web 2.0, Streaming, Web services, AJAX, Flash, FLEX, Mobile, etc.

Dashboards, Metrics & Analytics

Dashboards, Metrics & Analytics

Recording & ProvisioningRecording & Provisioning

Education &Best PracticesEducation &Best Practices

APIs &Data Feeds APIs &

Data Feeds Alerting & DiagnosticsAlerting & Diagnostics

HTML, Web 2.0, Streaming, Web services, AJAX, Flash, FLEX, Mobile, etc.

World’s Most Comprehensive Testing NetworkWorld’s Most Comprehensive Testing Network162 countries 2,500 ISPs162 countries 2,500 ISPs

100+ commercial‐grade nodes and data centers

100+ commercial‐grade nodes and data centers

500+ combos of browsers and O/S

500+ combos of browsers and O/S

4,600+ supported mobile devicesdevices

4,600+ supported mobile devicesdevices

100,000+ consumer‐grade desktops

100,000+ consumer‐grade desktops

S f O C t

Find Out How Gomez Can Help You

Some of Our CustomersGomez Solutions help companies ensure their web applications perform

ti ll f ll b d ioptimally for all users, browsers, devices, and geographies.

Optimal performance and availabilityQuality operation across all browsers and mobile devices Rapid notification of issues with actionableRapid notification of issues with actionable diagnosticsInsight into how these issues affect business

L M li tLearn More online at :www.gomez.com

Speak To A Gomez Solutions ExpertToll Free: 877.372.6732

5 IT Trends That Reduce Cost And Improve Web Performance - A Forrester and Gomez Webinar

Technology

application15 entire

performance hypothesis

end usersthrough

web application complexity

application architecture

normal performance levels

thecharacteristic end

actual end user experience