An AppDynamics Business White Paper

October 2013

End-to-End Performance Monitoring

End-to-End Performance Monitoring

Application architectures have changed significantly over the past decade. The
days of simple monolithic applications are drawing to end in favor of applications
that combine service-oriented architectures (SOA), big data, cloud computing,
in-memory data caches, AJAX, mobile devices, and more. This paper describes
the core principles of troubleshooting modern, distributed applications and the
key features and functionalities to look for in a performance monitoring solution
for these applications.
The Problem
If youve ever been involved in troubleshooting an application outage, you know how
painful of a process it can be. It usually starts with an email, page, or text message to join
a conference call because some application you are involved with is experiencing issues.
When you call into this conference line there are usually a number of other support personnel
joining or already on the call discussing the issue at hand. People start checking log files,
probing databases, looking at network data, analyzing message queues, testing from different
network segments, etc., all in an effort to isolate the cause of the problem.
This method of troubleshooting used to suffice when applications were simpler and
performance was less critical. Today, troubleshooting a performance issue this way can take
hours, and your end users dont want to wait that long. You need an application performance
management solution thats designed for the challenges modern applications are facing.

The Solution
From a high-level perspective, the troubleshooting process is pretty straightforward; detect,
isolate, repair. Lets look at what this means, practically speaking.
Detect: The definition is simple enough: identify that there is a problem. The reality is that
very little monitoring happens from an end user perspective. IT organizations have more
information about whats occurring inside of their data center than what the end user is
experiencing on the outside. To make matters worse, modern applications rely on remote
services for some of their functionality. This means that even if your application is running
fine, a slow or unavailable third party service could be degrading your end user experience.
Problem detection is not as easy as it seems. The last thing you want to do is leave it to end
users to call into your help desk and tell you that your application is broken.
Isolate: Most application problems can be tracked down to a single problematic component
or server. This is the step that can cost an organization countless hours of troubleshooting
and downtime without the right people, processes and technology in place. Each of these
components is equally important, even though process is often omitted while its true that
it takes the right people and the right tools to quickly identify issues, its also critical to make
sure you have the proper processes in place to fully streamline your troubleshooting efforts.
Repair: Fix it! This process can be fast or slow depending on the root cause of the issue. The
most efficient way to repair applications is by automating known fixes to known problems.
Another important aspect of repair is gathering the right troubleshooting information when
a known problem is detected. For example, when there is a code deadlock the immediate
resolution is to take a thread dump (so that you can identify the locking and blocked code),
and restart the locked application node to restore service. This is a temporary solution until the
software defect can be fixed in a later release, but it is a perfect opportunity to use automation
for lightning-fast service restoration.
Now that weve created these high-level buckets, lets explore the role that end-to-end
application performance monitoring and management can play for each one. Well start with
the end users perspective and work our way to the backend of the data center.

An AppDynamics Business White Paper

End-to-End Performance Monitoring

End User Monitoring

Description: With End User Monitoring you should see exactly how your end users experience
your application starting at the first byte time. Its imperative to understand application
performance at the browser level, with granular browser response time snapshots broken
down by device, geographic region, browser type and more. In addition, the ability to drill
down from a browser snapshot to the corresponding call stack trace in the application server
gives you complete visibility from the browser to the application server to the database.
Important Functionality:

Monitor page, AJAX and iFrame performance

Navigation timing API measurements
Geographic/Location performance
Device & OS performance
Browser Type & Version performance

Benefits:

Trace a transaction from the end user to the database

Monitor end user experience by geographic region, device and browser type
Get visibility into transactions, pages, AJAX requests and iFrames
See all JavaScript errors at the line of code level

Important Considerations:
Designed for extreme scalability up to billions of devices and sessions
Easy to install and use
Fully integrated from browser to app server and database

Navigation Timing API for a real user browser snapshot.

Geographic dashboard showing errors and poor performance by region.

An AppDynamics Business White Paper

End-to-End Performance Monitoring

Application Performance Monitoring

Description: Application Performance Monitoring (APM) has many definitions. In this paper,
well use Gartners definition from their 2012 magic quadrant for APM:
1. End-user experience monitoring - the capture of data about how end-to-end application
availability, latency, execution correctness and quality appeared to the end user
2. Runtime application architecture discovery, modeling and display - the discovery of
the various software and hardware components involved in application execution, and
the array of possible paths across which those components could communicate that,
together, enable that involvement
3. User-defined transaction profiling - the tracing of events as they occur among the
components or objects as they move across the paths discovered in the second
dimension, generated in response to a users attempt to cause the application to execute
what the user regards as a logical unit of work
4. Component deep-dive monitoring in an application context - the fine-grained monitoring
of resources consumed by and events occurring within the components discovered in the
second dimension
5. Analytics - the marshaling of a variety of techniques (including behavior learning engines,
complex-event processing (CEP) platforms, log analysis and multidimensional database
analysis) to discover meaningful and actionable patterns in the typically large datasets
generated by the first four dimensions of APM
Important Functionalities:
Self-updating application flowmap showing you how different components interact
Automatic detection and tracking of Business Transactions (functions performed within an
application that have an associated downstream flow of activity)
Intelligent dynamic baselines for each Business Transaction
Full call stack traces with timing correlation for distributed transactions
Monitoring all exit points of your application
In-context drill down between end user, application server, and database server
Easy to use for rapid adoption and fast time to value.
Zero or minimal initial and ongoing configuration
Benefits:

Wide and deep visibility

Fast time to value
Reduce MTTR from hours to minutes
Improve collaboration between development, application support, and operations
Accelerate the software development lifecycle in agile environments
Fastest possible restoration of service using automation

Important Considerations:

<2% overhead to reduce chance of impact on production workload

Automatic instrumentation for complete code-level diagnostics
Automatic correlation of data anomalies to poor-performing transactions
Automatic scoring of Business Transactions based on deviation from normal to quickly
detect problems

An AppDynamics Business White Paper

End-to-End Performance Monitoring

Dynamic application flow map showing all application components.

Dynamic application flow map showing all application components.

Deep application diagnostics showing call stack with timings and exit calls available for
drill-down inspection.

Slowest remote service calls with associated server-side snapshots.

An AppDynamics Business White Paper

End-to-End Performance Monitoring

Database Monitoring
If you want true end-to-end monitoring, its not enough to just monitor the end users and the
application servers. You also need to monitor the backend databases from the inside out so
that you can quickly remediate problems within the database tier.
Description: Your database monitoring solution should continuously monitor database
instances and provide deep visibility into SQL execution, resource consumption and wait
states to help users rapidly pinpoint the root cause of bottlenecks. The ideal approach is to
remotely monitor production environments with less than 2% overhead so users can manage
database performance 24/7.
Important Functionality:

SQL & stored procedures

Explain plans
Wait states
Resource consumption
Database objects
Schema statistics
User sessions
Data files
Change events
Locking and blocking analysis

Benefits:
Rich granular data provides insight into root cause
Application and database support can see the same information which enables
cross-silo collaboration
Production-ready monitoring
Important Considerations:
Remote data collection (agents incur unnecessary overhead in the DB tier)
Less than 2% overhead

Database monitoring high level statistics including hardware metrics and database wait states.

An AppDynamics Business White Paper

End-to-End Performance Monitoring

Database locking tree showing blocking and

blocked sessions. From each session you
can drill down to the associated queries.
Database execution plan (explain plan) for a long-running query with associated wait states.

Extensible Monitoring Platform

The reality of monitoring from end to end is that there may be application and/or
infrastructure components that are not supported by specialized monitoring tools within your
organization. This is where an extensible monitoring platform is required to fill in your blind
spots and provide actionable intelligence.
Description: A proper extensible monitoring platform is more than just a data collector it
is a flexible and scalable platform capable of turning data into intelligent information. Its
a platform where you can build your own monitoring extensions to collect data from any
component of your infrastructure (databases, web servers, caches, queues, network devices,
hardware, etc.). Its a platform that can integrate with other tools to perform actions or even
auto-scale your application in your favorite cloud. Its a platform that you can use to build
custom alerts, remediation mechanisms and visualizations for your application.
Important Functionality:
Capable of integrating ANY data
Automatically baselines ALL data
Able to create alerts and remediate based on imported data
Benefits:
Plugs gaps in visibility
Allows support for bleeding-edge technology
Automatic baselines show which metrics have deviated from normal values during
troubleshooting
Important Considerations:
Community-built integrations
Vendor supported integrations

Operating system process level details gathered using a pre-built community integration.

An AppDynamics Business White Paper

Putting It All Together

End-to-end monitoring starts with the end user experience, continues into the data center
and follows the user transaction all the way to the backend. A good monitoring platform will
automatically detect code issues, slow transactions, memory issues, resource contention, and
poor-performing services.
Rapid detection of problems makes it easier to quickly isolate root cause. Most problems are
eventually tracked down to a single application component or server, so its important that
your monitoring strategy and tooling provides visibility into every application component to
increase the likelihood of discovering the problem.
Once the issue has been detected and isolated to the problematic node its time to repair the
platform and restore service. An intelligent APM solution knows if problems are isolated to a
single node, multiple nodes, or even whole services. Having this information enables actions
to be performed automatically to only the impacted components. This capability is called
Application Run Book Automation (ARBA), and it allows you to rapidly restore service by
automating common tasks.

Conclusion
Your end-to-end monitoring strategy should be based around a holistic monitoring
architecture that is designed to minimize visibility gaps. The tools that support your strategy
need to be capable of working effectively and efficiently in the most demanding modern
application environments. Selecting the proper technology will enable your organization to
detect, isolate, and repair faster than ever before.

About AppDynamics

Try it for FREE

AppDynamics is the next-generation application performance management solution that

simplifies the management of complex, business-critical apps. No one can stand slow
applicationsnot IT Ops and Dev teams, not the CIO, and definitely not end users. With
AppDynamics, no one has to tolerate slow performing apps ever again.
Visit us at www.appdynamics.com.

www.appdynamics.com

AppDynamics, Inc. 303 Second Street, Suite 450 North Tower, San Francisco, CA 94107 www.appdynamics.com
Copyright 2013 AppDynamics, Inc. All rights reserved. The term APPDYNAMICS and any logos of AppDynamics are trademarked or registered trademarks of AppDynamics, Inc.