Application Performance Management

When evaluating their IT investments, companies are focusing more and more on evaluating the “quality“ of IT from an end user’s perspective, i.e. its functionality but also its availability and performance. As a consequence, traditional monitoring parameters such as server availability, server usage or network availability are losing their importance. The only thing that counts for the end user is the performance at his desk. For this phenomenon the term End User Experience Management (EUE) was created. New monitoring techniques are able to capture this end user experience; even for large numbers of users and for many locations or Internet users (e-commerce, field staff, branch offices, etc.), thus providing an important contribution to business/service level management (TBM/SLM).

For this reason, End User Experience Management is considered to be just one discipline of Application Performance Management (APM). While user experience primarily is following the target to visualize the IT quality experienced by the user, comprehensive APM requires additional disciplines such as:

  • Impact Analysis: How many users and which business transactions are impacted by an availability or performance restriction? Since when? Where?
  • Fault Domain Isolation: Which IT domain is responsible for the constraint? Network or data center? Service provider? Citrix environment? The goal is a quick fault isolation leading to an efficient analysis and solution of the problem.
  • Deep Dive: A comprehensive APM solution in any case should provide an appropriate deep dive analysis feature for every fault domain (especially network and data center) to isolate the error.

Additional requirements for modern APM solutions:

  • User and transaction detection – A highly granular performance monitoring at user and transaction level enables a comprehensive performance analysis. Average performance monitoring data on various transactions or even whole applications prevent a real understanding for the problem causes. This feature differentiates APM solutions from traditional infrastructure performance tools.
  • Comprehensive coverage of monitoring techniques – Very often an APM initiative is started with synthetic monitoring of important applications to prove specific availability and SLA levels. Later on, real user monitoring for important business applications as well as deep dive analysis in the data center may be added. Finally, the cloud-based e-commerce platform has to be monitored from an end user’s perspective – APM always will be some kind of “journey” respectively an evolution process. A professional APM solution has to be able to cover this by various tightly integrated monitoring techniques.
  • Comprehensive coverage of technologies – Even if many applications today are web-based, especially in large companies there still will be a wide variety of additional application technologies such as SAP or mainframe. An APM solution has to provide coverage for a huge number of different technologies.

Triggers

The main trigger for the growing interest in user experience is higher “up-down” pressure. The reasons are evident. APM projects have a high corporate visibility – up to the highest board level. After all availability and performance data for business critical applications are captured, the results will be understood by non-IT personnel as well. Therefore, the IT organization will be forced to take action. It cannot hide any longer behind the relatively easy proven availability of single IT components (servers, database, network infrastructure), but has to take more and more accountability for the result at the end user’s desk. For this reason, it is no wonder that more and more APM projects will not be initiated by the IT, but by the business. The sales director for example calls for a guarantee of an appropriate performance quality of his field CRM solution and the web-based sales portal.

Future Market Trends

Reifegradmodell

Market analysts forecast significant market growth for Application Performance Management as well as SLM/TBM for the next years. This market development is fuelled by the fact that many companies have reached the second level of IT maturity (reactive) of the five-level IT maturity model (see chart) and now are trying to further develop their IT to reach level three (proactive) or level four (service-oriented).

amasol will support you with a comprehensive solution and service portfolio in the design and implementation of your specific APM initiative as well as the selection of the appropriate monitoring methods. In addition, we will assist you in implementing, operating and maintaining the corresponding solutions and accompany you on your “APM journey”.

APM Monitoring Techniques Overview

The following chart provides an overview on the typical methods, technologies and user scenarios for the major APM monitoring techniques as well as the products used by amasol.

Method Technology Short description Typical Use Cases APM solution
APM technologies
Synthetic Monitoring – periodic monitoring of selected reference transactions On premise robot Monitoring from workstations within the corporate network 24×7 availability and SLA monitoring for business applications Dynatrace SMPE, CA Unified Infrastructure Management
Synthetic Monitoring – periodic monitoring of selected reference transactions Cloud robot Monitoring from Internet monitoring points, at service providers and end users, typically as a Software-as-a-Service solution. 24×7 availability and SLA monitoring for e-commerce, web and mobile applications Dynatrace Gomez SaaS, CA Unified Infrustructure Management Cloud Monitor
Real User Monitoring (RUM) – Performance monitoring from an end user’s perspective, by passive monitoring of transactions carried out by “real” end users IP-flow-based Transaction monitoring and performance analysis by passive monitoring of the network data traffic Real user monitoring for business applications operated in the corporate data centre or hosted Dynatrace DCRUM, CA Wily CEM
Real User Monitoring (RUM) – Performance monitoring from an end user’s perspective, by passive monitoring of transactions carried out by “real” end users Browser-based Transaction monitoring and performance analysis by passive monitoring of browser-based user transactions Real user monitoring for web applications Dynatrace UEM, CA Wily CEM
Real User Monitoring (RUM) – Performance monitoring from an end user’s perspective, by passive monitoring of transactions carried out by “real” end users Client-based Transaction monitoring and performance analysis by passive monitoring of user transactions in the client application Real User Monitoring for web-based and mobile applications as well as business applications with proprietary rich clients Dynatrace UEM
APM deep dive technologies
Network-aware APMAPM deep dive focusing on the application performance in different delivery networks (LAN, WAN, Thin Client, Internet, mobile, …) IP flow analysis Transaction monitoring and performance analysis by passive monitoring of the network data traffic Fault-domain isolation network/data centre for business application operated in the corporate data centre or hosted Dynatrace DCRUM, CA Wily CEM
Agent-based APMAPM deep dive focusing on the application performance in the data centre Code instrumentation Transaction monitoring and performance analysis by using application server agents Deep dive analysis of the application performance in the data centre by multi-tier transaction monitoring Dynatrace Application Monitoring, CA Wily Introscope

Classification of APM Monitoring Techniques

The following overview explains the major APM monitoring techniques as well as their advantages/disadvantages.

Active Robots

With the so called robotic/synthetic testing technique, typical user transactions are recorded with a script recorder. Then the transactions will be repeated on a 24×7 base from robot workstations (capture relay method). The project success highly depends on the selection of the correct “most used” transactions, as complex but rarely used transactions will enormously increase the effort, but add little value to the application performance monitoring results.
The advantage of this most often used method with numerous vendor offers is the availability of continuous monitoring data from selected reference points. In addition, clearly defined, repeatable transactions are used that will provide perfect KPI data for SLM/TBM.

The disadvantage is the complex design and maintenance of the script. In addition, an impact analysis, i.e. a statement, how many users were impacted by performance alerts is not possible. Finally, the synthetic test causes additional workload for the IT infrastructure. At amasol – depending on the use case – we are using the following products as “active robots”:

  • On premise business applications: Dynatrace Synthetic Monitoring for Enterprise Apps and CA Unified Infrastructure Management
  • Cloud/Internet applications: Dynatrace Synthetic Monitoring Web and Mobile and CA Unified Infrastructure Management Cloud Monitor

Passive Client Agents

With this method the user experience is monitored by software agents at the browser, on the desktop PC or when thin clients are used at the terminal server. For browser-based monitoring the instrumentation is executed in most cases by Java script injection on the web and/or application server which does not require an intervention at the device and makes it possible to use the method for Internet applications as well. Application-specific rich clients in contrast most of the time require a specific application instrumentation.

The advantage of this method are very accurate monitoring data from a client’s perspective including the client’s processing times. The disadvantage especially in the German-language region is the “surveillance image” of this monitoring method making it more difficult to implement it in companies from this region.

Depending on the use case amasol is using the following products:

  • Dynatrace/Gomez UEM
  • CA Wily CEM

IP Flow Analysis

This method captures performance data by transaction through passive monitoring of the network traffic via probes/appliances at a central location, for example at the data center gateway.

The advantages of this method are the clear performance statements for every user transaction leading to a facilitated impact statement. Implementation is very easy, maintenance efforts are small. Neither agents are required nor changes to the IT infrastructure. For this reason, this method may be used as well for users that are outside the corporate IT sovereignty e.g. business partners or Internet users.
The accurate transaction monitoring via IP flow analysis requires a deep analysis of the corresponding communication protocol – many APM products provide this analysis only for http(s), nevertheless for using it successfully in a corporate environment a broader technology coverage is necessary, e.g. for SAP, databases or Citrix.

At amasol depending on the use case we are using the following products:

  • Dynatrace Real User Monitoring for Enterprise Apps
  • CA Wily CEM

Application Agents

For an APM deep dive analysis, i.e. the monitoring and performance analysis of single transactions through all application tiers in the data centre as well as an analysis of the methods/function calls within the single application server components generally application agents are used. For Java and .Net applications generic agents are available that instrument the bytecode when the application is started. For other application platforms often a code instrumentation will be necessary.

For agent-based APM deep dive we at amasol are using the following products depending on the use case:

  • Dynatrace Application Monitoring
  • CA Wily Introscope

Best Practice: Combination of Methods

As shown in this overview there is not a single “best” APM method. Therefore depending on the specific corporate requirements a useful mix of methods is recommended. When choosing a vendor you should select a vendor who masters all required methods to avoid the higher integration effort that otherwise will be necessary.


Contact

For further questions and information please contact

Stefan Deml
amasol AG
Campus Neue Balan
Claudius-Keller-Straße 3 B
81669 München
Tel.: +49 89 1894743-12
stefan.deml@amasol.de