application performance management (APM)

Many companiesare experiencing performance bottlenecks that can be resolved with Application Performance Management( APM). Rapidly increasing data volumes, business processes that are executed across multiple locations, heterogeneous networks and the growing number of Internet applications result in bottlenecks that lead to performance problems and impair the work of both employees and customers.

Application Performance Management (APM) is part of the tasks of a comprehensive system and network management of every company. Along with Fault Management( FM), Security Management( SM), Accounting Management( AM) and Configuration Management( CM), often referred to as Fault, Configuration, Account, Performance, and Security Management ( FCAPS), Application Performance Management continues to play an important role in ensuring acceptable quality of service for users. Among these, infrastructure components such as servers, storage, networking, operating systems, security systems, and monitoring are constantly changing.

APM management basically involves two levels of mapping:

1. Dependence of business processes on applications.
2. Dependence of applications on infrastructure components.

A clear and presentable hierarchy helps with all management tasks. Unfortunately, this type of visualization is rarely used due to legacy issues.

APM management is closely linked to service level agreements(SLAs), as most key performance indicators(KPIs) are included in the SLAs by APM management. Examples are data throughput, response times and quality of service( QoS). Customer satisfaction is also mentioned, or user perception of quality( QoE).

Many other KPIs are specified in industry standards and benchmarks, such as Information Technology Infrastructure Library( ITIL), OSI, ETSI and Control Objectives for Information and Related Technology (CoBIT). From a practical point of view, however, it is necessary to keep to a realistic minimum.

The factors influencing performance are always very important. They include, among others:

• The quality of the program structure, the language and the quality of the code,
• the power of the resources where the applications run,
• the fire width in the network and the nature of the competing applications,
• the degree of support for the mobility of the applications,
• the number of concurrent instances of the applications,
• the depth and thoroughness of testing and
• of load emulation for capacity planning.

The question of accountability for these factors is an open one: Information Technology( IT) or business units?

APM architectures consist of the following three components:

1. Acquisition of data and information.
2. Analysis of the data.
3. Provision of the results.

Performance measurements are always very important. The technology is available to measure with software or hardware agents in all infrastructure components. So-called Infrastructure Access Points( IAP) are also provided by manufacturers. In Cisco routers, for example, performance data is provided by NetFlow. There are many standards on what, where and how to measure, and where to store the measurement data. Well-known examples of this are SNMP in versions v1, v2 and v3 and RMON in versions 1 and 2, which are very often used. Other alternatives are also gaining importance, especially web agents, which dominate in the penetration of our economic and private life by web applications. XML-basedprotocols such as Simple Object Access Protocol( SOAP) and Netconf are replacing traditional protocols. They represent the new standards for several management tasks, including application performance management. However, a mapping procedure allows both protocol families to coexist for a long time.

In the meantime, the collected data is stored in Management Information Bases( MIB), then transferred to a more powerful database. ARM MIBs define the application programming interfaces( API) that allow agents to be embedded in applications. They then provide the basic information for APM management. This is followed by processing and analysis. This is where company interests play an important role. It can be simple evaluations over given time periods or correlative analyses between multiple KPIs. This step leads towards business intelligence and business analytics( BI/ BA).

The results are provided, printed and distributed in different ways. For example, as a periodic email attachment, as a download (pull) on a web page, or they are sent to the subscriber in a timely manner (push).

There are various alternatives for carrying out the measurements:

• Measurement on the fly with minimal overhead (many tools).
• Measurement with synthetic transactions to test the response behavior of the infrastructure (some tools).
• Measurement in an emulated environment (e.g.: test database) where the load is generated by a load driver based on co-recorded transactions (few tools).

The arrangement of sensors and agents is similar in all three cases. In most cases, end-to-end measurements are targeted.

The cost of APM management comes from tools, analytics software, reporting software, training, and system and network overhead. But on the other side of the equation is business efficiency and agility, supported by high-performance applications.

Virtual environments present a new challenge. By sharing virtual resources, the performance of a given application is not readily interpretable. Virtualization vendors strive to leverage the virtualization layer with the hypervisor between the applications and the infrastructure for management tasks. The hypervisor is the data source with a defined interface for downstream procedures. Agents can also be positioned within the virtual machines. Examples include: BlueStripe, Compuware, OpTier, DynaTrace and Quest Software.

Tool offerings come from the large companies such as IBM (Tivoli Suite), HP (Netview, SiteScope), BMC (ProactiveNet, Analytics, Performance Manager) and Computer Associates( eHealth, Introscope) or from smaller companies such as NetIQ (AppManager), Compuware (Vantage Suite) NetScout (nGenius, Sniffer), Paessler (PRTG) and Zott (Saturn Suite). Many options exist with open source that can be used either directly or indirectly through other products. Some examples are: OpenSmart, Ganglia, Nagios, NetMeter, Osmius, Nimsoft and Zenoss.

APM management is a continuous process. Improvements through tuning are always possible. Tuning can only be successful if the real causes of the performance problems are known. There are also so-called accelerators as AdWare - mostly hardware-based - which mechanically improve performance in some components of the infrastructure. However, solutions where resources are switched on as needed - e.g. cloud resources- so that application performance can be guaranteed even at peak load are promising.