• application performance management case studies

AppsOne® Architecture Diagram

The block diagram below demonstates how AppsOne® captures a comprehensive range of real user data (Data Capture and Monitoring layer), then analyzes that information in real time, leveraging the unique capabilities of the platform (Analytics layer), and finally helps companies assess what action or actions should be taken before problems occur (preventative vs. reactive) as well as automatically initiating many of these tasks and remediations (Operations layer including the AppsOne® automation engine).

AppsOne® includes a comprehensive library of extensions and APIs to connect to various inbound data sources and outbound operations services and applications (like ITSM service desks). AppsOne® also has integrated enterprise level security features to ensure the system and it’s related data is protected, while not introducing other vulnerabilties into an IT environment.

The sections below provide more detail about each of the key capabilities identified in the AppsOne® stack diagram.


Application Behavior Learning (ABL)

At the core of the AppsOne® solution is Appnomic®’s patent pending behavior learning and analytics technology – Application Behavior Learning. ABL is a powerful application of “Big Data” analytics applied to the enormous and growing volume of metrics in the complex application environments of enterprise and cloud IT infrastructures. The core of ABL is the analytics approach that applies a number of new cluster, regression, and machine learning analytical techniques to correlate three dimensions of metrics:

Different from other APM or analytics solutions, the resulting correlations are not just time based (like comparing last Friday to this Friday) or event driven which the ABL approach will also capture. ABL’s correlations are driven by application usage patterns - actual usage patterns of real users AppsOne® identifies and which reflect the volumes of concurrently occurring transactions that are contending for underlying infrastructure component resources.

Patterns are automatically discovered by AppsOne® for comparison to future behavior to assess if future behaviour is consistent with patterns reflected in desired or accceptable conditions. If a pattern identified on a Friday afternoon also appears at on a future Tuesday evening, AppsOne® will identify this application usage as statistically the same and do its work accordingly. If a news piece on a company in “off season” generates a high load reflective of a high season shopping day, AppsOne® will pick up this pattern – even though it is not high season – because of the actual work load approacing the application.
More details about how ABL identifies and matches these patterns are available through dicussions with the company.

Application Usage Pattern Horizontal Bar Chart

  • Real end-user application transaction experience
    (e.g., responsiveness, slowness, availability)
  • IT infrastructure key performance indicators
    (KPIs – like cpu utilization, database IOPs, active connections, etc.)
  • Naturally occurring load or Application Usage Patterns (AUPs)

Application Usage Patterns (AUP)

A core component of Appnomic®’s approach to the next generation of application performance manage-ment is the ability to view application transactions and data center metrics from the end-user “layer” of the application / data center stack down into the lower layers of the application stack. AppsOne®’s ABL identifies naturally occurring combinations of end-user transaction types and volumes of these transactions in short monitoring intervals. If an adjacent monitoring interval shows a similar, but slightly different mix AppsOne® will identify if that combination is a statistically different pattern or essentially the same.

Each of these patterns is then correlated to "good" behavior of the infrastructure components that support the application. In deployment of the softwaare, this occurs when the software is placed in learning mode.

When the application usage pattern is seen again, if performance metrics associated with the components are out of alignment (above or below baseline trigger levelss established by AppsOne®), then AppsOne® knows there is a problem and provides an alert. The platform also simultaneously begins to capture diagnostic and forensic information to provide IT operations professionals (see Forensics).

AUPs have proven insightful for developers as well who can sometimes get visibility into unanticipated combinations of concurrent transaction types which can lead to application redesign and optimization. They also are used for application load testing to generate a more typical set of real production environment combinations of concurrent transaction types generating load within the application and on the data center components.

Automated Threshold Discovery (ATD) and Dynamic Alerts

ABL automatically discovers and sets the trigger levels for alerting and eliminates the need for users or administrators to do this work. The alerting trigger settings result from the identification, by AppsOne® analytics, of Application Usage Patterns. AppsOne® analytics correlates those patterns to underlying infrastructure component behavior (e.g., servers, databases, hosts) -- their performance KPIs. When load is high, alerting trigger levels are set high – for example when a batch job is running in parallel with other production end-user activity. When load is low, alerting trigger points are low. Unlike typical monitoring systems in the market today, these alerts are dynamic as the trigger levels change throughout the day based on what application usage pattern is active at any given time.


Full Stack Visibility

Among the more popular screen views of AppsOne® is the “full stack view” screen where all key layers of the application stack are visually presented in one “pane of glass” and the relationships between the different layers can be viewed in a unique manner. This enables IT operations professionals to more quickly identify relationships between the data center components, the application layer, and end-user Application Usage Patterns. As a result, they can better prevent issues that may affect the application layer where users are living the app experience. This visualization is also helpful in the fixing phase of dealing with problems that may still occur.

The X-axis in this view is time so the visualization scrolls from right to left as time passes. In the illustration below, the top horizontal bar represents different Application Usage Patterns that are generating load on the platform. As AUPs change, the impact on the data cener layers below can be visibly observed. Alerts from one layer can be seen relative to alerts in other layers of the application stack.

Automation Module Actionable Analytics

AppsOne® includes an embedded automation module that is triggered by Early Warning Alerts™. This automation module enables the safe, secure and auditable execution of preventive, remedial, restorative and performance enhancing actions, tasks and scripts for infrastructure services, application components and related infrastructures. These actions can be conditionally configured to be executed whenever a specific Trigger Range deviation is observed. Actions can be as simple as stopping and restarting an application or as complex as ‘flexing’ your application by adding or removing capacity to enable an application to ‘self-modulate’ its performance. Existing and In-house developed scripts can be easily incorporated into AppsOne® for use by the automation module as well.

To build new preventive or remedial automations or enhance existing scripts, Appnomic®’s OpsOne® IT Process Automation (ITPA) solution can be used in conjuction with AppsOne® to develop automations ranging from the very simple to very complex, leveraging OpsOne®’s easy to use, efficient graphical process designer (the Enterprise Solution Designer or ESD). The process designer helps design interactive and rule-based workflows with the desired mix of automation, human interaction and manual checkpoints. It can be easily used by IT Operations staff and does not require developers. The automation module can also be used to completely automate the diagnosis and troubleshooting actions of L1 and L2 resources and directly pass the information for root cause analysis to L3 resources.

Early Warning Alerts™

AppsOne® leverages ATD to deliver accurate early warnings of impending performance issues when either application or infrastructure KPI’s deviate from normal behavior. Deviations often occur in the depths of the infrastructure before they impact the end-user experience. AppsOne®’s analytics and correlation capabilities are uniquely designed to provide these accurate, detailed warning alerts needed to prevent user experience and service quality from being impacted. Early Warning Alerts™ are generated when KPIs deviate – up or down from the expected range of performance. Many APM and ‘analytics-based’ systems today alert when a threshold is exceeded, but not when a KPI is below an expected level. These types of alerts are particularly helpful in catching upstream hotspots where a problem upstream is preventing activity from proceeding -- resulting in KPIs below expected levels for downstream systems.

Automated Fault Remediation and Self-Modulating Applications

Once Appnomic® has identified an issue, there are cases where clients request the remediation or restoration of services be automatically implemented. There is an automation capability in AppsOne® that enables these remedial actions. In addition, AppsOne® may message to an external automation system like the Appnomic® OpsOne® solution to implement such automations. This function of AppsOne® enables the analytics to be as actionable as possible – not just informative to the IT operation.

The most advanced users of Appnomic® technologies are now achieving self-modulating application performance. One bank is providing a premium online experience for premier banking clients by using AppsOne® Real User Monitoring (RUM) and OpsOne® automation capabilities. Envision ramping up an ecommerce stack of components to deliver extra fast response times and customer purchase conversion for respondents to an online marketing campaign. Conversely, when demand is low, users of Appnomic® technologies can configure applications to turn down resources and conserve OpEx budget or capital utilization.


To aid in the diagnosis and preventive response to emerging performance issues, Early Warning Alerts™ can automatically trigger the ‘in-time’ collection of context-specific, diagnostic KPIs and data across related infrastructure components. This capability to collect transaction, application component or infrastructure data from a wide variety of sources, attach this data to the associated Early Warning Alerts™ and provide it easily for use through AppsOne®’s graphical user interface is a key feature that helps accelerate the preventive actions involved in an AppsOne® alerting situation and also in remedial actions should an incident occur.

Data Capture and Monitoring Layer

Real User Transaction Response Time
One of the core differences between Appnomic® and many others is that we believe the approach to predictive analytics for application performance management should start with the end user experience and then move from there. Trying to solve cross-application problems coming at the problem from the infrastructure components like network or servers cannot enable a wholistic and accurate view of impacts to the ultimate goal of the application which is to enable a user to achieve a specific task.

Technology advancements have also now enabled ITSM systems to operate quickly enough to process real user activity versus taking proxy measures or by implementing synthetic monitoring to assess an application’s performance. These advances in technology producivitiy, efficiency, and cost reductions enable executives and IT operations professionals to now monitor real user experience and “health of the business” with the AppsOne® Transaction Performance Screen that provides a quick visualization of end-user transactions that meet target response time requirements, those that are slow, and those that are failing.

AppsOne® Real User Monitoring may be used for HTTP, HTTPs and non-HTTP transaction types as is often necessary for hybrid environments. There is out of the box support for client server based applciations like SAP, Finacle, Fino, and ATM transactions. Data collected can include detailed analysis, with an in-depth view of response times up through end-user desktops, network latency and processing time in data-center.

Infrastructure Component KPIs

The Appnomic® philosophy and approach to monitoring is to start with the least invasive and lowest overhead methods that achieve the best results.

Collecting application transaction data is accomplished through the installation of our Protocol Stack Agent (the PS Agent) at the entry points to the application, typically the front end web server where users or applications interact with the target environment. The PS Agent monitors the traffic as it passes into the network port of the application server, in much the same way a network sniffer does, and captures administrator defined traffic for analysis by the AppsOne® server. This data is then packaged up by the agent and forwarded on to the AppsOne® server. To collect infrastructure component KPIs, AppsOne® uses a variety of methods depending on the KPI. The collection methods are low impact and use existing application or system tools wherever possible. For example, to monitor Oracle databases, a standard login is used to collect metrics. For UNIX and LINUX, we login with an SSH session and collect KPIs such as CPU and memory utilization with existing system tools. Agents are only employed on Windows based platforms where these functions require special access to obtain the data.

AppsOne® collects data and monitors a wide array of business and application transactions, as well as all layers of application infrastructures- web servers, application servers, database, as well as storage. This comprehensive - End-User to Storage – data collection and monitoring provides the right data (transactions, application components, infrastructure) to provide accurate, preventive Early Warning Alerts™.

As Appnomic®’s differentiating and competitive advantage relates to analytics and the unique implementation of analytics – not data capture -- with monitoring and automation, we work to be flexible on data capture methods – leveraging monitoring infrastructure in place or deploying collection methods where appropriate.