This is an epic list of mistakes. These are the top reasons why APM deployments fail, according to many of the industry's top experts – including consultants, analysts, vendors and users – and also some of the solutions that help you overcome these challenges.
The list is not meant to be in order of importance. There is quite a lot of variety in the list, and any one of these pitfalls can cause your APM deployment to fail, if you are not careful:
1. Organizational Silos
The biggest challenge to APM is organizational readiness. APM, like BSM, is a cross-domain endeavor that requires multiple stakeholders to collaborate and share data. While there are a lot of management technologies that are evolving to support this, habits of mind, along with siloed processes and politics, often get in the way. There is no simple solution to this - but strong executive leadership or at least sponsorship is key - with the expectation of increased dialog to complement more automated, more efficient ways of working.
VP at Enterprise Management Associates (EMA)
The real obstacle to application optimization isn't technological, it is organizational. Application performance depends on two fundamental elements: application design and network configuration. Unfortunately, in too many organizations these two elements are responsibility of two separate and sometimes competing groups - the application development and network operations teams. Fortunately, the nature of today's Cloud-based apps dictates that the traditional network and application silos disappear. Network and application teams should use today's Cloud-based APM solutions as a mechanism to put aside their differences and start working together more effectively.
Managing Director of THINKstrategies and Founder of the Cloud Computing Showplace
Surprisingly, one of the biggest roadblocks to successful APM Deployments is not technology. It is ownership. Who owns APM? At many enterprises it is a cross-department effort. While there are specific groups who own web servers, others who are in charge of databases, networking owners and even business process owners, APM is often an orphan - albeit very cute, and everyone talks about adopting it. As a result, the implementation may be comprised and focused on its adopted parent’s siloed perspective. When that happens we can end up with volumes of very detailed information, but still lack clear visibility into the application. Make sure that the cross-department effort includes the application owners and that the technology deployed is focused on monitoring the results of the application and not just its activities.
VP Product Management and Marketing, Nastel Technologies
APM deployments can face challenges when IT staff is too siloed to see the full impact. Instead of having the operations staff (including system administrators, the infrastructure team, network engineers, etc.) each having their own tool to solve performance issues, an APM solution should offer access to data that is relevant to the operations staff, while being easy to use and time synchronized between the network, the application and the client to isolate problems as quickly as possible for all the operations staff to capitalize and work from, not just a subject matter expert.
Senior Product Manager at Visual Network Systems
2. Not Focusing on Business Impact of APM
A research report published by Quocirca earlier in 2012 shows that businesses are too often failing to correlate the metrics gathered by APM tools with relevant business metrics. For example if a customer facing web application is performing poorly, how does that impact sales?
Analyst and Director, Quocirca
APM solutions can collect a lot of data, the problem is turning this data into operational intelligence so IT can be more productive, thus helping IT and the business to become more agile. Focusing on low-level APM technical metrics and data is interesting but they don't provide any indicators of end user experience or business impact. If APM is to create value, ROI and success then its users have to focus on managing business impact, so they can prioritize what to troubleshoot rather than being slave to infrastructure resource, health or capacity metrics.
Tech Evangelist, AppDynamics
Too often, application performance management is focused on monitoring and correlating information about web-application components or server performance. This does not consider the full implications of what it takes to deliver a service to the user. Nor does it identify all issues that can impact the performance of an application or service. The bigger question IT needs to focus on is “What am I trying to achieve to support the business?” IT must consider service delivery as a strategy and holistic responsibility. APM technologies should be evaluated based on the desired end goal of assuring service delivery, and must consider all elements that impact a user’s experience.
Vice President of Marketing, NetScout
3. APM-Generated Big Data
The huge amount of data collected and available from APM monitoring tools can be overwhelming, and translating into actionable information can be challenging. Setting up KPIs on a large number of performance metrics can be impractical or require too much time and resources. Consider using an analytics based solution to automatically learn the environment’s behavior and detect deviations from normal behavior before performance impacts systems and users, with major business impact.
Business Development, NEC Invariant Analyzer
APM tools are designed with the innate assumption that IT operation professionals understand how multiple tiers (vertical and horizontal) within an application stack interact during peak, non-peak and other load conditions. In reality, this assumption is incorrect. Enterprise IT Operations are supporting multiple applications built on disparate technology stacks and so it’s impossible to assume that the enormous amount of data that APM tools provide on the code, infrastructure performance metrics, and other factors for multiple application technology stacks can be interpreted and made truly actionable by IT operations in a timely manner. There is a need for APM products that start to understand the behavior of an application's entire stack. An APM product that can build load-behavior models, enhance them as more data or new scenarios emerge and use those models to monitor and manage the performance of business critical applications. This approach will enable early detection and prevention of application performance issues, self-healing, dynamic capacity management and the ability to leverage the true potential of the Cloud.
President, US Operations, Appnomic Systems
4. Dashboard Complexity
The traditional problem in APM is that the tools are designed for performance experts, and a lot of companies just do not have that luxury. The tools need to be accessible and understandable to the regular network and application teams.
Senior Research Analyst, IT Infrastructure, Aberdeen Group
The second generation of APM comes with a myriad of metrics, gauges, and dashboards, which can be overwhelming at first glance. Steer clear of flashy dashboards with dozens of metrics and animated icons that have little true meaning. Dial it down, and fine tune the message to how the application is performing from a business perspective. Present a dashboard that answers critical performance questions in one click or less. Focus on showing basic, easy to understand real-time monitoring metrics.
Director of Enterprise Application Services at the Auto Club Group
5. Rapid Rollouts
A challenge faced by many organizations is the need for rapid rollout of new and enhanced application-supported business processes. In order to increase speed to deployment, organizations are tempted to sacrifice instrumentation and simulation stages during production, but this can expose the team to rollbacks due to slower than expected application performance. To successfully meet the demands of the business, closer collaboration between development and operations is required to define, build and test applications that already include the monitoring instrumentation to handle production requirements. This results in a higher success rate of production rollouts and greater customer satisfaction.
Vice President, Tivoli Service Availability & Performance Management, IBM
6. Complicated Deployment and Maintenance
Rolling out an APM deployment can require a large amount of time and people power. IT organizations are already running around with their hair on fire and adding this to the list of projects they need to tackle is overwhelming.
The thing I hear most from clients is not that the deployments fail, but they often create deployment and management overhead which is unsupportable given the ever expanding scope of management needs.
Research Director in Gartner's IT Operations Research Group
An APM solution should be painless to implement and frictionless to maintain. The larger the startup and maintenance costs, the more likely an APM deployment is to fail.
Developer Evangelist, New Relic
Application performance management can be complex; but that doesn't mean deployment of an APM solution needs to be. If an APM tool requires manual work ("professional services") to get up and running and has the potential to lose value unless it's frequently maintained, it's going to be hard to use effectively. Fortunately, the pains of installation and maintenance have been greatly eased by SaaS and modern auto-instrumentation approaches. Pick an APM solution with zero time to value and dead-simple setup - save your time for improving the app.
APM Product Manager, AppNeta
I’ve seen failed deployments as the result of a project taking too much time to deploy. IT teams get distracted easily and a new fire drill appears. If the product can’t solve their immediate problem in a day or two, they move onto the next problem.
Marketing Manager, SolarWinds
One of the biggest reasons for APM failure is when the deployment doesn’t deliver the ROI - not just the software cost but the whole effort of deploying the APM environ and maintaining it.
VP of Products, Precise
7. Project Scope
Failing to address size and scope of the project is one of the reasons why APM deployments fail. IT may think it has the project properly sized, but could find their choice of APM solution, once implemented, cannot handle the large number of transactions being monitored, causing production-wide slowdowns. Or they might find there are blind spots in the coverage that limits the ability to troubleshoot problems. For large organizations in particular, an APM solution must be able to handle heavy transaction loads while covering the breadth of technology ranging from modern Web and cloud-based systems all the way through to legacy mainframe systems to help find the root cause of problems quickly.
Senior Director of Product Marketing, CA Technologies
Scoping deployments which are of a reasonable size and scaling up over time is a better approach. Trying to instrument 100 applications across 1500 servers within a deployment scope is probably not only too large, but with most organizations application diversity would take multiple years. This creates a high cost for software and maintenance. Taking a step back and understanding which dimensions must be met, and the most cost effective way to accomplish that visibility, is a way to speed up deployment times.
Research Director in Gartner's IT Operations Research Group
8. Today's Dynamic IT Environment and Advanced Applications
APM Deployments fail when they don’t take into account the application architecture vision and cloud strategy of the company. Most APM tools can cover one type of infrastructure well, such as J2EE with Oracle, but today, many companies are executing strategies around Cloud and Polyglot Programming (the use of multiple languages, databases and frameworks in an application). The failure to assess an APM tool's flexibility to adapt and work with today's modern, dynamic infrastructures, without compromising on the value it delivers, is a recipe for failure.
Director of Products, Boundary
The biggest challenge to APM deployments is that most of the tools and technologies were built for static environments. As a result, they are no longer tenable in today's enterprise IT environments, which are being transformed by IT mega-trends such as server virtualization, cloud computing, mobility, and agile development. To work in these environments, legacy APM tools require teams of consultants to configure and deploy, and they break when the environment changes. Modern APM solutions are built from the ground up to support dynamic environments. They automatically discover both physical and virtual devices and applications, detect changes in the environment, and are minimally invasive without perturbing the systems that they monitor.
CEO and Co-Founder, ExtraHop
The biggest challenge faced by APM deployments today is that many existing monitoring tools are unable to keep pace with today's advanced applications. While application performance monitoring tools have been available for many years, these older tools are focused on monitoring for problems at a course-grain level, providing e-commerce operators with no pre-failure warning and no valuable information capture to help isolate code-level root-cause after a failure has occurred. For example, if a user experiences a product search failure, it is important to quickly pinpoint and fix the cause. Using this level of instrumentation today applied to modern e-commerce applications with their rich, Web 2.0 functionality, multiple tiers, millions of users and dozens of web services is like using WWII fighter instrumentation for a modern B787 airliner. Just knowing a problem exists doesn't help you. You need to know where the problem lies. The solution to this challenge is something we refer to as “answer-focused APM”.
VP of Product Management for Compuware's APM Business
9. APM in Isolation
Application Performance Management (APM) in isolation is a recipe for failure. APM as part of an organization’s end-to-end monitoring is critical to the success of the APM project. A simple example illustrates the power of an integrated end to end monitoring capability. E-commerce in many organizations is a business critical service that both the business and IT care about. What do they both care about? Performance. While the business measures the business performance, IT also cares about the application performance. If the e-commerce service goes down or is too slow, it directly impairs the business performance. But how do you determine if it is an application, middleware, database, server, networking or storage issue? If you’re not monitoring the pieces as an integrated whole then you are not going to be able to pinpoint the root cause versus the noise in the system. Avoid isolation both on a personal basis and the monitoring front!
Product Marketing Lead for HP Application Performance Management
Trying to deploy APM tools in isolation can result in failure. APM is one component of your overall monitoring, and you need end-to-end infrastructure and application performance monitoring to get the full picture. Getting a "slow database transaction" alert from your APM tool is ineffective without having a monitoring platform in place for your database performance metrics, the server performance and the network infrastructure.
10. Setting Expectations Too High
Enterprise deployments of application performance management suites struggle when expectations are set too high, timelines are too ambitious, or when buy-in across Operations and Application stakeholders is lacking. If stakeholders lose patience, then the deployment may end up being scaled back before achieving significant results for IT Operations, Application Owners, and Business interests. Solutions should not be over-engineered to generate exaggerated expectations as it can lead to unsatisfied businesses.
Vice President, Tivoli Service Availability & Performance Management, IBM
11. Management Gaps
An APM deployment can meets its goals (solve production code problems) and ultimately be a failure. The focus of APM tools on specific platforms (like Java) creates too many management gaps that prevent IT Ops teams from efficiently managing application/transaction availability. The solution is to implement transaction-centric performance management solutions that go beyond code to see how all the pieces of an end-to-end transaction impact performance.
COO and Co-Founder, BlueStripe
12. Inaccurate Service Maps
APM deployments are being driven by the business side of the enterprise which is acutely interested in service availability and performance. With this business goal in mind, a strong APM deployment is complemented by maintenance of an accurate representation of the relationship between key CIs in the CMDB. A big challenge thus faced by APM projects is inaccurate service maps to which IT administrators turn when availability and performance warnings arrive. These warnings need to be quickly and efficiently turned into actionable information on the right CIs to reduce MTT-Identify problems and to repair (MTTR) them. In addition, customers environments are constantly changing and evolving so real-time service maps upon which change activities can be modeled and controlled is imperative. Customers can choose to “boil the ocean” in an asset management approach to populating the CMDB but experience demonstrates that this is often a futile, quixotic effort. It requires significant manual intervention which is never ending, as we are all familiar with the nature of today's IT environment: complex and always-changing. An alternative is to use real-time service modeling tools to address this lack of actionable, real-time service maps.
VP of Products and Co-Founder, Neebula
13. Too Much Overhead
Not only should an APM solution be low overhead operationally, it should be low overhead on the application itself. Be wary of APM adding latency and load to your application. In particular, APM data reporting mechanisms that have the potential to block critical web processes or hog resources on production nodes are highly dangerous. Of course, there's no need to be penny-wise and pound-foolish: deployment that adds some minimal overhead is ok - you'll be more than making up for it with increased efficiency due to visibility.
APM Product Manager, AppNeta
14. Missing the Big Picture
One of the biggest challenges faced by APM deployments is ensuring the right level of visibility. Depending on the tools that are used, IT may not be able to see the forest for the trees, perhaps getting views of application performance in pieces or just tied to specific infrastructure components. While these views are very valuable for troubleshooting, it's important to be able to build a service or application view that shows overall health, risk and availability, and where only service-impacting components are included in the “big picture”. Whatever APM tools you select, check to be sure that they allow you to operate at the service or application level - e.g., monitoring against customized SLAs, creating application-level eventing and alerts, and even kicking off automated remediation for issues that impact the application.
CTO at ScienceLogic
APM deployments most often fail when they are treated as yet another silo that peers only into the application layer. For an APM deployment to be successful, the APM solution must provide enough detailed data across all silos in the organization, beginning with the end-user experience; the application processing layer; the middleware; database; network; physical and virtual infrastructure; and the back end storage, too. All of this must be accessible via a single view into the data.
Director – Product Marketing for Quest Software (now part of Dell)
15. Choosing the Wrong APM Product
Picking the wrong product. There are literally dozens of APM vendors in the market and if you read their marketing materials, they all pretty much sound the same and claim to offer the best solution for anyone and everyone. Cut through the hype and talk to real users of these products to understand each product’s true strengths and most suitable use cases. Find these real users not only from the vendor’s reference program, but also from social networking sites such as twitter, LinkedIn, and IT Central Station.
Founder and CEO, IT Central Station