Welcome!

Linux Authors: Liz McMillan, Pat Romanski, Trevor Parsons, Ivan Antsipau, Michael Shaulov

Related Topics: DevOps Journal, Java, .NET, Linux, Cloud Expo, Big Data Journal

DevOps Journal: Article

Choosing the APM System that Is Right for You

A lot of the arguing in the APM space is about the fundamental approach to monitoring application transactions

In my role as technology evangelist I spend a lot of time helping organizations, big and small, make their IT systems better, faster and more resilient to faults in order to support their business operations and objectives. I always find it frustrating to "argue" with our competitors about what the best solution is. I honestly think that many APM tools on the market do a good job - each with advantages and disadvantages in certain use cases. There is no "one size fits all" - there is just a "this tool fits best for your APM Maturity Level" (not saying the others wouldn't do a good job).

A lot of the arguing in the APM space is about the fundamental approach to monitoring application transactions: monitor and capture ALL details vs. monitor and capture relevant details. Along with that come topics like "overhead impact", "scalability" and "data hording vs smart analytics".

Ultimately, you want to pick the right tool to solve your problems. As you have multiple tools to choose from let me - in my role as technology evangelist - highlight some of the use cases that our customers solve. As a technologist and a blogger, what I really care about is that the right technology is applied to the right problem. As such, I feel compelled to share what I have learned working with customers in the trenches. Hopefully, this will help you understand the technology and what problem it can solve in real life problems, and cut through the propaganda. Let me start with a few use cases today and follow up with some more in follow up blog posts.

Use Cases from Steven - A Performance Engineer
The first use cases are picked from Steven - whom I reached out to after I read his question on our APM Community Forum. His company decided to move from a competitor to our APM solution and I wondered why. In an email, he highlighted that he had some initial success with the tool, and had been able to solve a couple of low hanging problems. When they decided to start taking a strategic Continuous Delivery approach to software delivery, they realized that the current tool had certain shortcomings slowing their attempts to practice DevOps.

They identified the following key problems they need to solve and what they really required from an APM solution in order to get to where they are heading:

How a user got to a problem, and not just seeing the problem itself

  • Every transaction, with all details they need, out-of-the-box
  • Web request/response bytes, SQL bind values, exception details for every transaction

Number of transactions executed per user and tenant used for business and cost reporting

  • Capture custom business context data for every transaction
  • Business transactions based on "buried" context data as not every detail is in the URL

Eliminate homegrown tools which are costly to maintain

  • Provide application as well as system and infrastructure monitoring
  • Integrate with other tools such as JMeter, LoadRunner, Jenkins or HP Open View

Eliminate the need to make people look at other tools and data

  • Foster collaboration across Architect, Dev, Test & Ops by using same data set
  • Data must be shareable with a single click

Ability to extend to custom frameworks, systems and protocols

  • Bring in custom metrics from external tools via Java Plugin infrastructure
  • Follow transactions across any custom protocol or technologies outside Java & .NET

Full Automation to support Continuous Delivery

  • Use Metrics provided by APM for every build artifact along the deployment pipeline to act as quality gateway
  • Inform APM about new deployments to prevent false alerting

Replace traditional application logging

  • Eliminated log files which saves I/O and storage
  • Get the log messages captured in context of a transaction and the context of the user that triggered that log message

One solution for everything

  • Not just performance monitoring but also business reporting as well as deep dive diagnostics

Active community forum

  • Get answers right away
  • Leverage extensions already provided by the community such as plugins for Jenkins, PagerDuty, ...

Let me give you some examples for Steven's use case so that you can better decide on whether that is relevant for you as well:

Every Transaction with All Details
dynaTrace was built from the ground up to support the full software lifecycle. We as Compuware APM/dynaTrace understood that we needed a technology that captures every transaction with all details for root cause diagnostics as well as proper business monitoring without falling into a sampling mode where you lose critical information for both business and root cause diagnostics. Most of our customers claim they see little to acceptable overhead in production yet capturing 100% transactions including method arguments, SQL Statements, Log Messages or Exceptions. The magic word in our case is our PurePath (see the YouTube video) & PureStack Technology which allows dynaTrace to do exactly that. One of the several visualization of the PurePath is the Transaction Flow which is a great way to understand how your transactions flow through the system - where your hotspots are (3rd party impact, custom code issues or impact of Garbage Collection) and where your architectural issues (e.g: too many web service calls, too many SQL executions):

Transaction Flow: One View that tells it all to Devs, Architects and Operations Teams

What if you don't capture all transactions but be "smart" and focus on capturing the problematic ones? While this approach allows you to find and fix the easy-to-find problems that can be analyzed by analyzing those transactions that fail or violate the average response-time based baseline, it falls short when it comes to problems that are caused by transactions that are not "outside the norm". One example here is a database deadlock we recently analyzed for a customer. The "smart" approach only highlighted the transaction that hit the deadlock but no information was captured for those transactions actually causing the deadlock with their data manipulations. Being able to see which transactions executed which UPDATE statements at the time leading up to the deadlock is required to solve this problem.

As companies - such as Steven's - are getting into a maturity level where they grow out of "smart" average response time-based analysis it is important to have the ability to look at everything and not just the average problem. As a follow up read the blog Why Averages Suck and Percentiles are great!

Capture Custom Business Context
What is Custom Business Context? The actual business function executed such as a "Create Claim", "Transfer Money," or the name of the user or tenant of your system. Why is this not as easy as it sounds? Because many applications just don't show the business function as part of the URL or provide the user name in a cookie. A great example was given in a webinar by NJM Insurance (New Jersey Manufacturing Insurance). They were using a third-party claim management software which was designed to "hide" everything behind a claimCenter.do URL. In their case they needed dynaTrace to analyze every single transaction and pick a method argument invoked in the business layer of their app to figure out which function in their system was actually executed. On top of that they also needed to know the user that executed that function because they needed to understand which insurance office and group of employees created how many claims as they needed this for their quarterly business reports. The following shows business reporting based on the user role where the user role gets captured from a method argument within the business logic of the application:

Business Reporting requires Business Context data for every Transaction

This was only possible because dynaTrace allows you to selectively capture business context in the context of every single executed transaction. Along the PurePath you will then see things like method arguments, return values, bind values, session variables, HTTP parameters or cookie values. All to be later used for your business reporting or targeted root cause diagnostics. Here is a follow up blog post that explains business transactions in more technical detail.

For more APM Buyer's tips, and for further insight, click here for the full article.

More Stories By Andreas Grabner

Andreas Grabner has more than a decade of experience as an architect and developer in the Java and .NET space. In his current role, Andi works as a Technology Strategist for Compuware and leads the Compuware APM Center of Excellence team. In his role he influences the Compuware APM product strategy and works closely with customers in implementing performance management solutions across the entire application lifecycle. He is a frequent speaker at technology conferences on performance and architecture-related topics, and regularly authors articles offering business and technology advice for Compuware’s About:Performance blog.

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.