Virtualization Technology News and Information
Think Like an End-User, Monitor Like an Admin

Application performance monitoring in the age of cloud

By Patrick Hubbard, Head Geek, SolarWinds

Admit it. Stepping into a NOC of beautiful dashboards with hundreds or thousands of live operations metrics feels...good. For engineers and executives too, familiar equipment reporting low utilization and green status reassures everyone that all is right with the world. The problem? Recent changes in how we deliver applications are making traditional monitoring techniques difficult or even impossible, while the business expects even greater performance. Worse, the same decisionmakers pushing us to cloud and encouraging us to mix on and off-premises as hybrid operations, don't want to let go of the traditional metrics they trust. But after years of false-starts, operations has a chance to rethink-critically-the practical definitions of "good" when it comes to making users happy. It's a rare chance to catch up with recent technology changes and regain control of application delivery monitoring.

At an APM Crossroads

If you're using cloud the way you're supposed to-or at least the way more likely to meet your budget-you're learning a great deal in a hurry. You might not be capturing mythical, digital transformation unicorns, but at a minimum you're modernizing more applications than expected. However, it's also possible you're actually deconstructing legacy applications to fully utilize public cloud services and cloud-native technologies. Either way, many businesses are discovering traditional monitoring isn't enough because the foundation of how we monitor our applications has substantially changed.

We're at a critical application performance monitoring (APM) inflection point, and there are two realities we must accept. First, the advent of cloud computing is driving a long overdue rethink of traditional strategies and metrics because user needs for our applications have changed. And second, the "application stack" we can see in our sleep, increasingly can't be monitored by design. Public cloud providers are the world's largest MSPs, and like any MSP won't and can't give you root access to their systems. That's not how multi-tenancy security works-full stop. But while that might be unnerving, this disruption is intentional, forcing organizations to evolve their strategies if they're to get APM back under control.

On one hand, cloud's pitch to hardware-weary IT is compelling-freedom from the day-to-day management of infrastructure. Most application owners are happy to stop fiddling with base metrics like network traffic, IO, storage, CPU, memory, and more. But we've historically done that because they've always been the tea leaves we read to infer application performance at the user end. As long as well-understood metrics stayed within specific boundaries, users didn't seem to complain. We were using the tools we had in the shed, and it mostly worked.

But we should have evolved 10 years ago, because the fat-client, LAN connected web pages that drove stack monitoring then are no longer typical of user needs today. The green glow of device dashboards now has less bearing on whether you're actually delivering high-quality user experiences. At the same time, there are fewer out-of-the-box tools to deduce whether cloud applications are performing well-especially when it comes to custom applications or hybrid operations.

This is where the core tenets of modern APM will have a profound impact moving forward; they empower businesses to place service delivery goals over resource utilization thresholds. Better, it lets them define success metrics based on evolving real-world, end-user experiences. Incorporating Service Level Indicators (SLI's) like latency and error rates, along with custom data like user satisfaction and user behavior, can greatly help organizations identify performance indicators driving improvement and growth.

For many engineers, this process starts with a simple change in perspective-learning to think like an end-user to monitor like a sage admin.

Plan for the Future by Looking Backwards

In this approach, consider your application users as endpoint monitors or system agents, and instrument them accordingly. Tools are finally available to turn code injection and custom application instrumentation into reasonable configuration tasks, allowing business experts to outline strategic delivery goals. Then, from those real user metrics you can follow the troubleshooting path backwards to include associated infrastructure performance, making inevitable issue resolutions easier.

To be clear, traditional infrastructure metrics still matter greatly. Even in the age of cloud things will break, requiring understanding and the reconfiguration of code, app platform components, databases and more. If end-user metrics indicate slow delivery and correlated infrastructure data reveals high database wait times, you'll still need a DBA or an ops engineer with query chops to identify root causes. But for day-to-day and planning purposes, back-end details aren't particularly helpful when answering key questions like, "Is our app good? Do people like it? Are our systems generally working as expected and promised?"

This evolved approach is particularly important when it comes to custom applications-especially those we know are business differentiators. Even if these apps are built out of standard vendor parts, the way they've been assembled is unique to your business. It makes sense then that measurement and monitoring should include real-user behavior data specific to these apps. Are you tracking users' average purchase prices or instances of checkout ‘add-ons'? Are you more concerned with how your app encourages dwell time on a certain page or general levels of interactivity? Are mobile users equally satisfied on different wireless carriers or Wi-Fi networks? Modern APM can deliver 80 - 90% of this information while adding new insight the business will truly appreciate.

It Pays to Start Small

So, where to start? Time and money are the canonical answers, and at least some budget and time will be needed. You're not completely remodeling ops to achieve the same monitoring richness you had with traditional infrastructure monitoring. But it's no longer necessary to bring in an army of developers or consultants. Simply adding end-user monitoring and digital experience monitoring as you modernize apps is a low-friction first step. APM is a great shortcut to restore much of the usefulness provided by the stack monitoring you used to rely on.

As with most new things, it's a good idea to start small and remain flexible. If you're new to APM, you won't make any real progress if your first attempt is going to cost a million dollars or requires a specialist or even more staff. The best APM monitoring is the APM monitoring you have, not pie in the sky monitoring you can't afford. Try layering APM into the monitoring of a critical system or an application where performance is a known issue and prove to your organization that it's a powerful tool ultimately resulting in better application delivery.

This approach also lends itself well to renegotiation with leadership, offering a relatively risk-free educational opportunity that can in turn help reeducate and reset expectations going forward.

What Now?

Perhaps the silver lining to updating your APM strategy is it can buy you freedom to innovate. We've trained IT management to use specific infrastructure metrics and dashboards as key performance indicators. And you'll never be able to change those views, reports, and alerts because it creates unease. Managers like it when green things remain green. But if they learn how to measure application delivery quality as the primary indicator of success, a bit of a wobble on the back end while you evolve it feels less risky. For example, if you provide a metric of real users' shopping cart error rates, it's easier for the back end to make changes as long as error rates are reasonably steady.

This pivot in monitoring strategy ultimately serves a higher purpose-it's a foot in the door to start more meaningful conversations with the CIO's office. Three to five years ago, most teams were too busy wrapping their heads around cloud-driven changes to socialize ideas, and worse, APM was expensive. But now's a great time to begin a conversation about catching up with the needs of changing usage and new platforms. Leadership wants us to succeed and (recently), is more open to learning new ways to make customers happy. After all, they don't want to mess with racks of servers any more than we do. We both want to deliver apps customers love, while keeping our fingers safe from hungry rack rails.



Patrick Hubbard 

An accomplished technologist with over 20 years of experience, Hubbard's career includes software development, operations, product management and marketing, technology strategy, and advocacy. An unapologetic market-hype deconstructionist, Hubbard is passionate about arming technology professionals with the tools and skills to deliver services that delight, not just satisfy, users. Hubbard's current focus is helping enterprises adopt cloud-native and DevOps techniques that deliver the business transformation CIOs increasingly demand.

Since joining SolarWinds in 2007, Hubbard has combined his technical expertise with an IT customer perspective to drive product strategy, launch the Head Geeks, develop and manage the SolarWinds Certified Professional (SCP) and SolarWinds Academy Training Classes programs, and create the SolarWinds online demo platform. Today, most admins recognize Hubbard as the executive producer of the Telly award-winning SolarWinds Lab, and SolarWinds THWACKcamp.
Published Thursday, October 03, 2019 7:09 AM by David Marshall
Filed under: ,
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
<October 2019>