blobbotronic -

News Stay informed about the latest enterprise technology news and product updates.

Dynatrace monitoring faces market ambivalence in NoOps push

Dynatrace is gung-ho about NoOps with training services and an open source tool for CI/CD automation, but may face skepticism from a wary enterprise IT audience.

Dynatrace plans to expand the influence of its DevOps monitoring tools beyond production infrastructure environments and give IT pros a path to NoOps, but enterprise IT pros remain ambivalent about that concept.

NoOps IT management relies totally on unsupervised machine-based automation, often AI-driven, to manage application deployments and infrastructure, eliminating the need for human intervention. NoOps has waxed and waned over the years as a hot IT topic while DevOps and IT infrastructure automation have advanced. Each new wave of automation and abstraction, from functions as a service to AI-driven DevOps monitoring tools, has spurred new talk about a potentially "opsless" future.

Dynatrace monitoring competitors such as New Relic and AppDynamics, along with AI-driven IT automation specialists including Moogsoft and BigPanda, all envision a much less hands-on future for enterprise IT pros, but Dynatrace's strategy may be the most ambitious.

After the company re-emerged as a publicly traded entity in September, its executives outlined an aggressive approach to fully unsupervised IT automation within the company itself, and stated plans to bring that NoOps approach to customers as well. This week, the company took more concrete steps to advance that strategy with the launch of new IT training workshops and professional services focused on NoOps concepts, dubbed the Autonomous Cloud Enablement practice. It also touted a newly production-ready open source tool for DevOps monitoring and AI-driven automation integration with CI/CD pipelines called Keptn.

"Keptn is an open source pluggable control plane that can specify pipelines and generate them automatically, and [invoke Dynatrace] to do distributed transaction tracing and react to failed deployments," said Alois Reitbauer, VP, chief technology strategist and head of Dynatrace's Innovation Lab.

Keptn can be used independently of Dynatrace monitoring tools, but "it works better with Dynatrace," Reitbauer said. Keptn arose as a way to standardize the integration of Dynatrace monitoring with third- party DevOps and IT automation tools.

As of this week, Keptn is primarily focused on CI/CD automation, but Dynatrace plans to differentiate it from other CI/CD tools that also offer automated rollback features, such as Harness and Spinnaker, by broadening its features to include automated incident management and remediation.

"The key difference [between Keptn and tools such as Harness] is that Keptn isn't script-based, and covers more than continuous delivery to include IT ops automation," Reitbauer said.

Dynatrace's NoOps ambitions face skepticism

Dynatrace deployed a completely hands-off, NoOps approach to the back-end management for its own SaaS platform and Dynatrace customers have already expressed interest in following suit, including at least one customer that recently put Keptn into production use, according to company officials. Other longtime Dynatrace monitoring users such as Barbri, a Dallas firm which offers legal bar review courses, already used Dynatrace AI features for automated root cause analysis in 2018, and expressed a desire to move to automated remediation as quickly as possible.

However, in the broader market, NoOps faces significant challenges, from consensus on its definition to its perceived threats to the livelihoods of IT professionals. Early adopters point to unsolved challenges around data cleansing in data repositories that feed AIOps tools and the need for maturity among available tools as challenges to the NoOps vision. While Dynatrace users such as Barbri have a large appetite for NoOps, companies such as American Fidelity have prioritized other initiatives, including BizDevOps.

Automation tools themselves will require constant tweaking -- if you don't want to call that ops, that's fine.
Nancy GohringAnalyst, 451 Research

Furthermore, IT industry research reflects ambivalence among IT pros about whether unsupervised AI and advanced automation mechanisms will actually simplify IT management. For example, 451 Research's Voice of the Enterprise survey on storage budgets and outlook in 2018 asked 540 IT decision-makers whether they agreed or disagreed that the addition of machine learning and AI capabilities to vendor products will simplify IT management. Most of the respondents, 55%, answered "slightly agree," while 16% slightly disagreed and 9% strongly disagreed. Only 20% of respondents strongly agreed with the statement.

"Just automating a lot has value, and it's increasingly clear that automation will be important," said Nancy Gohring, analyst at 451 Research. "But automation tools themselves will require constant tweaking -- if you don't want to call that ops, that's fine."

Nancy GohringNancy Gohring

There's also confusion and apprehension among IT pros as they evaluate advanced IT automation tools, especially those driven by AI and machine learning, Gohring said.

"Vendors tell me that their customers are asking, 'Do we need to add a data scientist to our team to use your AIOps tool?'" Gohring said. "That's something that by definition they should not need, but the question tells you something; end users see this as a complicated concept and are concerned they don't have enough background to understand how [IT automation tools] come to certain conclusions."

Dig Deeper on DevOps Team Organization

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

How do you define NoOps?

NoOps is an interesting concept, and can perhaps be achieved for certain net-new cloud-native workloads. But to assume NoOps will become the de facto operations model for the typical G2K enterprise is naïve. The reality is that today’s enterprise is dealing with hundreds if not thousands of applications, hosted in a combination of physical, virtual, private, and public cloud environments.


What we hear from our customers is that there are significant challenges that need to be addressed in order to realize more autonomous operations.  Some examples:


  1. Data that feeds autonomous models is a mess – most enterprise data lakes simply are not comprehensive or accurate enough (in real-time) to inform an ML model that can reliably drive autonomous operations.  For a NoOps approach to work, there is a core, underlying assumption that enterprise IT can trust it’s data to be an accurate reflection of operational reality – and that gaps, latency, and redundancy don’t pollute or bias a model that drives automated actions. This is possible only with newer, cloud native workloads that are designed to operate in a homogenous environment and leverage native apis for data access and automation.  But these kinds of workloads represent but a small fraction of the typical enterprise application estate today.
  2. Autonomous systems must establish trust thru transparency of operation – Data science has evolved enough such that ML algorithms can be applied in targeted ways to provide reliable and relevant insights and recommendations.  Before autonomous actions can be trusted, however, the recommendations and suggested actions require validation.  Not enough work has been done in this area to ensure truly hands-off operations of most production workloads. Before humans turn over the operational “keys”, they need to understand how the machine intends to respond to various events and situations over time. 
  3. Autonomous models assume real-time and intelligent integrations with a heterogenous ecosystem – Not all data models or APIs are the same between vendors.  In a heterogenous ecosystem you need a sophisticated data ingestion engine that can map, transform, and align disparate data sources / types to a common model.  NoOps assumes that you have transformed the source data into a normalized state for quantitative analysis and can automate processes that span multiple systems.  Few IT Ops platforms have been able to accomplish this task given the wide variety of hosting environments, data sources, and protocols that comprise the typical enterprise IT estate.


To begin to realize the vision of more automated IT operations, ScienceLogic has decided to focus on tackling these data-centric problems as a first-order priority. Over the last decade we have helped enterprises of all sizes build a more comprehensive, authoritative, and real-time operational data lake. Moving forward, we are increasingly focused on leveraging ML techniques to reason over that data and provide highly qualified insights that drive automated actions. Our product capabilities and vision of desired state are more closely aligned to what the industry has begun to refer to as AIOps – intelligent IT automation driven by real-time data and machine learning. In contrast to NoOps, we believe AIOps represents a more pragmatic approach to managing the breadth, complexity, and velocity of today’s enterprise IT. By employing machine learning techniques and process automation, we can evolve from traditional “observability” tools, to an “advisory and action” platform that does not displace IT, but instead makes IT radically more efficient.


Conclusion: while fully autonomous operation may be possible in theory, in practice, there will be human professionals working with the machine for the foreseeable future.  Think of autopilot and adaptive cruise control, not SkyNet. 

- Michael Nappi, CPO, ScienceLogic