Why Observability Is Becoming a Hybrid Cloud Imperative
As applications increasingly span mainframes, on-premises systems, containers and AWS services, enterprises need a unified approach to monitoring and logging
Many applications no longer run on a single platform. A transaction might begin on a mobile app, pass through an API gateway hosted on AWS, invoke containerized microservices, retrieve data from a cloud database and complete on an IBM z/OS application.
While hybrid architectures have enabled enterprises to adopt cloud services without replacing core systems, they have also created a new operational challenge: understanding what happens when something goes wrong.
The growing need for observability reflects a shift in enterprise IT, Roi Mor, co-founder and CTO at OpenLegacy, tells TechChannel. “The workload on the mainframe is growing and growing. Why? Because digital demand is growing, and eventually it needs to go back to the mainframe,” Mor says.
The promise of AI—and the applications it enables—is feeding enterprises’ renewed interest in cloud migration, “which they lost … 10 years ago,” Mor says.
While AI may assist in that migration, moving decades of business-critical systems remains a daunting process. “Taking tens of millions or hundreds of millions of lines of code that were written in 40 years and expecting to be running in the cloud in three years, it’s not realistic even with all the AI,” Mor says.
Instead, enterprises are accepting that hybrid architectures will remain in place for years to come. “No one did it in three years, but let’s say realistically the smart CIO says it’s going to take seven years. There is an understanding that in reality, at least for the next 5-10 years, we’re going to be in a hybrid world,” Mor explains.
Why Hybrid Environments Create Observability Challenges
Many organizations operate siloed mainframes, with overnight batch data replication or limited API-based synchronization between environments. But applications are becoming tightly integrated across on-premises infrastructure and cloud services, creating far more hybrid execution paths.
Cloud providers acknowledge that reality. “One of the AWS executives that I talked to admitted that they would say 30% is a reasonable workload to move in a couple of years,” Mor says. “Clients will still have 60, 70% workload remaining on the mainframe, because it doesn’t make operational sense to move it.”
The operational challenge, then, is understanding operations across the entire stack. “There will be more and more hybrid use cases. … This area is going to get more and more complex,” Mor says.
That makes unified observability increasingly critical. “If you need to live in a hybrid world, you need to take observability more seriously across your entire networks,” Mor says. “It’s not just about your cloud services, or your digital services or your certain environments across the networks, your entire networks. … You need to have your mainframe able to talk with your observability tool.”
From Infrastructure Monitoring to Observability
Traditional infrastructure monitoring was designed to answer relatively simple questions, such as “Is the server running? Is the CPU overloaded? Has storage reached capacity?” While the answers remain important, they provide only a partial picture in hybrid environments where a single business transaction can traverse a mainframe, Kubernetes cluster, APIs and cloud services. Teams need to understand system status as well as system behavior.
“The challenge is once you start making the leap,” Mor says. “Observability is all about what went wrong, and the more hops you need to make on the network, and more layers of software you need to go through, the problem becomes more complicated.”
Each additional layer introduces another potential point of latency, failure or misconfiguration, making it increasingly difficult to identify the source of performance issues through traditional monitoring alone.
Handling the Complexity of Distributed Systems
The challenge grows as applications become more distributed. “Move the business logic off the mainframe, and things even become more fragile,” Mor says, because previously business logic resided with mainframe data. “It was concrete. Now you have logic that maybe with AI was rewritten and running somewhere else, and maybe some portion of the business logic remains on the mainframe, and some portion of the data remains on the mainframe, so it’s becoming more and more of a problem.”
This is where observability differs from conventional monitoring. Rather than simply reporting the health of individual systems, it combines metrics, logs and distributed traces to provide context around how services interact, where dependencies exist and how users experience an application.
Distributed tracing assigns a unique trace ID that provides visibility into the path an individual request takes end-to-end as it moves between services, platforms and environments. API, microservices, databases and back-end systems record telemetry against the identifier.
That allows operations teams to determine why a failure occurred and pinpoint the exact service, API call or dependency responsible for the degradation in performance. As hybrid architectures continue to grow in complexity, this visibility is key to reducing downtime and improving service reliability.
Managing Scattered Telemetry Data
One of the biggest challenges is ensuring telemetry data remains consistent across different systems. OpenTelemetry has emerged as an open standard designed to address this by providing a framework for generating and managing telemetry data such as metrics, logs and traces.
Teams can send data to multiple back ends, which is particularly important for hybrid and multi-cloud environments. That helps connect observability data across platforms, reducing tool fragmentation in complex distributed systems.
In parallel with open standards, AWS’s cloud-native observability platform Amazon CloudWatch can collect and analyze metrics, logs and events to provide a centralized view of cloud workloads through log aggregation, dashboards and automated alarms. But it can also integrate with OpenTelemetry to extend visibility beyond AWS, consolidating data from on-premises systems and other cloud providers.
Building on CloudWatch’s role, AWS X-Ray helps developers to analyze and debug distributed applications, providing insights into the performance of underlying services. And AWS Systems Manager (SSM) acts as a central operational hub for hybrid environments, enabling teams to manage servers across AWS EC2 instances and on-premises infrastructure, deploy updates and roll out agents such as the CloudWatch agent across servers.
Aligning the People Element
While much of the focus on observability is on the tools and telemetry, the bigger challenge is often organizational. In many enterprises, different teams are responsible for different parts of the stack, including mainframe operations, infrastructure, cloud platforms and application development.
“Typically, the problem is that there’s lack of alignment,” Mor says. “There is big silo, a Chinese wall in many cases between on-prem workload and cloud workload, and there are different teams, different personas, different agendas, etc.”
Each group might have its own monitoring tools, dashboards, alerting systems and incident workflows. Even when tools are capable of providing end-to-end visibility, that can prevent organizations from acting effectively and slow down resolution times.
“You really need a strong CIO and CTO to dictate … to try to have a unified stack, especially if you’re understanding that you need to live in a hybrid world,” Mor says. “This is about the challenge to get the people aligned on the same stack at least, or the cross-traffic concern aspect.”
Why Observability Is Becoming Foundational
As enterprises move toward AI-driven operations, observability is more than a nice-to-have. AI operations (AIOps), automation and predictive incident management all depend on high-quality, consistent telemetry that spans the entire hybrid environment.
AI can help with interpretation as well as automation. “AI can give a better trackability, if you feed it with good information, or AI can do the actual analysis,” Mor says. “If something fails on the mainframe because of wrong data that was passed from … some edge case that was not found on the mainframe, because it caused an error, you’re able to track it back.”
This ability to correlate events across systems enables AI to move beyond log aggregation toward meaningful root-cause analysis. “I see those things in CI/CD, for example,” Mor says. “CI/CD is getting smarter and able to tell you if the bill fails, why it fails, and what the root cause, etc., not just dumping the logs.”
In hybrid environments in particular, the principle remains that you cannot automate what you cannot observe. Ultimately, observability succeeds when enterprises move from platform-based ownership to service-based thinking, with shared telemetry and unified incident response models.