DEV Community

Cover image for Which Platform Gives the Best Visibility and Control Over Cloud Instances, Containers, and VMs? (2026): A See-Versus-Act Matrix by Workload Type
Noah Grinstein
Noah Grinstein

Posted on Originally published at controlplane.com AI-assisted

Which Platform Gives the Best Visibility and Control Over Cloud Instances, Containers, and VMs? (2026): A See-Versus-Act Matrix by Workload Type

No single tool gives both full visibility and full control over cloud instances, containers, and VMs on every cloud, so the best choice depends on whether you need to see workloads or act on them. Observability platforms such as Datadog and Dynatrace see the most, control platforms such as HPE Morpheus Enterprise and Azure Arc offer the broadest ongoing lifecycle control, and platforms that also run the workloads, such as Control Plane, give both for the workloads they run.

Disclosure: Control Plane publishes this guide.

TL;DR

Seeing a workload and acting on it are different jobs, and the right platform depends on which job you need for cloud instances, containers, and VMs.

  • Hyperscaler tools such as AWS Systems Manager, Azure Arc, and Google Cloud’s fleet tools go deepest in their own cloud and reach other environments through agents, connectors, or attached Kubernetes clusters.
  • Azure Arc and HPE Morpheus Enterprise offer the broadest lifecycle control across environments, while Datadog and Dynatrace see nearly everything and act only through workflow automations that call other tools’ APIs.
  • Red Hat OpenShift with OpenShift Virtualization and SUSE Rancher with Harvester manage containers and VMs together on Kubernetes through KubeVirt.
  • Control Plane fits teams that want to run containers across AWS, GCP, Azure, Oracle Cloud Infrastructure, and on-premises locations, and VMs on Control Plane’s Managed Kubernetes (MK8s) locations, without having to operate Kubernetes clusters, with visibility and control built in.

Which platform offers the best visibility and control over cloud compute instances, containers, and virtual machines?

The answer depends on which workload types you run, how many clouds and on-premises environments you span, and whether you need to see, act, or both.

Platform Category Cloud instances Containers VMs (on premises / hypervisor) Clouds and environments What it does not cover
AWS Systems Manager + CloudWatch Hyperscaler native See and act (patching, inventory, automation) See (ECS and EKS metrics and logs) See and act on hybrid-activated servers AWS; Azure and Google Cloud VMs through Cloud Connector; on-premises servers and other clouds through hybrid activations No container orchestration actions; Oracle Cloud Infrastructure and other machines outside Azure and GCP need a manual agent install and hybrid activation
Azure Arc + Azure Monitor Hyperscaler native, multi-cloud control See and act (Arc-enabled servers on AWS, GCP, other clouds, and on premises) See and act (Arc-enabled Kubernetes on any conformant cluster) See and act (Arc-enabled VMware vSphere and SCVMM) Azure, AWS, GCP, other clouds through the agent, on premises, VMware, SCVMM Automatic onboarding covers AWS and GCP (preview) only, so Oracle Cloud Infrastructure VMs need a manual agent install; container runtime protection outside Azure needs Defender for Cloud
Google Cloud (Compute Engine, GKE, Google Cloud Observability) Hyperscaler native See and act (Compute Engine) See and act (GKE, plus fleet management of attached EKS, AKS, and on-premises clusters) See and act (Google Cloud VMware Engine for vSphere VMs) GCP; Kubernetes clusters on AWS, Azure, and on premises through fleets Reach outside GCP is Kubernetes only, with no control of other clouds’ instances
HPE Morpheus Enterprise Cloud management platform See and act (provision, resize, decommission across clouds) See and act (Kubernetes cluster provisioning and container management) See and act (VMware, Hyper-V, Nutanix, SCVMM lifecycle) AWS, Azure, GCP, Oracle Cloud, VMware, Hyper-V, Nutanix, bare metal No deep observability; relies on integrations for metrics and traces
VMware Cloud Foundation (vCenter, VCF Operations) Virtualization platform No lifecycle control of public cloud instances See and act (vSphere Kubernetes Service clusters on VCF) See and act (full VM lifecycle on vSphere) VMware on premises and VCF-based hyperscaler services (Amazon EVS, Azure VMware Solution, Google Cloud VMware Engine) No management of native public cloud instances
Red Hat OpenShift + OpenShift Virtualization Kubernetes platform No See and act (container orchestration, monitoring, policies) See and act (KubeVirt VMs alongside containers) On premises, AWS, Azure, GCP, IBM Cloud, Oracle Cloud Infrastructure No management of cloud instances outside its clusters
SUSE Rancher + Harvester Kubernetes platform No See and act (multi-cluster container management) See and act (Harvester VMs integrated in Rancher) Containers on premises, AWS, Azure, and GCP; Harvester VMs on on-premises bare metal No management of cloud instances outside its clusters
Datadog Observability See (infrastructure monitoring, cloud integrations) See (container and Kubernetes monitoring, APM) See (agent-based VM monitoring) AWS, Azure, GCP, on premises, any host the agent supports No native provisioning, patching, or orchestration; actions run through Workflow Automation calls to cloud APIs
Dynatrace Observability See (full-stack monitoring, cloud integrations) See (Kubernetes and container monitoring, root-cause analysis) See (agent-based VM and host monitoring) AWS, Azure, GCP, on premises, any host OneAgent supports No native provisioning or patching; remediation runs through Workflows connectors (Kubernetes, AWS, Azure, Ansible)
IBM Turbonomic Resource optimization See and act (resize, scale, suspend, start, delete unused volumes) See and act (container resource resizing) See and act (resize, move, suspend VMs on vSphere, Hyper-V, Nutanix) AWS, Azure, GCP, Oracle Cloud Infrastructure, VMware, Hyper-V, Nutanix, Kubernetes, OpenShift Does not deploy new workloads from templates; actions are limited to optimization
Terraform / OpenTofu Infrastructure as code Act (provision, modify, destroy across clouds) Act (provision Kubernetes resources and container services) Act (provision VMs on vSphere, Hyper-V, and clouds) Any environment with a provider No runtime metrics, logs, or traces; visibility is limited to declared state and drift detection
Control Plane Production workload platform (containers and VMs) See and act on Managed Kubernetes worker nodes, which Control Plane provisions and autoscales on supported providers; does not manage standalone instances See and act (runs containers on AWS, GCP, Azure, Oracle Cloud Infrastructure, and Control Plane’s Bring Your Own Kubernetes (BYOK) and MK8s locations, with built-in logs, metrics, traces, and cost view) See and act (VM workloads through the KubeVirt add-on on MK8s locations with KVM-capable nodes) AWS, GCP, Azure, Oracle Cloud Infrastructure, on premises through MK8s or BYOK Instances, VMs, or clusters not connected to Control Plane

Hyperscaler stacks give the deepest integration in their own cloud. AWS Systems Manager reaches Azure and Google Cloud VMs through Cloud Connector and on-premises or other-cloud servers through hybrid activations, Azure Arc brings servers, Kubernetes clusters, and vSphere and SCVMM VMs into Azure’s management plane, and Google Cloud reaches outside GCP only for Kubernetes clusters, through fleets and GKE attached clusters.

Observability platforms cast the widest net for visibility. Datadog and Dynatrace collect metrics, traces, logs, and cost data from instances, containers, and VMs everywhere, and their workflow features can trigger scripted fixes, but they do not manage workload lifecycle. IBM Turbonomic sits between observability and control. It sees resource use and executes resizing, placement, suspend, and capacity actions, though it does not deploy new workloads.

Infrastructure-as-code tools such as Terraform and OpenTofu act broadly, provisioning and changing resources across clouds, but they track only declared state and drift, with no runtime monitoring. Platforms that run workloads directly, such as Control Plane, see and change those workloads through the same API.

What do visibility and control mean for each workload type?

Each workload type needs different things to see and do, and tools fall short in predictable places.

Workload type Visibility you need Control you need Where tools usually fall short
Cloud instances (EC2, Compute Engine, Azure VMs) Inventory, configuration state, patch status, cost attribution, security posture, performance metrics Provisioning, patching, resizing, stopping, tagging, access management Hyperscaler consoles show other clouds’ instances only after agent or connector onboarding
Containers on managed Kubernetes (EKS, GKE, AKS) Pod and node health, resource use, network topology, image vulnerabilities, deployment status Scaling replicas, rolling updates, admission policies, RBAC, network policies Observability tools see containers but cannot enforce policies; Kubernetes dashboards show state but lack cost attribution
Containers on serverless container services (Cloud Run, Fargate, Azure Container Apps) Request metrics, latency, error rates, cold starts, cost per request or per vCPU-second Deployment, revision management, concurrency limits, IAM bindings Little access to underlying infrastructure metrics; scaling behavior is opaque; control does not carry across clouds
VMs on vSphere, Hyper-V, or KVM Guest OS health, hypervisor metrics, storage IOPS, snapshot state, license compliance Provisioning, live migration, snapshots, patching, decommissioning Cloud-native tools reach on-premises hypervisors only through agents or connectors; VM platforms often lack container awareness
VMs on Kubernetes (KubeVirt) VM health as a Kubernetes resource, node resource pressure, storage provisioning status VM lifecycle through Kubernetes APIs, live migration, resource limits, scheduling affinity Observability tools without a KubeVirt integration treat VMs as pods; Kubernetes dashboards may not surface VM-specific metrics
Bare-metal servers Hardware health, firmware versions, IPMI or BMC status, power state Firmware updates, BIOS configuration, OS provisioning, power cycling Most cloud management tools assume a hypervisor or container runtime; bare-metal visibility needs dedicated agents or out-of-band management

Which platform fits which situation?

Choosing a platform starts with your current environment, team, and main workload types.

Situation Start with Why What you still need
One cloud, mostly instances Your hyperscaler’s native tools (Systems Manager, Azure Monitor, Google Cloud Observability) Deepest integration, lowest setup effort, no extra licensing A cross-cloud observability tool if you add a second cloud; an IaC tool for repeatable provisioning
Azure-heavy with on-premises servers Azure Arc + Azure Monitor Arc-enabled servers extend Azure management to on premises and other clouds; Arc-enabled VMware vSphere manages existing VMs A configuration management tool for complex OS fleets; a container security platform if you run Kubernetes
VMware estate moving to containers VMware Cloud Foundation with vSphere Kubernetes Service, or HPE Morpheus Enterprise VCF manages existing VMs while vSphere Kubernetes Service adds Kubernetes; Morpheus manages VMs and containers across clouds An observability platform for application metrics and traces; a FinOps tool for cost attribution
Kubernetes everywhere, some VMs OpenShift with OpenShift Virtualization, or Rancher with Harvester Both run containers and VMs on Kubernetes through KubeVirt A cloud management platform for non-Kubernetes instances; an observability stack if built-in monitoring falls short
Many clouds, need one dashboard Datadog or Dynatrace for visibility; HPE Morpheus or Azure Arc for control Observability platforms see across every cloud; Morpheus and Arc act across clouds and on premises An IaC tool for provisioning; a policy engine for governance; a FinOps platform for cost optimization
Teams that want production containers and VMs with built-in visibility and control, without having to operate Kubernetes clusters Control Plane Runs containers natively on four clouds and VM workloads on MK8s locations, with observability, cost, and changes through one API A hyperscaler tool or cloud management platform for resources not connected to Control Plane

How do you combine tools without overlap?

Combining tools works when each one has a clear job, because overlap means two tools acting on the same resource, conflicting inventories, or duplicated cost data.

  • Designate one inventory source of truth. If you use Azure Arc, let Arc’s resource graph be the canonical list of servers and clusters; if you use Datadog, let its infrastructure map be the canonical view.
  • Assign one place to act per workload type. Instances go to Systems Manager or Morpheus, containers to your Kubernetes or workload platform, and VMs to vCenter, Morpheus, or your Kubernetes-based VM tool.
  • Enforce a tagging standard across clouds. Tags drive cost allocation, access policies, and automation scope, and a tag that exists in AWS but not in Azure breaks cross-cloud cost and access reports.
  • Keep one audit store. Forward CloudTrail, Azure Activity Log, and platform audit logs into one searchable store so every change to a workload lands in one place, and standardize governance across dozens of cloud accounts so the trail stays consistent.
  • Split actions by lifecycle phase. If Terraform provisions a VM and Morpheus also changes its configuration, one will overwrite the other, so let the IaC tool create and destroy resources, let the operations tool patch and resize them, and run drift detection to catch crossover.

The 10 best multi-cloud management platforms compares the wider landscape.

How does Control Plane give visibility and control over containers and VMs?

Control Plane runs workloads across clouds and on-premises locations, so the same platform that runs them shows and changes them, and Control Plane handles Day 2 operations for you, including autoscaling, built-in observability, and security.

For containers, Control Plane runs serverless, standard, stateful, and cron workloads natively on AWS, GCP, Azure, and Oracle Cloud Infrastructure, and on Kubernetes clusters you already operate through BYOK or on clusters Control Plane provisions through MK8s. A Global Virtual Cloud (GVC) is Control Plane’s named set of locations, and a workload runs in every location of its GVC, active-active across regions and clouds.

For VMs, Control Plane runs VM workloads through the KubeVirt add-on on MK8s locations with KVM-capable nodes, either bare metal or nested-virtualization instances. On MK8s locations, Control Plane also provisions and autoscales the worker nodes, so the instances under your workloads are visible and managed too.

Logs and metrics are collected automatically and explored in a built-in Grafana, and traces use Control Plane’s OpenTelemetry tracing provider once a sampling rate is set. The console’s Cost & Usage view shows spend per org and per resource with a projected total, and a tamper-proof audit trail records every platform action, showing who changed what across the org. A single view of cost and security across multiple clouds covers how to join this with data from other tools.

The CLI, console, Terraform and Pulumi providers, and the Control Plane MCP server all work through the same Control Plane API. The MCP server lets AI assistants deploy, diagnose, scale, and roll back workloads, and access is controlled through Control Plane policies. Control Plane is PCI DSS Level 1, SOC 2 Type II, HIPAA, and GDPR compliant.

Control Plane’s scope ends at what it runs, so instances, VMs, or clusters not connected to Control Plane stay with a hyperscaler tool or a cloud management platform such as Azure Arc or HPE Morpheus Enterprise, and teams evaluating hybrid cloud platforms or Kubernetes management platforms often pair one of those with Control Plane.

FAQ

Short answers on visibility and control over instances, containers, and VMs.

Which platform offers the best visibility and control over cloud compute instances containers and virtual machines?

No single platform is best at both across all three workload types on every cloud. For visibility, Datadog and Dynatrace cover instances, containers, and VMs across all major clouds. For lifecycle control, Azure Arc and HPE Morpheus Enterprise reach the most environments. For teams that run containers and VMs on one platform, Control Plane provides both for the workloads it runs on AWS, GCP, Azure, Oracle Cloud Infrastructure, and on-premises locations.

What is the difference between cloud visibility and cloud control?

Visibility means seeing what exists and how it performs, including inventory, health, metrics, cost, and security posture. Control means acting on workloads by provisioning, scaling, patching, stopping, migrating, and enforcing policy. Observability tools see broadly and execute changes only through workflow automations, while orchestration platforms execute changes but may not see resources outside their own scope.

Can one tool manage both containers and virtual machines?

Yes. Red Hat OpenShift with OpenShift Virtualization and SUSE Rancher with Harvester use KubeVirt to run VMs as Kubernetes resources alongside containers, HPE Morpheus Enterprise manages both across clouds, and Control Plane runs containers and VM workloads under the same API.

Is Azure Arc good for managing AWS and on-premises servers?

Yes, especially when Azure is your governance center. Arc-enabled servers bring AWS instances, GCP instances, other clouds’ machines, and on-premises servers into Azure for inventory, policy, monitoring, and patching. Arc-enabled Kubernetes projects any conformant cluster into Azure, and Arc-enabled VMware vSphere and SCVMM manage on-premises VMs through the Azure portal.

Does Datadog let you control infrastructure?

Partly. Datadog is primarily an observability platform. Its Workflow Automation and App Builder can trigger actions such as starting, stopping, or rebooting EC2 and Azure VMs and scaling Auto Scaling Groups, but it does not provision infrastructure, patch operating systems, or orchestrate containers. Teams pair it with Terraform for provisioning, Systems Manager for patching, or a Kubernetes platform for orchestration.

Can Kubernetes run virtual machines?

Yes, through KubeVirt, an open-source project that runs VMs alongside containers on Kubernetes. OpenShift Virtualization is Red Hat’s supported implementation, Harvester is SUSE’s, and Control Plane uses KubeVirt to run VM workloads on its Managed Kubernetes locations.

What is the best single pane of glass for multi-cloud compute?

For visibility, Datadog and Dynatrace come closest, covering instances, containers, and VMs through agents and cloud integrations. For control, HPE Morpheus Enterprise and Azure Arc offer the broadest reach for provisioning and lifecycle management. No tool gives complete visibility and control for every workload type on every cloud, so most teams combine an observability platform with a control platform.

How does Control Plane manage containers and VMs?

Control Plane runs containers as serverless, standard, stateful, and cron workloads on its managed AWS, GCP, Azure, and Oracle Cloud Infrastructure regions or on your own Kubernetes clusters, and runs VM workloads on its Managed Kubernetes locations. Logs, metrics, traces, cost, and the audit trail are built in, and every change goes through one Control Plane API.

Do I need separate tools for visibility and control?

In most cases, yes. Observability platforms collect and correlate data across environments and act only through automations, while orchestration and management platforms execute changes but have narrower cross-environment visibility. The cleanest setup assigns one inventory source of truth, one place to act per workload type, and a shared tagging standard.

Talk to the Control Plane team to see containers and VMs run across clouds under one API, or sign up free to try it.


Originally published on the Control Plane blog.

Top comments (0)