toddcretacci.com

Google Distributed Cloud Virtual (GDCV), VMware, Kubernetes, Hybrid Cloud, Platform Engineering

Since publishing my Google Distributed Cloud Virtual guide, I frequently get asked this question:

Should we deploy GDCV on VMware or bare metal?

The answer I give is:

It depends.

There is no universally correct answer because organizations are solving different business problems, working with different operational constraints, and starting from different levels of infrastructure maturity.

In many technology discussions, architects immediately jump into performance comparisons, infrastructure specifications, or Kubernetes design patterns. While those considerations matter, the real decision often comes down to operational strategy. The better question is:

Which product is a better fit for the organization?

If a company is just starting on its Kubernetes journey, starting out with a cloud provider like Google Cloud may be the best bet. If there’s a desire to deliver on-prem Kubernetes, both Google Distributed Cloud Virtual (GDCV) options can be considered a good solution.

GDCV supports both hypervisor-based and bare metal deployments, giving organizations flexibility in how they modernize infrastructure, adopt Kubernetes, and build hybrid cloud platforms. The key is understanding where each architecture excels.

As discussed in my previous article, A Guide to Google Distributed Cloud Virtual: Real-World Lessons from the Field, successful GDCV deployments are rarely determined by technology alone. They are determined by how well the platform aligns with existing operational practices.

WHY THIS DECISION MATTERS

Many organizations are interested in:

• Kubernetes adoption
• Platform Engineering
• Application modernization
• AI-ready infrastructure
• Hybrid cloud architectures
• Infrastructure automation

However, the path toward those goals looks very different depending on the environment. A company with a mature VMware footprint may already have:

• Established backup processes
• Operational runbooks
• Resource management procedures
• Disaster recovery workflows
• Monitoring platforms
• Virtualization expertise
• Experienced technology professionals

Meanwhile, another organization may be building a dedicated Kubernetes platform from the ground up and wants maximum performance and complete control over resources. The architecture choice should support long-term operational success rather than short-term technical preferences.

WHEN VMWARE IS THE RIGHT CHOICE

For many enterprises, VMware remains the easiest and least disruptive path toward GDCV adoption.

YOU ALREADY HAVE A VMWARE INVESTMENT

As discussed in my article How Organizations Can Extend Existing VMware Investments with Google Distributed Cloud Virtual, modernization does not require abandoning existing infrastructure. Many organizations have spent years building:

• VMware expertise
• Operational procedures
• Monitoring systems
• Security controls
• Disaster recovery strategies

Replacing those investments simply to deploy Kubernetes rarely provides a strong business case. By deploying GDCV on VMware (GDCVV), organizations can modernize while leveraging platforms they already understand.

GDCVV also provides a simpler deployment model. During deployment, VMWare templates are downloaded that not only include Google Kubernetes Engine (GKE) but also include the Ubuntu operating system, meaning both the OS and cluster are set up at the same time. This is a huge savings in labor costs, because the organization doesn’t need to license, or maintain them.

YOUR TEAM ALREADY KNOWS VMWARE

Technology transformations are often viewed as technical projects. Most are actually people projects. A VMware team already understands:

• Resource allocation
• Virtual machine lifecycle management
• Capacity planning
• High availability
• Disaster recovery
• Infrastructure troubleshooting
• Hardware and network requirements for VMWare

Rather than asking teams to learn everything at once, VMware allows DevOps teams to focus on learning:

• Kubernetes
• GDCV administration
• Platform Engineering practices
• Cloud-native operations

This reduces organizational risk significantly.

YOU NEED FLEXIBILITY

Virtualization introduces operational flexibility. Administrators can:

• Add resources more easily
• Clone environments
• Create development clusters
• Recover systems quickly
• Standardize deployments

For organizations operating multiple environments:

• Development
• Test
• Pre-production
• Production

GDCVV often simplifies lifecycle management because most of the hardware aspects will be maintained by the VMWare team.

YOU ARE MODERNIZING EXISTING APPLICATIONS

Many enterprises aren’t building greenfield environments. They’re supporting:

• Legacy applications
• Traditional middleware
• Existing databases
• Virtualized workloads

VMware allows organizations to modernize incrementally while keeping existing workloads operational. The journey toward cloud-native platforms becomes much smoother.

WHEN BARE METAL IS THE RIGHT CHOICE

GDCV on Bare Metal (GDCVBM) also have compelling advantages. In some situations, bare metal becomes the obvious architecture choice.

YOU WANT MAXIMUM RESOURCE EFFICIENCY

Virtualization introduces another layer of abstraction. Although modern hypervisors are highly efficient, bare metal eliminates that extra layer entirely. This means:

• Direct hardware access
• Reduced overhead
• Simpler architecture
• Full utilization of physical resources
• Operating system flexibility
• Better disk performance

Organizations operating large Kubernetes environments often appreciate those efficiencies. Being able to use their Linux version of choice means less abstraction for patching and management processes.

YOU ARE BUILDING A DEDICATED PLATFORM

Some organizations deploy Kubernetes as a strategic application platform rather than an extension of existing infrastructure. In those situations, bare metal can make sense because:

• Hardware is dedicated
• Resources are predictable
• Capacity planning becomes simpler
• Infrastructure ownership is clear

Teams can focus exclusively on Kubernetes operations rather than simultaneously managing virtualization layers.

YOU HAVE PERFORMANCE-SENSITIVE WORKLOADS

Some applications are sensitive to:

• Latency
• Throughput
• Hardware acceleration
• GPU access
• Storage performance

While VMware performs extremely well in most scenarios, certain workloads benefit from direct access to physical hardware.

This can be particularly relevant for:

• AI workloads
• Machine learning platforms
• High-performance data processing
• Specialized analytical systems

multiple hypervisor support

One of the downsides to running GDCVV is that VMWare licensing can be expensive. Recent changes to Broadcom’s licensing model have made it difficult for companies to want to operate an expansive environment.

As of this writing Google will support GDCVBM on other hypervisors like Microsoft’s Hyper-V. Having the advantage of using a per-existing hypervisor environment gives an organization a lot of flexibility. Before going forward verify with Google on what hypervisors are currently supported.

YOU WANT A KUBERNETES-FIRST OPERATING MODEL

Organizations building platform engineering practices may eventually reach a point where Kubernetes becomes the primary operational platform.

In these environments:

• Kubernetes is the platform
• Virtual machines become less important
• Automation becomes the operating model
• GitOps becomes the source of truth

Bare metal aligns naturally with this approach by reducing infrastructure layers and focusing directly on platform operations.

THE OPERATIONAL TRADEOFFS

Too many architecture discussions focus entirely on technology. Operational realities matter just as much.

VMWARE BENEFITS

VMware typically provides:

• Mature operational tooling
• Existing organizational expertise
• Familiar disaster recovery practices
• Flexible resource management
• Easier environment creation
• Patching and GKE versions in a pre-configured VMWare template

VMWARE CHALLENGES

VMware also introduces:

• Additional infrastructure layers
• Hypervisor dependency
• Added operational complexity
• Server-level monitoring and security tools may not be installable
• Cluster nodes are dynamic – created and destroyed with each upgrade / update

BARE METAL BENEFITS

Bare metal typically provides:

• Direct hardware access
• Simplified architecture
• Strong resource efficiency
• Kubernetes-focused operations
• Cluster nodes are static and do not change

BARE METAL CHALLENGES

Bare metal often requires:

• Additional infrastructure planning
• Additonal hardware lifecycle management
• Greater operational maturity
• Dedicated platform expertise
• Multi-level updates – OS and GKE

A PRACTICAL DECISION FRAMEWORK

When evaluating GDCV deployment options, I recommend asking the following questions:

Choose VMware if:

• You already have significant VMware investments
• Your operations team is virtualization-focused
• Risk reduction is a priority
• You want a gradual modernization path
• You need flexible development environments

Choose Bare Metal if:

• You are building a dedicated Kubernetes platform
• Maximum resource efficiency matters
• You have strong Kubernetes expertise
• Hardware performance is critical
• Platform Engineering is already mature

Consider Both if:

Many large enterprises operate both models. For example:

• Production AI clusters on bare metal
• Development and testing environments on VMware
• Different environments supporting different business requirements

Hybrid architecture decisions, including GKE on-prem and in the cloud, deliver the greatest flexibility.

WHAT I’VE LEARNED FROM REAL-WORLD DEPLOYMENTS

After working with GDCV environments, one pattern appears repeatedly:

Organizations rarely fail because they selected the wrong hardware platform.

They struggle because they underestimate:

• Operational readiness
• Documentation requirements
• Network planning
• Automation needs
• Platform governance
• Skills development

As I discussed in both Infrastructure as Code Beyond Terraform and A Guide to Google Distributed Cloud Virtual: Real-World Lessons from the Field, operational consistency matters more than infrastructure purity.

A poorly managed bare-metal cluster will not outperform a well-managed VMware-based platform from a business perspective. Technology decisions should support operational excellence, not replace it.

FINAL THOUGHTS

The reality is that both architectures can successfully support Google Distributed Cloud Virtual, but in my opinion, a company that is diving into Kubernetes for the first time should consider a cloud provider, like Google Cloud. A cloud provider will manage a large amount of the hardware lifecycle, which allows teams to focus on learning Kubernetes and DevOps practices.

If your organization already operates a mature VMware environment, GDCVV may provide the fastest path toward Kubernetes adoption and platform modernization. Of the two GDCV products, GDCVV allows Platform and Devops teams to focus on GKE and not about maintaining servers.

If you’re building a dedicated Kubernetes platform and want maximum control over infrastructure resources, GDCVBM may be the better choice.

The goal should not be selecting the architecture that appears most modern. The goal should be selecting the architecture that best aligns with your organization’s operational model, modernization objectives, and long-term strategy.

Here’s a chart that summarizes several of the pros and cons. Let me know what your experience has been like with these products.