Increase VM density with HPE Morpheus Software memory overcommitment

[HPE_Experts]

Learn how memory overcommitment in HPE Morpheus Software helps increase VM density, improve utilization, and reduce host requirements.

In virtualized infrastructure, memory allocation often limits virtual machine (VM) growth before physical memory is fully consumed. As AI, analytics, and other data-intensive workloads expand, memory has become one of the most valuable resources in the data center. IT teams must add capacity while controlling hardware costs, power consumption, supply-chain constraints, and physical footprint. Yet many environments still strand usable memory because placement decisions are based on configured allocation rather than real utilization.

Memory overcommitment in HPE Morpheus Software helps unleash that stranded capacity by placing workloads based on actual utilization patterns instead of worst-case allocation assumptions. The result is higher VM density, better hardware utilization, and a smaller infrastructure footprint while preserving operational control. In one large HPE hybrid cloud engineering labs deployment, memory overcommitment supported a 32% reduction* in host requirements while maintaining resiliency objectives.

HPE Morpheus is available in three variants, allowing organizations to start with the capabilities they need and expand over time. HPE Morpheus Software – VM essentials provides unified VM management, automation, HVM, and VMware vCenter coexistence. Advanced adds on-premises private cloud capabilities, including self-service provisioning, HPE Kubernetes Service, automation, and software-defined networking. Enterprise extends the platform with full hybrid cloud orchestration, broader integrations, FinOps, microsegmentation, and governed agentic operations. Memory overcommitment and the HVM hypervisor are available in all three variants, so the density benefits described here apply across HPE Morpheus Software – VM essentials, advanced, and enterprise.

What is memory overcommitment?

At the virtualization layer, HPE Morpheus provides a single control plane for managing and automating VMs across VMware vCenter and HVM environments. HVM is HPE KVM-based hypervisor, available in every variant, and can run alongside existing VMware estates to support modernization without requiring an all-or-nothing migration.

For VMs, the HPE Morpheus manager monitors parameters including VM memory assignment and host memory capacity to determine placement. A portion of host memory is reserved for the host operating system. For example, the manager may place nine VMs with 10 GB of assigned memory on a 100 GB host, using 90 GB of VM-available capacity. By default, a tenth VM of the same size would be denied because total assigned memory would exceed the host memory available for VM resources.

With HPE Morpheus Software 9.0, a tunable parameter was introduced to allow administrators to overcommit memory on HVM hosts.

While a VM may be configured with 8 GB of memory, actual guest OS consumption may be much lower. In the previous example, if all VMs were using only 25% of assigned memory, actual consumption would be roughly 22.5 GB, yet placement would still be capped at nine VMs. The overcommit feature allows administrators to define an overcommit ratio as a percentage. Setting it to 200%, for example, would allow twice as many VMs to be placed on the same host, assuming each VM’s actual utilization remains below 50% of assigned memory.

Benefits of memory overcommitment

The most immediate benefit of enabling memory overcommitment is increased VM density. More VMs can run on the same physical hardware without additional investment. For organizations managing large numbers of workloads with predictable memory utilization, this can dramatically improve return on existing infrastructure.

Overcommitment also provides greater scheduling flexibility. HPE Morpheus has more placement options available when memory headroom is calculated against actual utilization trends rather than worst-case assigned values. This can reduce placement failures, improve overall cluster efficiency, and support management tasks that require live migration, such as host updates and reconfigurations.

For organizations actively migrating away from legacy virtualization platforms, overcommitment can ease the transition by allowing more VMs to be hosted during a consolidation period without requiring immediate hardware expansion.

Risks and considerations

Memory overcommitment requires discipline. If VM memory consumption spikes because of workload demand, runaway processes, or sudden bursts of activity, the host may experience memory pressure and degraded performance across affected VMs. Not every workload is a good fit, and some require a more conservative approach. The table below provides examples.

Good candidates	Use with caution
Dev/test environments	Databases
CI/CD workloads	In-memory caches
Predictable application servers	Analytics engines
VDI deployments	Memory bursty/greedy applications
Low utilization VMs	Real-time processing

It is also worth noting that overcommitment ratios that are too aggressive amplify risk. A 200% ratio on a host where actual utilization frequently approaches 60–70% of assigned memory leaves very little safety margin.

These risks can be mitigated with disciplined monitoring and data-backed configuration. Use the included monitoring in HPE Morpheus, and complement it with tools such as HPE OpsRamp Software where available, to track utilization trends and identify pressure before it affects workload performance.

Best practices

Before enabling overcommitment, establish a baseline. Use HPE Morpheus or HPE OpsRamp Software monitoring tools to observe actual memory utilization across your VMs over a representative period. This gives you a realistic picture of typical consumption and peak demand.

Start conservatively. A ratio of 150% is a reasonable starting point for most environments. This provides a meaningful increase in VM density while maintaining a comfortable buffer against utilization spikes. Increase gradually, and after validating that existing workloads remain stable.

Ensuring sufficient swap space is configured on each host is essential when enabling memory overcommitment. Swap acts as a last-resort buffer, preventing out-of-memory conditions when actual consumption approaches or exceeds physical RAM, though at the cost of performance. In general, higher overcommit ratios require more swap headroom. Configure the ratio so swap remains a last-resort buffer and, ideally, is rarely or never used.

Set up alerting on actual host memory consumption. If a host begins approaching its physical memory ceiling, you want to know before it impacts workload performance.

Finally, consider segmenting your clusters. You may find it best to run dev and test workloads on highly overcommitted clusters, while running critical databases on clusters without the feature enabled at all. You can configure overcommit on a per-host basis within a cluster, and pin VMs to specific host sets, so you can guarantee certain VMs always run on hosts without overcommit enabled while others take full advantage of it.

How to enable

The HPE Morpheus Software documentation explains how to enable memory overcommitment for HVM hosts. Edit the host under Infrastructure > Compute > {Host} > Edit > Advanced

Real-world example

Our HPE hybrid cloud lab team manages a large HPE Morpheus deployment consisting of 248 hosts and between 5500 and 6000 running VMs at any given time. Without memory overcommitment, many hosts had reached their maximum allocated capacity while actual memory utilization across the fleet averaged only around 39%. The team analyzed long-term resource utilization trends and segmented its deployment into CPU-optimized and memory-optimized clusters, targeting 60–65% actual memory utilization across primarily six-host clusters.

For this team, that 60–65% utilization target is ideal for more than just efficiency; it also provides built-in resilience. In a six-host cluster, two hosts can be lost and the remaining four will be able to support the full workload. This approach makes a projected 32% reduction in host footprint viable, freeing approximately 80 hosts—about two racks of hardware—for redeployment elsewhere.*

Conclusion

For organizations modernizing virtualization, memory overcommitment in HPE Morpheus offers a practical way to increase VM density, improve infrastructure efficiency, and reduce hardware requirements. Because HVM and memory overcommitment are available across HPE Morpheus Software  –  VM essentials, advanced, and enterprise, customers can improve utilization while choosing the operating model that fits their needs today and supports their modernization path tomorrow. With the right monitoring and disciplined configuration, unused memory capacity can become measurable business value without compromising performance or resilience.

Ready to improve VM density, utilization, and virtualization economics? Learn more about HPE Morpheus Software and how HPE private cloud solutions can help modernize your infrastructure with greater flexibility and control.

Visit HPE.com/Morpheus to learn more about HPE Morpheus Software

*Based on HPE internal study

Meet the authors:

James Shupe, Sr. Principal Engineer
Jamie Reed, Systems/Software Engineer
Praveena Patchipulusu, VP Engineering