{"id":18668,"date":"2026-06-09T19:40:56","date_gmt":"2026-06-09T19:40:56","guid":{"rendered":"https:\/\/made4it.com.br\/blog\/alternatives-to-vmware-in-2026-discover-huawei-fusioncompute\/"},"modified":"2026-06-29T19:40:27","modified_gmt":"2026-06-29T19:40:27","slug":"alternatives-to-vmware-in-2026-discover-huawei-fusioncompute","status":"publish","type":"post","link":"https:\/\/made4it.com.br\/en\/blog\/alternatives-to-vmware-in-2026-discover-huawei-fusioncompute\/","title":{"rendered":"Alternatives to VMware in 2026: Discover Huawei FusionCompute"},"content":{"rendered":"\n

Anyone who works with virtualization has already realized that VMware has become a different conversation since the Broadcom acquisition. The focus has shifted from \u201cWhat features do I need?\u201d to \u201cHow much will this cost at the next renewal?\u201d For many customers, this shift has opened the door for Proxmox. It makes sense. But there\u2019s a third option that almost never comes up in initial meetings: Huawei FusionCompute<\/strong>. <\/p>\n\n

The platform is neither new nor experimental. It is currently at version 8.10, has extensive technical documentation, and is used by telecommunications companies, banks, and government agencies around the world. <\/p>\n\n

In Brazil, it still rarely comes up in discussions about replacing VMware. And maybe it’s time to take a closer look at it. <\/p>\n\n

What is FusionCompute?<\/h2>\n\n

FusionCompute is Huawei’s virtualization layer for data centers. It runs on physical servers, provides VMs, manages CPU, memory, networking, and storage, and is part of Huawei’s DCS solution. <\/p>\n\n

The architecture revolves around two components:<\/p>\n\n

CNA (Computing Node Agent):<\/strong> Runs on each physical server, implements the UVP hypervisor, and manages VMs locally.<\/p>\n\n

VRM (Virtual Resource Manager):<\/strong> The VRM is the central brain and handles cluster management. It controls resource scheduling, VM lifecycle management, IP and VLAN allocation, and provides the web-based administration interface. It runs in active\/standby mode with automatic failover within 1 to 2 minutes if the active node goes down. <\/p>\n\n

The UVP hypervisor is based on KVM, just like Proxmox, but it runs on a bare-metal architecture without an intermediate host OS, similar to ESXi.<\/p>\n\n

Comparison table of the three platforms<\/h2>\n\n

<\/a> The table below helps distinguish between actual architectural differences and differences that are merely due to commercial packaging.<\/p>\n\n

Functionality<\/th>VMware ESXi \/ vSphere 8<\/th>Proxmox VE 9.2<\/th>Huawei FusionCompute 8.10<\/th><\/tr><\/thead>
Hypervisor type<\/strong><\/td>Bare-metal (ESXi)<\/td>Type-1 (KVM + QEMU)<\/td>Bare-metal (UVP\/KVM)<\/td><\/tr>
Hosts per cluster<\/strong><\/td>up to 96<\/td>No explicit limit*<\/td>up to 128<\/strong><\/td><\/tr>
VMs per cluster<\/strong><\/td>up to 8,000<\/td>No explicit limit<\/td>up to 8,000<\/strong><\/td><\/tr>
Licensing<\/strong><\/td>Subscription required; min. 72 colors\/CPU <\/td>Open-source (optional paid support)<\/td>Huawei Business License<\/td><\/tr>
VM HA<\/strong><\/td>\u2705<\/td>\u2705<\/td>\u2705 Automatic, independent of the management node<\/td><\/tr>
Live Migration<\/strong><\/td>\u2705 vMotion<\/td>\u2705<\/td>\u2705<\/td><\/tr>
Storage Live Migration<\/strong><\/td>\u2705 Storage vMotion<\/td>\u2705<\/td>\u2705<\/td><\/tr>
Automatic load balancing<\/strong><\/td>\u2705 DRS (included in the bundle)<\/td>\u2705 Dynamic Load Balancer (since PVE 9.2)<\/td>\u2705 Integrated DRS + DPM<\/td><\/tr>
Automatic Power Saving (DPM)<\/strong><\/td>\u2705<\/td>\u274c<\/td>\u2705 Automatically shuts down idle hosts<\/td><\/tr>
Intelligent Memory Overcommit<\/strong><\/td>\u2705<\/td>\u2705 (ballooning)<\/td>\u2705 (ballooning + sharing + swapping; x86 and Arm)<\/td><\/tr>
Thin Provisioning<\/strong><\/td>\u2705<\/td>\u2705<\/td>\u2705<\/td><\/tr>
Distributed Virtual Switch<\/strong><\/td>\u2705 vDS (in the paid bundle)<\/td>\u2705 OVS\/SDN<\/td>\u2705 Native DVS<\/td><\/tr>
SR-IOV<\/strong><\/td>Limited<\/td>\u2705<\/td>\u2705<\/td><\/tr>
GPU Passthrough<\/strong><\/td>\u2705<\/td>\u2705<\/td>\u2705<\/td><\/tr>
GPU Virtualization<\/strong><\/td>\u2705 (NVIDIA vGPU)<\/td>\u2705 (NVIDIA vGPU, starting in 2025)<\/td>\u2705 (Integrated Intel vGPU)<\/td><\/tr>
Containers \/ Native K8s<\/strong><\/td>\u274c (Tanzu, separate and paid)<\/td>\u2705 (LXC; no native K8s support)<\/td>\u2705 Integrated K8s<\/td><\/tr>
Fully multi-tenant<\/strong><\/td>\u2705 (with add-ons)<\/td>\u274c<\/td>\u2705 (VPC, ECS, ELB, NAT, etc.)<\/td><\/tr>
Orchestrated DR<\/strong><\/td>\u2705 SRM (paid)<\/td>\u274c<\/td>\u2705 Built-in UltraVR<\/td><\/tr>
Agentless Backup with CBT<\/strong><\/td>\u2705 (through third parties)<\/td>\u2705 Proxmox Backup Server<\/td>\u2705 Native eBackup<\/td><\/tr>
Arm Support<\/strong><\/td>\u274c<\/td>\u274c<\/td>\u2705<\/td><\/tr>
Diagnostic Black Box<\/strong><\/td>\u274c<\/td>\u274c<\/td>\u2705<\/td><\/tr><\/tbody><\/table><\/figure>\n\n

The official Proxmox documentation does not specify a limit on the number of nodes per cluster. The practical upper limit depends on network hardware and latency; there are reports of clusters with more than 50 nodes in production using enterprise-class hardware. <\/em><\/p>\n\n

What stands out about FusionCompute<\/h2>\n\n

Cluster scale<\/h3>\n\n

FusionCompute supports 128 hosts and 8,000 VMs per logical cluster, with up to 4,096 hosts and 80,000 VMs managed per instance. VMware supports up to 96 hosts per cluster in vSphere 8. Proxmox does not have a documented upper limit, but stability in very large clusters depends heavily on how the management network was designed. <\/p>\n\n

For large enterprise data centers, support for 128 hosts per cluster with a total scale of 80,000 VMs is a major advantage.<\/p>\n\n

DRS and DPM: Energy-Saving Scheduling<\/h3>\n\n

FusionCompute’s DRS monitors host load in real time and automatically migrates VMs to balance CPU and memory usage across the cluster, without manual intervention. This is equivalent to what VMware offers in the vSphere bundle and what Proxmox now offers with the Dynamic Load Balancer in PVE 9.2. <\/p>\n\n

What sets FusionCompute apart here is DPM (Dynamic Power Management)<\/strong>: when the cluster\u2019s load drops, the system consolidates the VMs onto fewer hosts and automatically shuts down idle servers. When demand rises, the hosts are restarted. For environments with varying workloads throughout the day or week, this results in real energy savings without the need for external automation. <\/p>\n\n

Three-Layer Memory Overcommitment<\/h3>\n\n

FusionCompute combines ballooning, memory sharing (identical pages across VMs are consolidated into a single physical copy), and swapping to allow a server to provide more virtual memory than the available physical memory. Proxmox uses ballooning but does not natively implement memory sharing between VMs. The practical result is a higher density of VMs per host on FusionCompute, especially when workloads have similar memory usage profiles. <\/p>\n\n

This feature works on both x86 and Arm, which is a key advantage in environments with Kunpeng servers or similar systems.<\/p>\n\n

Three-dimensional resource-based QoS per VM<\/h3>\n\n

While Proxmox offers basic controls for CPU shares and limits, FusionCompute allows you to configure three independent parameters per VM:<\/p>\n\n

Quota:<\/strong> the proportion of CPU that the VM receives during resource contention among VMs on the same host.<\/p>\n\n

Reservation:<\/strong> the guaranteed minimum amount of CPU that the VM will always receive, even under high contention.<\/p>\n\n

Limit:<\/strong> The absolute maximum amount of CPU that the VM can consume, even when the host is idle.<\/p>\n\n

The same model applies to memory, networking, and storage IOPS. This matters when the same cluster hosts a mix of production, validation, and less predictable workloads. Without a well-defined resource policy, a resource-intensive VM can become a problem for another area. <\/p>\n\n

Five-Layer VM Isolation<\/h3>\n\n

The UVP hypervisor implements isolation between VMs in five dimensions: physical resources, vCPU scheduling (via VMCS), memory (with three levels of addressing that prevent VMs from accessing each other\u2019s memory), internal networking (with VRF and packet filtering), and disk I\/O.<\/p>\n\n

In addition, each host has an embedded Black Box<\/strong> that records critical kernel information in non-volatile memory when an exception occurs. In the event of any system failure, the diagnostic data is preserved for post-mortem analysis, which greatly simplifies troubleshooting in production environments. <\/p>\n\n

Multi-tenancy as part of the architecture<\/h3>\n\n

Proxmox was not designed for multi-tenancy. VMs from different customers can coexist on the same cluster, but network and storage isolation between them is the responsibility of the person configuring the system, not the platform. <\/p>\n\n

FusionCompute, integrated with the full DCS stack via eDME, offers native multi-tenancy with a VPC per tenant, fully isolated ECS, ELB, NAT, a virtual firewall, and separate storage policies. For those who need to deliver IT resources as a service to different business units or customers, this is an architectural difference. <\/p>\n\n

Backup and Disaster Recovery within the Huawei stack<\/h3>\n\n

In Proxmox, the Proxmox Backup Server handles backups well, but orchestrated disaster recovery (DR) across sites requires third-party solutions. In VMware, Site Recovery Manager covers this, but it is an additional paid product. <\/p>\n\n

In the DCS stack, UltraVR<\/strong> provides orchestrated disaster recovery (DR) with support for topologies involving two, three, or more data centers, with zero RPO in active-active and automated failover scenarios. eBackup<\/strong> performs continuous incremental backups of VMs using CBT (Changed Block Tracking), without requiring an agent to be installed inside each VM. <\/p>\n\n

Who does this make sense to?<\/h2>\n\n

FusionCompute doesn’t really fit into every discussion. For small environments with few hosts and teams looking for low licensing costs, Proxmox remains a very strong option. <\/p>\n\n

The game changes when you have dozens of hosts, a formal SLA, cross-site disaster recovery, segmentation by area or customer, and the need to operate everything with fewer loose ends. In this type of environment, FusionCompute starts to make more sense. <\/p>\n\n

It also works more seamlessly when the customer is already using Huawei hardware, especially servers, OceanStor storage, or CE switches. Integration doesn’t solve everything, but it reduces operational friction. <\/p>\n\n

Arm could also be a factor here. As Arm chips gain ground in data centers, FusionCompute\u2019s native support for Arm processors\u2014which offer feature parity with x86\u2014could be a deciding factor in environments that are already considering Kunpeng servers or similar architectures. <\/p>\n\n

Migrating from VMware requires careful consideration. Migrating to Proxmox may be the right decision in many cases, but when the requirement is to maintain enterprise features such as DRS, orchestrated DR, per-VM QoS, and multi-tenancy, FusionCompute delivers these without requiring the setup of a large stack of external tools. <\/p>\n\n

Market Context<\/h2>\n\n

A 2024 Gartner survey showed that 74% of IT leaders were actively evaluating alternatives to VMware. vSphere 7 reached end-of-support in October 2025, which accelerated migrations for those who were still putting off the decision. And customers who accepted 1-year bridge agreements in 2024 are now facing renewals with full VCF pricing. <\/p>\n\n

The move away from VMware is no longer just a theoretical discussion. In many environments, it has become a project, a budget item, and a risk. It is in this context that FusionCompute needs to be evaluated: not as a curiosity from Huawei, but as an enterprise-grade option for those who do not want to trade one business problem for an operational one. <\/p>\n\n

Next Steps<\/h2>\n\n

If the discussion about VMware has already reached the budgeting stage, it might be worth considering FusionCompute. Made4it can assist with this technical review: architecture, comparison with Proxmox and VMware, migration risks, operations, and disaster recovery (DR) design. <\/p>\n\n

Contact us to see if it makes sense for your environment before deciding on a migration path.<\/p>\n\n

<\/a> <\/p>\n\n

Sources: FusionCompute Product Documentation v8.10.0 (Huawei, May 2026) \u00b7 DCS Product Documentation v2.2.0 (Huawei, February 2026) \u00b7 Proxmox VE 9.2 Release Notes (May 2026) \u00b7 Broadcom VMware Configuration Maximums vSphere 8.0<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

Anyone who works with virtualization has already realized that VMware has become a different conversation since the Broadcom acquisition. The focus has shifted from \u201cWhat features do I need?\u201d to \u201cHow much will this cost at the next renewal?\u201d For many customers, this shift has opened the door for Proxmox. It makes sense. But there\u2019s a third option that almost never comes up in initial meetings: Huawei FusionCompute. The platform is neither new nor experimental. It is currently at version 8.10, has extensive technical documentation, and is used by telecommunications companies, banks, and government agencies around the world. In Brazil, it still rarely comes up in discussions about replacing VMware. And maybe it’s time to take a closer look at it. What is FusionCompute? FusionCompute is Huawei’s virtualization layer for data centers. It runs on physical servers, provides VMs, manages CPU, memory, networking, and storage, and is part of Huawei’s DCS solution. The architecture revolves around two components: CNA (Computing Node Agent): Runs on each physical server, implements the UVP hypervisor, and manages VMs locally. VRM (Virtual Resource Manager): The VRM is the central brain and handles cluster management. It controls resource scheduling, VM lifecycle management, IP and VLAN allocation, and provides the web-based administration interface. It runs in active\/standby mode with automatic failover within 1 to 2 minutes if the active node goes down. The UVP hypervisor is based on KVM, just like Proxmox, but it runs on a bare-metal architecture without an intermediate host OS, similar to ESXi. Comparison table of the three platforms The table below helps distinguish between actual architectural differences and differences that are merely due to commercial packaging. Functionality VMware ESXi \/ vSphere 8 Proxmox VE 9.2 Huawei FusionCompute 8.10 Hypervisor type Bare-metal (ESXi) Type-1 (KVM + QEMU) Bare-metal (UVP\/KVM) Hosts per cluster up to 96 No explicit limit* up to 128 VMs per cluster up to 8,000 No explicit limit up to 8,000 Licensing Subscription required; min. 72 colors\/CPU Open-source (optional paid support) Huawei Business License VM HA \u2705 \u2705 \u2705 Automatic, independent of the management node Live Migration \u2705 vMotion \u2705 \u2705 Storage Live Migration \u2705 Storage vMotion \u2705 \u2705 Automatic load balancing \u2705 DRS (included in the bundle) \u2705 Dynamic Load Balancer (since PVE 9.2) \u2705 Integrated DRS + DPM Automatic Power Saving (DPM) \u2705 \u274c \u2705 Automatically shuts down idle hosts Intelligent Memory Overcommit \u2705 \u2705 (ballooning) \u2705 (ballooning + sharing + swapping; x86 and Arm) Thin Provisioning \u2705 \u2705 \u2705 Distributed Virtual Switch \u2705 vDS (in the paid bundle) \u2705 OVS\/SDN \u2705 Native DVS SR-IOV Limited \u2705 \u2705 GPU Passthrough \u2705 \u2705 \u2705 GPU Virtualization \u2705 (NVIDIA vGPU) \u2705 (NVIDIA vGPU, starting in 2025) \u2705 (Integrated Intel vGPU) Containers \/ Native K8s \u274c (Tanzu, separate and paid) \u2705 (LXC; no native K8s support) \u2705 Integrated K8s Fully multi-tenant \u2705 (with add-ons) \u274c \u2705 (VPC, ECS, ELB, NAT, etc.) Orchestrated DR \u2705 SRM (paid) \u274c \u2705 Built-in UltraVR Agentless Backup with CBT \u2705 (through third parties) \u2705 Proxmox Backup Server \u2705 Native eBackup Arm Support \u274c \u274c \u2705 Diagnostic Black Box \u274c \u274c \u2705 The official Proxmox documentation does not specify a limit on the number of nodes per cluster. The practical upper limit depends on network hardware and latency; there are reports of clusters with more than 50 nodes in production using enterprise-class hardware. What stands out about FusionCompute Cluster scale FusionCompute supports 128 hosts and 8,000 VMs per logical cluster, with up to 4,096 hosts and 80,000 VMs managed per instance. VMware supports up to 96 hosts per cluster in vSphere 8. Proxmox does not have a documented upper limit, but stability in very large clusters depends heavily on how the management network was designed. For large enterprise data centers, support for 128 hosts per cluster with a total scale of 80,000 VMs is a major advantage. DRS and DPM: Energy-Saving Scheduling FusionCompute’s DRS monitors host load in real time and automatically migrates VMs to balance CPU and memory usage across the cluster, without manual intervention. This is equivalent to what VMware offers in the vSphere bundle and what Proxmox now offers with the Dynamic Load Balancer in PVE 9.2. What sets FusionCompute apart here is DPM (Dynamic Power Management): when the cluster\u2019s load drops, the system consolidates the VMs onto fewer hosts and automatically shuts down idle servers. When demand rises, the hosts are restarted. For environments with varying workloads throughout the day or week, this results in real energy savings without the need for external automation. Three-Layer Memory Overcommitment FusionCompute combines ballooning, memory sharing (identical pages across VMs are consolidated into a single physical copy), and swapping to allow a server to provide more virtual memory than the available physical memory. Proxmox uses ballooning but does not natively implement memory sharing between VMs. The practical result is a higher density of VMs per host on FusionCompute, especially when workloads have similar memory usage profiles. This feature works on both x86 and Arm, which is a key advantage in environments with Kunpeng servers or similar systems. Three-dimensional resource-based QoS per VM While Proxmox offers basic controls for CPU shares and limits, FusionCompute allows you to configure three independent parameters per VM: Quota: the proportion of CPU that the VM receives during resource contention among VMs on the same host. Reservation: the guaranteed minimum amount of CPU that the VM will always receive, even under high contention. Limit: The absolute maximum amount of CPU that the VM can consume, even when the host is idle. The same model applies to memory, networking, and storage IOPS. This matters when the same cluster hosts a mix of production, validation, and less predictable workloads. Without a well-defined resource policy, a resource-intensive VM can become a problem for another area. Five-Layer VM Isolation The UVP hypervisor implements isolation between VMs in five dimensions: physical resources, vCPU scheduling (via VMCS), memory (with three levels of addressing that prevent VMs from accessing each other\u2019s memory), internal networking (with VRF and packet filtering), and<\/p>\n","protected":false},"author":5,"featured_media":18670,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1178],"tags":[1183,920,1182,1179,1180,1181,1184],"class_list":["post-18668","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-virtualization","tag-broadcom","tag-huawei","tag-huawei-fusioncompute","tag-proxmox","tag-virtualization","tag-vmware","tag-vmware-migration"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t