Integrates virtualization, distributed storage, software-defined networking, Kubernetes, backup and disaster recovery, and multi-cluster management in one platform. It runs traditional virtual machines, containerised applications, and on-premises AI inference at the same time.
Hyperconverged infrastructure (HCI) uses a single piece of software to integrate compute and storage on the same batch of standard x86 servers, managed end to end through one web interface. Servers (compute) and a SAN disk array (centralized storage), which used to be purchased separately, are combined into one. There is no longer a need to buy dedicated storage hardware separately.
Arcfra Enterprise Cloud Platform (AECP) takes a "full-stack enterprise cloud" approach. Beyond compute and storage, it also folds in software-defined networking (using software to segment networks and set firewall rules), the Kubernetes container platform (the standard tool for managing containerised applications), backup and disaster recovery, and multi-cluster management across data centers - all within the same platform. For a company with only two or three IT staff, the multiple software products that used to be purchased, learned, and renewed separately become a single license and a single console.
Related reading: Hyperconverged Infrastructure & VMware Alternatives | VergeIO VergeOS | Hyperconverged infrastructure selection guide
AECP is a set of modules running on the same cluster operating system. Below is a plain explanation of what each one does.
Arcfra Cloud Operating System. Installed directly on physical servers, it handles node management, cluster state, and resource scheduling - the foundation of the whole platform.
Arcfra Virtualization Engine. An in-house hypervisor based on KVM that splits one physical server into multiple virtual machines.
Arcfra Block Storage. Pools the SSDs/NVMe drives from each node into shared storage, keeping multiple copies of data, replacing an external disk array.
Arcfra File Storage. Provides a shared file service so multiple virtual machines or applications can read and write the same data at once.
Arcfra Network Service. VPC network segmentation, a distributed firewall (micro-segmentation, separating internal traffic between production, office, and DMZ), and load balancing.
Arcfra Kubernetes Engine. Creates a Kubernetes cluster through a graphical interface; the vendor claims three steps, about ten minutes. Virtual machines and containers share the same underlying resources.
Arcfra Backup & Disaster Recovery. Agentless backup, snapshots, and off-site replication.
Arcfra Operation Center. Centralized viewing and operation across data centers and sites, so multiple locations do not need to be logged into separately.
Migration tool. Moves existing virtual machines from VMware to AECP. Supports agentless, online migration and can be done in batches.
An open-source AI model serving platform (Apache License 2.0). Companies can host large language models (LLMs) in their own data center, paired with GPUs for on-premises AI inference. The commercial edition, Neutree Enterprise, is available on enquiry.
Scope: ANS only operates at the virtual layer, isolating internal traffic within the data center. It cannot replace physical switches or an outward-facing perimeter firewall. For external networking and security builds, see Enterprise Networking and Web & API Protection (WAF).
AECP's pricing logic is what procurement and finance teams find easiest to understand, and ask about most often.
Prices are not listed on this page. The actual quote depends on node count, core count, selected modules, and scope of services; please contact us for a quote.
Not every data center needs to switch platforms. The following three situations are where AECP's difference actually shows.
An objective account of the legal registration, investor structure, and technology origin, for reference during architecture evaluation and internal review.
Arcfra's legal entity name is ARCFRA PTE. LTD., incorporated in Singapore in May 2024, with headquarters also in Singapore. Investors include Vertex Holdings, part of Singapore's Temasek, and Samsung Ventures, part of Korea's Samsung.
The technology and founding team originate from the spin-off of hyperconverged vendor SmartX's business outside China: in 2024 Arcfra acquired SmartX's global business and product intellectual property outside China, and AECP and ACOS are the direct evolution of that software, while SmartX continues to operate in the China market.
Whether ICT product origin review applies, and under which regulation, is determined by your legal compliance, security, or procurement unit based on internal rules and the requirements of the competent authority. We can help request a country-of-origin declaration and written compliance documentation from the vendor for internal review.
If your organization has ICT product origin restrictions, the VergeIO option, from a US vendor, is also available. The two options suit different customer conditions and can be evaluated side by side based on review scope and workload type.
Choosing between the two comes down to three things: supply chain origin review, storage data volume and growth rate, and whether container and AI workloads already exist. The decision logic is laid out in the hyperconverged infrastructure selection guide.
We source vendor products and support through authorized channels in Taiwan, and provide planning, deployment, migration, training, and annual maintenance services.
For the full process and deliverables, see Services.
The questions most often asked in actual projects. Answers are based on publicly available vendor information and test results.
No. The Migration Tool supports agentless, online migration. AECP's specifications also support taking over an existing VMware ESXi environment, so this can be done in stages. The actual cutover sequence is finalised after validation in your environment during the POC stage.
AECP installs on standard x86 servers, and a cluster can mix machines of different brands and generations. Each server still needs to be checked against the vendor's hardware compatibility list (HCL), particularly disk and network card models. This is completed during the architecture planning stage.
Whether it "runs" and whether the application vendor issues a supported statement are two different things. Applications with strict certification lists, such as SAP or Oracle, need to be checked case by case. We help send enquiries to application vendors and validate on real hardware during the POC. If you use Odoo, see ERP Business Management.
We source vendor products and support through authorized channels in Taiwan. Our engineers handle first-line support, escalating to the vendor when an issue needs their involvement. The annual maintenance contract covers business hours, with scope and response method clearly stated in the contract.
Data center infrastructure typically has a five- to eight-year lifespan, so we recommend managing both the contract and the technical side. The contract should clearly state license continuation terms and the data export and exit path. On the technical side, the POC stage should verify that backups can be restored independently and virtual machine images can be exported, so data portability stays in your own hands.
Want to assess the current state first and confirm feasibility and migration scope? We recommend starting with a current-state inventory and TCO estimate.
Contact us View the hyperconverged infrastructure optionArcfra and AECP are trademarks or registered trademarks of Arcfra. GN-AI is an independent systems integrator. This page is not authorized, sponsored, or endorsed by Arcfra.
Leave your contact details and requirements, and an engineer will get in touch to help assess architecture planning, migration approach, and the scope of annual maintenance.