Sangfor HCI

Two hosts minimum for virtual storage (aSAN); three or more recommended for production, and required for striping or multiple datastores.

Up to 64 hosts per cluster, ideally no more than 48 to limit fault-domain size. Scale beyond that by adding clusters under SCP.

Not supported for HCI: virtual storage is not available on a single host and Sangfor says not to deploy a one-host HCI cluster.

Supported as a 2+1 design: two hosts plus a witness node, either the aServer-J-100Z witness device or a witness VM (4 vCPU, 16 GB) on HCI or VMware. 6.12.0 adds active/standby witnesses.

Stretched cluster (active-active) with RPO 0 and about 30 s RTO: at least four hosts plus a witness, RTT of 1 ms or less, up to about 50 km.

Compute-storage decoupled clusters using external storage are supported, and datastores can be carved by host or by disk; there is no separate storage-only node type for aSAN.

FC, iSCSI and NFS external storage are supported natively, with HA and live migration; 6.12.0 adds NVMe over RoCEv2 and lets one VM spread disks across aSAN and external LUNs.

Flash volumes and NVMe devices are supported; SPDK-based Turbo mode bypasses the kernel for I/O-heavy VMs, and RDMA storage networks are optional. aSAN design remains cache-plus-capacity.

No inline deduplication or compression is documented for aSAN primary storage; compression applies to backups and archives only. Data reduction and erasure coding are features of the separate aStor product.

Not available in aSAN, which protects data with 2 or 3 replicas; Sangfor offers erasure coding only in aStor, its standalone storage product.

2 or 3 replicas per volume or per VM policy, server-room fault domains in stretched clusters, rebuild up to 1 TB per 30 min, bad-sector prediction and sub-health disk isolation.

Redirect-on-write snapshots, scheduled snapshots, consistency-group snapshots (e.g. Oracle RAC), linked clones and instant full clones.

Transparent virtual-disk encryption before data is written to storage, TLS for external traffic and encrypted internal replication; backup copies can be encrypted. External KMS support is not documented publicly.

Storage policies include automated QoS tiers (high, default, low) per VM; aSV also offers CPU, memory and I/O resource guarantees.

Not part of aSAN; SMB/NFS shares require Sangfor aStor (a separate product) or a third-party NAS.

Not part of aSAN; S3 object storage comes from Sangfor aStor, which SCP 6.12.0 can manage from its console. Backup copies can target object storage.

Virtual iSCSI: aSAN can present iSCSI targets to physical servers, third-party hypervisors or other devices.

Native DR between Sangfor HCI or MCS sites with a minimum RPO of 1 s (CDP-based) and minutes-level RTO; no distance limit.

Zero-RPO protection via stretched clusters (within about 50 km, RTT 1 ms or less) and CDP hot-standby DR for critical systems.

Disaster Recovery Management covers active-passive DR and active-active stretched clusters for homogeneous and heterogeneous sources, with one-click switchover; runbook depth is less than dedicated SRM tools.

Built-in backup (full then forever-incremental with CBT), instant VM recovery, backup health checks, archiving, backup copies and CDP with second-level recovery points. CDP is a licensed feature.

aSEC adds ransomware protection, antivirus and vulnerability management inside the platform; immutable (WORM) snapshot retention is not documented for aSAN.

Veeam Backup & Replication 13.1 supports aSV through the Veeam Plug-in for Sangfor aSV 1.0; other vendors can integrate via Sangfor OpenAPI and SFVDDK for agentless backup.

DR to Sangfor Managed Cloud Services (MCS), which runs the same HCI stack; no native DR target in AWS, Azure or Google Cloud.

aNET provides VXLAN overlay networking with virtual switches, routers and a drag-and-drop topology; SCP adds VPCs and elastic IPs. NFV appliances (firewall, ADC, SSL VPN) plug in.

VM-level micro-segmentation with policy auto-generation and a distributed firewall; aNI can recommend DFW policies from observed traffic.

SCP uses a three-level role model (platform administrator, tenant, tenant user) with approval workflows. SSO/SAML and MFA options are not detailed publicly.

Hardened kernel, non-root aCLI troubleshooting access, platform security baseline checks via aSEC and security advisories; no published STIG or secure-boot attestation.

Sangfor Cloud holds CSA STAR Level 1 and SAP certification; no FIPS 140 or Common Criteria certificate for HCI is published on sangfor.com.

SCP manages many HCI clusters, VMware vCenter clusters and bare-metal hosts, with hierarchical multi-site management and pooled licences.

Quick, rolling and offline upgrade methods with pre-checks (aDeploy); patches and upgrades from SCP. Server firmware lifecycle is handled for aServer appliances only.

OpenAPI, a VMware-compatible CLI with over 230 PowerCLI-equivalent commands, and aCLI. No official Terraform provider or Ansible collection is listed on Sangfor pages.

SkyOps AIOps covers 150+ failure scenarios, predicts issues 7-30 days ahead, predicts disk and memory failures, and forecasts capacity 90 days out.

SCP provides multi-tenancy, quotas, self-service work orders with approval, and built-in billing per resource unit.

Live migration (including across clusters, in bulk), node and VM HA, proactive HA and AI-assisted DRS.

SCP offers a PaaS layer for containers managed by Kubernetes on Sangfor cloud; Sangfor publishes little detail on versions or CSI support.

Homogeneous hybrid cloud with Sangfor MCS (same HCI stack) managed from on-premises SCP, plus AWS integration for heterogeneous management.

Sangfor aDesk VDI runs on the same aSV stack (the aSV licence covers VDI use); Citrix or Omnissa Horizon validation is not published.

PCIe passthrough and licensed aGPU virtual GPU support, plus an SCP GPU scenario solution; no packaged AI inference platform.

SCP includes PaaS Database-as-a-Service and middleware-as-a-service options; scope is narrower than dedicated DBaaS products.

Install from ISO or USB, or use pre-installed aServer appliances; cluster initialisation and the aDeploy checker run from the web UI. Typically a day per cluster once racked.

Sangfor aServer appliances, or third-party servers verified in the HCI Compatibility Chart (574 entries at review, covering servers, components and guest OSes).

Two hosts plus a small witness device or VM suits branch sites; central management from SCP. No one-node HCI option.

Offline licensing by exported device-information file and offline upgrade packages are supported, so sites can run disconnected.

Sangfor Cloud Migration Tool (SCMT) and aHM: assessment, agentless or agent-based live sync, 5-10 min cutover, automatic driver injection; also P2V and OVA/OVF import.

No free edition; trial licences are available through Sangfor or partners.

Licences are bound to cluster device information; SCP can reallocate pooled licences between its HCI clusters. No public-cloud portability.

On aServer appliances Sangfor supports hardware and software; on third-party servers the hardware stays with the server vendor.

24x7 remote support on every plan; Sangfor does not publish a P1 response-time target on its site.

Veeam integration, SAP certification and a partner-led channel across Asia, Europe and the Middle East; a smaller third-party ecosystem and skills pool than VMware or Nutanix.

KVM-based with qcow2 disks and VM export; but the storage, networking, management and licensing are proprietary, and the aSAN/aStor split limits storage choices.