Overview

• Supports the following cluster types:
  • Gateway proxy cluster: Manages baremetal nodes in the cluster through the gateway node. You need configure a gateway node and a provision network for the cluster in advance to enable the nodes' OS deployment and lifecycle management.
    • Provision network: A provision network is a dedicated network for PXE boot and image downloads while creating elastic baremetal instances in a gateway proxy cluster.
      • Before create a gateway proxy cluster, you need to deploy an IPv4 provision network to provide PXE services for baremetal nodes in the cluster.
      • Provision networks require high network performance. We recommend that you use at least 10 Gigabit NICs for your provision network.
      • You can configure a gateway for your provision network. This way, the provision network can be connected to other networks. If you do not need to connect your provision network to other networks, you do not need to configure a gateway for your provision network.
      • You can attach only one provision network to a gateway proxy cluster. However, you can attach a provision network to multiple gateway proxy clusters.
    • Gateway node: A gateway node is a node where the ingress and egress traffic of the Cloud and elastic baremetal instances in gateway proxy clusters is forwarded.
      • You can attach multiple gateway nodes to a gateway proxy cluster. However, you can attach only one gateway node to a gateway proxy cluster.
      • A gateway node is used to take over primary storages and assign storage space for elastic baremetal instances.
      • A gateway node provides iPXE, DHCP, and other services. It is used to deliver configuration settings to elastic baremetal instances.
  • DPU accelerated cluster: Directly deploys and manages baremetal nodes through DPU devices. You need to install DPU devices for each baremetal node in advance and do not need gateway node and provision network.
    • You must install a DPU device for each baremetal node before you can add it to a DPU accelerated cluster and the cluster manage the node via the dedicated DPU device.
• You can attach an L2 network to an elastic baremetal cluster to provide an extended L2 business network for elastic baremetal instances in the cluster. Elastic baremetal instances and VM instances that share the same L2 network can access each other without using the gateway. The L2 network that you can attach to an elastic baremetal cluster can be of the VLAN or NoVLAN type.

• You can set the startup methods for the baremetal nodes. The following two startup methods are supported:
  • Startup from a volume: Uses a volume as the system volume of the baremetal node to install and deploy the operating system.
  • Startup form a local disk: Uses a local disk as the system volume of the baremetal node to install and deploy the operating system.
  • If you select to startup the baremetal node from a local disk, you can choose whether to take over the original operating system.
    • If you choose not to take over the original operating system, the Local Disk (Non Take-Over) is used. When you use the baremetal node to create an elastic baremetal instance, this method helps download an operating system from the Cloud and install it on the instance. Meanwhile, the local system volume is formatted.
    • If you choose to take over the original operating system, the Local Disk (Take-Over) method is used. The elastic baremetal instance created from the baremetal node used the original system operating system stored on the local disk directly.
• The management node must be connected to the IPMI network to remotely manage baremetal nodes.
• Baremetal nodes must be configured with the BMC interfaces, IPMI addresses, ports, usernames, and passwords, and be connected to the IPMI network.
• A baremetal node can be distributed to only one elastic baremetal instance and an elastic baremetal instance can only be assigned one baremetal node.
• You can provide compute resources for elastic baremetal instances by using a baremetal node or elastic baremetal offering.

• You can get elastic baremetal offerings of baremetal nodes by obtaining their hardware information. Baremetal nodes with the same offering can be managed in a unified way.
• The baremetal offerings obtained from the node hardware information can be classified into 3 types according to their startup methods: startup from a volume, startup from a local disk (take-over), and startup from a local disk (non take-over).
• You can use an elastic baremetal offering to create an elastic baremetal instance. You can also release the advanced settings of baremetal nodes to avoid resource idling.
• You can create a pricing list for elastic baremetal instances based on elastic baremetal offerings. Then bills are generated for the elastic baremetal instances based on their usage.

• The following two startup methods are supported for elastic baremetal instances:
  • Volume: Uses a volume as the system volume of the elastic baremetal instance to install and deploy the operating system.
  • Local Disk: Uses a local disk as the system disk of the elastic baremetal instance to install and deploy the operating system.
  • The Local Disk (Non Take-Over) and Local Disk (Take-Over) methods are supported:
    • Local Disk (Non Take-Over): When you use a baremetal node to create the elastic baremetal instance, the operating system is downloaded from the Cloud and installed on the elastic baremetal instance. This method will format the local system disk.
    • Local Disk (Take-Over): When you use a baremetal node to create the elastic baremetal instance, the original operating system on the local system disk is used as the operating system of the elastic baremetal instance.
• The following describes the resources supplied to elastic baremetal instances of different startup methods:
  • Compute resource: To elastic baremetal instances of both the volume startup method and local disk startup method, the compute resources are provided by corresponding baremetal nodes,
  • Network resource:
    • Elastic baremetal instances of both the volume startup method and local disk startup method uses L3 networks on the Cloud as their business networks.
    • To elastic baremetal instances of volume startup method in gateway proxy clusters, provision networks provides the PXE service. Elsatic baremetal instances in DPU accelerated clusters rely on DPU agents for deployment instead of provision networks.
  • Storage resource:
    • Storage space for system volumes: Primary storage on the Cloud provide space for system volumes of elastic baremetal instances of the volume startup methods. Baremetal nodes provide space for system volumes of corresponding elastic baremetal instances of the local disk startup method.
    • Extended storage space: You can attach data volumes or block storage volumes to elastic baremetal instances to provide extended storage space.
      • Data volume: Provided by Shared Block or Ceph primary storage on the Cloud.
      • Block storage volume: Provided by Ceph or Vhost (ZHPS) primary storage on the Cloud. Ceph block storage volumes apply to elastic baremetal instances of both the volume startup method and local disk startup method. Vhost block storage volumes to elastic baremetal instances apply to elastic baremetal instances of local disk startup method only.
• We recommend that you create an elastic baremetal instance by using an image that has installed the agent. Otherwise, you cannot perform the following actions on the instance: open the instance console, modify the instance password, attach a volume to or detach a volume from the instance, and attach a network to or detach a network from the instance.
• By default, you can you an image whose BIOS mode is UEFI to create an elastic baremetal instance. If you need to use an image with a Legacy BIOS mode, contact the official technical support.
• You can configure business networks for elastic baremetal instances. If you attach an L2 network to the cluster where your baremetal nodes reside, elastic baremetal instances and VM instances that share the same L2 network can access each other without using the gateway.
• You can enable the elastic baremetal instances of the volume startup method to automatically release the associated baremetal node when it is powered-off. The baremetal node released can be used by other elastic baremetal instances, thus avoiding the resource idling.

The financial and insurance industries have high requirements over business deployment compliance and data security. In these scenarios, you can use Baremetal Management to secure dedicated resources, data isolation, easy management, and operation-tracking. This way, you can ensure the reliability and security compliance of your key business system and data.

Supercomputing, genome sequencing, and other high-performance computing scenarios have high requirements over the computing performance, stability, and timeliness. However, the virtualization may cause performance losses and hyperthreading may negatively influence the business. Deploying a reasonable number of baremetal clusters can solve these problems, meeting the high-performance computing requirements.

To meet business requirements, you may not want to deploy some key databases on VM instances while want to deploy the databases on physical servers that feature dedicated resources, network isolation, and guaranteed performance. In these scenarios, you can use Baremetal Management to provide dedicated high-performance physical servers for your applications.

Elastic Baremetal Management provides dedicated physical servers for your applications to ensure high performance and stability. In addition, this feature allows elastic scaling. You can apply for and scale resources based on your needs.

A single management node allows you to manage 10,000 baremetal nodes and scale-out according to your needs.

An elastic baremetal instance no longer depends on IPMI/BMC network (optional), and the NIC can be used as a baremetal node as long as it supports PXE boot or installed with a DPU device. The realization of the elastic baremetal instance does not associate any CPU or virtualization technology. As long as there is a system image that can be started, and no matter if it is an ISO, qcow2, or raw image, it can be used as a virtual hard disk by an elastic baremetal instance. That is, the elastic baremetal can be directly used on domestic servers.

In addition, the elastic baremetal management feature can be achieved by software only, without purchasing any proprietary hardware. It is compatible with all x86 and most of domestic ARM CPU architectures and supports mainstream x86 operating systems and some ARM operating systems.

In addition to turn on/off baremetal servers and deploy system, elastic baremetal management also supports the use of virtual resources on the Cloud, including VPC/flat/public networks, volumes, primary storage and other resources, which seamlessly connects physical resources and cloud resources, and greatly improves the availability of cloud resources.

All features of a VM instance can be directly operated on an elastic baremetal instance. And an elastic baremetal instance can use any advanced network services without purchasing additional network hardware.

An elastic baremetal instance supports both volume deployment method and local disk deployment method, respectively using the primary storage resource and the local disk on the Cloud, which combines the scalability of resources on the Cloud with the stable I/O and high throughput of local disks.

The primary storage resources used by an elastic baremetal instance can be shared by the KVM clusters of the Cloud, which relieves you from deploying additional storage resources. In addition, local disk deployment method supports take over the original system, which effectively ensures the business continuity.