Abstract: Eight features of elastic bare metal server are summarized. So ali Cloud elastic bare metal server (Shenlong) performance

The “elastic bare-metal server” we define here is a “new species”. It emphasizes “elasticity”, that is, through technological innovation, traditional bare-metal services can achieve the same use experience and business agility as virtual machines.

So, there are eight key characteristics of a “resilient bare metal server” :

(3) Characteristic one: Minute delivery

Rapid and flexible cloud service scaling is the core feature of IaaS cloud services. Generally speaking, VM services can achieve minute-level elastic service scaling.

In October 2017, Lu Han and Guan Xiaotong’s love affair caused the breakdown of Weibo. This kind of business is characterized by the sudden surge of business flow, which requires the background to have the elastic expansion capacity of minutelevel business. In this case, elastic bare metal servers must be able to deliver quickly at the minute level to handle sudden spikes in traffic.

(2) Fully compatible with cloud platform services such as VPC, SLB, and RDS. For customers who have used the cloud, it is important to be fully compatible with existing cloud platform services when protecting their current investment. In a VPC network, an elastic bare-metal server must have the same virtual network access and security isolation capabilities as a VM.

(3) Feature 3: Compatible with VM images The system is compatible with VM images, enabling smooth migration between elastic physical machines and VMS, and providing service deployment flexibility.

(4) Cloud disk startup and dynamic hot swap of data cloud disk Startup is a core feature of an elastic physical machine. If a cloud disk cannot be started, you must rely on the traditional PXE installation. PXE machine cannot deliver minute-level services, console image password repair and other key functions. Dynamic hot swap of data cloud disks is a key feature of elastic storage services, ensuring pay-as-you-go storage performance and capacity requirements

(5) The VNC, Web console, and openAPI user experience are consistent with the VIRTUAL Machine Web H5 VNC and Web console, which can greatly reduce the learning cost of users. Only openAPI interfaces that are consistent with VMS can protect customers’ investment in existing resource management systems.

Fault diagnosis Feature 6: Migration and recovery After a breakdown Automatic migration and recovery after a breakdown greatly reduces the fault recovery time of IaaS infrastructure. When a SOFTWARE or hardware fault occurs on a VM or elastic dedicated server, the O&M management and control system automatically migrates instances to minimize service interruption. The implementation of downtime migration relies on the local “stateless” of the elastic physical machine, and depends on key technologies such as BIOS configuration migration, cloud disk startup, and data cloud disk.

(7) Characteristic seven: free automatic operation and maintenance of human flesh

Below is a typical deployment process for a traditional physical server. (Source: 4 Hours: Breaking the rules and Building Blackstone’s Physical Server to Its ultimate Delivery)

While the deployment of traditional physical machines usually takes three working days, this mode requires at least four hours of delivery time for maximum optimization. The core feature of IaaS cloud service is human-free automatic operation and maintenance

(3) Characteristic 8: The complete features and performance of physical machines, as well as high isolation and high security virtual machines have the same seven characteristics, while flexible bare metal servers have the core competitiveness that virtual machines do not have:

Full processor features of physical machines: Technologies such as Intel VT-X and VT-D cannot be provided by VMS. Physical machines have these processor characteristics naturally. Due to the lack of technical support such as VT-X, nested virtualization is implemented on VMS, resulting in unacceptable performance. So for scenarios such as public cloud deployment on a private cloud, Typical examples are Openstack on public Cloud, VMware on public Cloud, Clear Container on public Cloud, and runV on public Cloud. Elastic physical machines are the only choice. When Intel vTune support, performance tuning, and profiling are dependent on Intel vTune, elastic machines are the only choice for newly imported processor instruction features such as Skylake AVX512. Elastic machines are available without waiting for Hypervisor adaptation. And because there is no complexity of Hypervisor software adaptation, its functional stability is naturally guaranteed

The lossless and high-performance virtualization technology of physical machines inevitably brings performance overhead, while physical machines do not have CPU and memory virtualization overhead, which has definite advantages.

High isolation Of dedicated servers Dedicated servers are completely isolated from each other in terms of resources, such as CPUS, memory, and I/OS

Physical machines with high security are truly user-exclusive, naturally immune to problems such as hypervisor escape, CPU microarchitecture side channel attacks, and so on

Ali Cloud flexible bare metal server (Divine Dragon) covers all the above features. Ali Cloud X-Dragon architecture and flexible bare metal server will broadcast “open box evaluation” for the first time at 3:30 PM on May 16th (This Wednesday). Brother Drill is invited to live together. !

The original link