5. Data center blade virtualization application

1) I. Virtualization application analysis

1.1 The challenges for data center

With the rapid development of IT technology, the expanding Internet value-added services, the software and hardware resources of the data center room is greatly enriched. The data center based on the traditional IT mode regards one server as one minimum dividing unit, or one application as one partition unit, or one user as one partition unit, which presents more problems when in application. In such application allocation model, some resources can not be effectively integrated, some applications have serious waste of resources and process, some applications are facing serious resource shortage, but there is no way to integrate resources.

The following problems exist in the data center:

Low resource utilization, high idle rate, and low operating efficiency: a considerable number of servers only run the very small B / S or C / S structure application procedure, such as basic service types as OA office system and website etc., while the new type multi-core processors are in serious idleness, but these free resources can not be integrated or applied.

The emergency response time is long, the service guarantee is poor: when failure occurs, after the failure is located at the server, the operation and maintenance personnel often have to spend a few hours to back up data, replace hardware, re-install operating system, re-install the application software, re-import the previous backup data and reinforce security. Failure recovery in the traditional IT architecture, it is not the operation level of operation and maintenance personnel that impact the emergency response time, which is determined by the features of the IT infrastructure mode in the traditional application mode. Normally, it takes about 4 hours to complete the hardware replacement and restart the process. 

The operation and maintenance cost is high: in the traditional IT architecture, the high operating cost usually is not the manual operation and maintenance personnel cost, but the electricity consumption, air-conditioning refrigeration and other resources consumed by the service, the greater the number of the server sets, the more the idle quantity and the higher the operation and maintenance cost.

1.2 An overview of data center virtualization

There are different definitions on the data center virtualization, while the core idea is the same, that is, it is a way to distinguish the priority of resources and allocate the server resources into the workload needing them most so as to simplify management and increase efficiency, thereby reducing resources reserved for a single workload peak value.

Through the virtualization of the data center, IT infrastructure is changed in structure fundamentally. IT infrastructure minimum unit changes from per server to per Virtual Environment. Through the deployment of virtualization products, according to the types of different applications, each server is divided into several virtual containers, each container accommodates an application, or allocates to a client, that is, it achieves the application integration of all physical servers on the logic level, you can achieve centralized management on the thousands of Virtualization Environment on several hundreds of servers, configure the centralized storage or the high-reliable applications on the unified management panel.

2) The advantage of Inspur virtualization

2.1 An overview of development of Inspur virtualization

As the first Research & Development and production manufacturer of servers, Inspur owns the first national high-performance server, storage key laboratory, Inspur has been focused on the research on server virtualization technology for many years, having an extensive client base in China server market, and Inspur has a more accurate grasp of client needs and a better understanding on the server virtualization technology. At the same time, as one of Intel Worldwide 16 Server Strategic Partners (Country Leader), Inspur benefits greatly from the years of close co-operation and development, long-term sharing of the forefront virtualization technology, thus Inspur has a higher starting point of virtualization technology and also plays a leading role in the promoting of virtualization application. Currently, Inspur cooperates well with Parallels, XenSource, Vmware, Microsoft, etc. Inspur can provide advanced, stable and reliable virtualization solution basing on users' demand in the industry.

2.2 Introduction of Inspur virtualization

In the mode of adopting the operating system virtualization, install the basic host operating system on one physical server, then install the operating system virtualization software on the host operating system. The former host system is divided into a number of Virtual Environment by the resources partition capability, each Virtual Environment acts as one minimum partition unit in IDC infrastructure, install and deploy the corresponding user, and will get a virtual dedicated server.

The virtualization technology adopted in operating system is greatly different from other virtualization technology, virtualization is located in the infrastructure layer, realizing the software partition and isolation, each Virtual Environment is part of the original operating system, there is no need to install a separate operating system, and also has all performance of the original system.

Overall solution of Inspur operating system virtualization system

3.1 The design objective of Inspur operating system virtualization system

Inspur experience in virtualization design and implementation: choose Inspur blade server NX660D as the physical server applied in virtualization, each blade can virtualizes a number of Virtual Environment. In order to ensure business continuity and sufficient expansibility at the same time, the operating system virtualizes the integrated backup function and configures the centralized backup.

The main technical indicators and characteristics are as follows:

End-to-end management: clients can understand the hardware layout through the Web interface, examine the status of the system and its components, understand the hardware information at any time, and the option card and the navigation bar can provide intuitive navigation. The network administrator can have more time to do forward-looking maintenance, and respond quickly to unexpected events.

Remote management, the "distant" and "instant" of IT maintenance: the management module integrates BMC and KVM over IP, wherever the administrator is, as along as he can log in the Internet, he is near the server virtually even though he is far from the server in the physically. After the administrator logs in the Internet remotely, through the concise interface, he will be very well informed of the status of the entire system, and he can perform the system configuration, distribution of storage pools and virtual disks, switch of faulty equipment, restart of power switch and server, the system and data backup, and all operations except hardware operation.

Rapid disaster recovery, and enhance the availability: the application born by the faulty server can be switched into another server easily within just a few minutes by virtue of the integration of disk-free server and virtualization storage, at the same time, the files can be rapidly documented through the system snapshot function and centralized backup, which can further protect the reliability of critical data. In addition, the system reliability and application is guaranteed by the a multiple security measures as N +1 power redundancy, hard disk advanced RAID, server mutual redundancy, virtual storage, etc..

Flexibility extending, effect dynamic triangle: adopt the hardware platform and software platform supporting Extended Memory 64-bit Technology, each blade can provide a number Virtual Environment (compatible with 32-bit application) which can switch between X32-bit and X64-bit operating systems, without changing the hardware.

3.2 Topology structure diagram of system solution

The blade server front-end can adopt the 100 mega network connection blade built-in giga switch, and the bandwidth is allocated appropriately in accordance with the node mummer. The blade server back-end adopts giga internal network connection so as to enhance the data migration rate of the centralized backup. If the giga internal network connection requirement is not fulfilled, the solution of mixing the backup network and data network transmission can be adopted, but this will reduce the throughput performance of the remote backup storage, the remote management end of the centralized management tool can either be deployed in the internal network, or in the public network to facilitate operation.

3.3 Hardware configuration of virtualization system

The operating basis of the operating system virtualization is the underlying operating system, known as the physical node, also known as the hardware node. The load of the entire operating system virtualization system depends on the actual configuration of physical nodes and software architecture. Factors restricting the virtualization capacity constraints are: CPU, physical memory, disk space, disk throughput, and operating system kernel resources.

Operating system virtualization technology can divide, administrate and schedule resources effectively; virtualization technology itself cannot create new resources without foundation. Therefore, it is of great importance to select hardware of high performance and reliability for the virtualization. In recent years, the rapid development of IT technology improves the server hardware configuration quickly, hyper-thread, 64-bit expansion and multi-core technology have become the guide technology of server; the high-capacity of memory particles enable the server memory generally exceed the 4GB limit of 32-bit system memory management; large-capacity high-performance storage is used in production system in large-scale. Therefore, the data center virtualization has adequate hardware support conditions and is suitable for virtualization.

3.4 Maintenance planning of the data center

Anti-network attacks

There are two ways of MAC address allocation for operating system virtualization: routing mode and bridging mode. In routing mode, all Virtual Environment and physical nodes share one MAC address. Viewing the switch and other network equipment, the physical server installed with virtualization is bound with a number of IP on one network card. In bridging mode, VPS can have its own MAC address, which is set by the system administrator manually. MAC address spoofing and other malicious network attacks from the same network segment can be avoided by binding the corresponding static MAC address on the switch.

Disaster recovery

In scene with centralized backup solution, historical image on the centralized storage can be used to perform rapid recovery. The recovery method is very simple, enter the interface of any management tool, click the corresponding VPS, click backup and bill all the backup history, choose the most recent backup and perform recovery. If failure of centralized storage equipment occurs, the backup data can not be extracted effectively, and then all Virtual Environment data can be recovered and rebuilt manually.