Discussion on data center management operation in cloud computing environment

Editor's note: Management is an important guarantee for IT system benign operation, different IT equipment has its own management system. In particular, large-scale data centers, must be centralized management system to run the management of computing, storage, network and other equipment to be able to quickly respond to and deal with data center business changes, abnormal events, continuous optimization. In the "IP Navigator" to the period of the article has focused on "data center management", but most of the focus on "network-centric" management, this article will be magnified to the entire cloud computing environment in the data center, the three kinds of operation and management model of the comparative analysis.

Traditional data centers, infrastructure-level equipment, through standardized connectivity and protocol interoperability, to ensure that the computing, storage, network equipment management systems separated from each other (as shown in Figure 1), so that different operations and maintenance teams can be based on their own business development and the evolution of the trend of continuous improvement and deepening of their management procedures, Meet the requirements of the continuous development of the data center business.

Figure 1 Traditional data center management running architecture

In the cloud environment, the separate operating mode can not support the deployment of cloud services, the new IT operation model challenges the traditional management architecture:

Virtualization: The traditional data center in each physical server is only a single or several applications of the fixed operation, the business is basically the binding operation with the host, the host management, in a sense that is the management of the business. Cloud computing environment in a large number of servers using virtualization technology, each physical network port will be distributed up to dozens of virtual machines, physical host running a number of different operating systems and applications, network application density greatly increased, the network performance, specifications, reliability are put forward higher requirements, However, the manageability of virtual machine Network properties is facing great challenges.

Dynamic: The traditional data center of the business for the physical host, while the physical server is generally fixed connected to a network port, and business attributes single, whether network policy, security control are relatively fixed. As long as the host and network operational interface clear, system ownership is clear, the business is easy to expand, and can run smoothly. However, with the deployment of high density virtual machines in the cloud computing environment, the dynamic migration of virtual machines is a prerequisite attribute in virtualized environment based on service change, disaster tolerance and distributed computing. If the network cannot perceive this kind of dynamic computation way, the continuous operation will cause the business disorder, the operation dimension is not controllable, this requires the management system to have the dynamic computation perception ability.

Relevance: The current network and computing operate in a loosely coupled way, there is basically no information correlation interaction between network management and host system, so, for virtualization data center, virtual machine Dynamic computing characteristics, network can not perceive, network management system can not locate virtual machine, network for business security, control, Configuration, regulation can not be associated with virtual machines, can not achieve the flexible deployment and scalability under cloud computing.

Automation: In the non-virtualized environment, the business deployment generally has a relatively fixed, that is, the host location, network access comparison to determine, the operation and maintenance of the target and physical machine, physical port consistent, this situation, host systems, network Management system deployment, debugging and docking relatively easy. But in large data centers, in particular, the business process under the cloud computing environment, based on the traditional separation debugging is unable to effectively support the business model of cloud services, which requires that the entire service should be able to provide a simple delivery, and different systems (basic computing, network, the top of the host, network management system) can interact with the information and to complete the automated deployment and operation of all components based on consistent business requirements.

The goal of cloud computing management

In order to support cloud computing virtualization, dynamic, relevance, automated service requirements, the entire cloud computing system needs to have a unified operational management platform, the cloud services can be end-to-end automated deployment, while quickly responding to resource scheduling and business changes in the service requirements (see Figure 2).

The Unified service platform can block the cloud service supply level to the bottom of the different architectures, so that the user or business operations focus on the service level, do not pay attention to cloud computing resources (computing, network, storage) the technical attributes of its own.

On the administrative correlation structure of automated responses, the provision of cloud services needs to transform the business requirements to the requirements of the deployment of the underlying resources, and to form the corresponding underlying configuration to the different devices, while in the process of service change (including disaster tolerance, virtual machine migration, extension and other resources operation and scheduling), It can adjust the configuration, function and docking of the underlying equipment to match the business requirement.

2 How to choose a reasonable operation management model

Model one: Centralized and unified cloud computing operation Management

To achieve flexible cloud computing services, some people have proposed a centralized approach to managing the data center infrastructure (as shown in Figure 3). In this mode, the cloud Operation management platform can integrate computing, storage and network, and form a single interface on the user's operation plane, which is very clear in the logical structure and operation structure, and has few management levels.

Although this kind of structure realizes the unified business deployment to some extent, the basic resources automatic dispatch, but the limitation is obvious. Different IT systems are inherently professional, network, calculation, storage systems monitoring operation, fault handling, software and hardware upgrades, capacity and planning are completely different, in a control system to achieve both the unity of the business, but also to achieve the overall basis of management, not only the size of the system itself, complexity, functionality, Professional challenges, but also to support the management of the operation of the team, the operation of the coordination, knowledge system, professional crossover has produced a huge degree of complexity.

Even if a factory can be a very high degree of professional integration of the operation of a number of basic resources to such a unified system, this system will also be very large and complex, its own operation and maintenance will also have great difficulty.

Mode two: dual-generic management

The second model is a dual-generic management model. As shown in Figure 4, under the architecture of a similar first model, in addition to the unified Operation management platform, in the computing, storage, network systems integrated their own professional management system. Compared with model one, model two has greatly enhanced, not only simplifies the complexity of unified operation management platform, but also introduces the traditional mature operation and maintenance management, and separates the service operation and infrastructure management of cloud computing, and forms an IT running structure with division and cooperation.

But the disadvantage of this model is that for the underlying physical devices, there are two sets of instruction systems: a unified management platform for the supply of cloud services and an independent operation and maintenance system, if there are operational deviations, the need for the two systems to define or determine a priority sequence, Otherwise, under certain conditions, the service exception will be caused by conflicting instructions from different systems. At the same time, for the basic equipment, the two sets of instruction system call interface or protocol may be completely different, even because of the current lack of standardization, for different cloud management platform has different customization requirements, bringing the basic equipment operation and design complex.

Pattern 3:3-Layer Management

The third model is a three-layer management model. As shown in Figure 5, the unified cloud management platform runs at a logical level (top Tier), providing cloud users with a service interface, cloud service provisioning operations, and not directly managing and operating the underlying devices. The middle tier (middle Tier) is the basic resource Operation management layer, accepts the cloud service call from the upper level, and transforms it into the configuration operation for the underlying device, while the middle layer functions as a specialized system to perform operation, maintenance and supervision of the basic equipment. The lowest layer is the basic equipment level (infrastructure Tier), is the computing, network, storage and other basic cloud computing resources connected to the formation of the physical layers, to receive instructions from the top to run and provide services.

For the three-layer model, the middle management unified the command from the cloud service management platform and its own operational change instruction, forming a consistent operation set issued to ensure the uniformity of operation. Especially for cloud computing, the deployment of upper-level services, changes will always involve the underlying multiple systems of the interaction between the changes, such as virtual machine dynamic computing characteristics of its network location changes, storage resources will be due to data migration to create location changes, which involve computing, network, storage of information interaction between the objects, Protocol notifications, connectivity checks, etc. to ensure continuity and sustainability of the cloud service. The flow of data and the underlying protocol interaction occurs in the third plane, but in the middle tier different resources management control system also carries on the initiative to carry on the information transmission, like the virtual machine management system and the network Management system Interaction computation migration, the status and the position and so on information, this makes the cloud service management process more accurate and controllable, Enables the connectivity of all IT infrastructure resources and enables the deployment of cloud computing to evolve towards more sophisticated automation.

The further benefit of the three-tier management model is that as the instruction layer of the basic resource level, the middle management layer has the ability of changing demand because it is composed entirely of the software, that is, it can encapsulate various information from the service level and the heterogeneous system, and form the instruction of the lower order to carry on the basic equipment. As shown in Figure 6, each of the basic resources and its management software constitutes a flexible on-demand IT systems, their external interface is mainly implemented by management software, the current universal Soap/restful interface has been widely used in software system calls, The eVB technology is implemented as an example: The network system is tightly coupled to realize the internal operation control management, the virtual Management Center and the server virtualization system completely tightly coupled to achieve the virtual computing internal operation control management; At the infrastructure tier level, Network and virtual machine system through the standard technology eVB to achieve data interoperability and protocol interaction, this is the entire cloud computing can be automated, dynamic, relevance of the basic interoperability standard requirements. In the control layer, the network management system and the Virtual Administration Center can flexibly define the information (Virtual machine identification, service type, network mark, network attribute, etc.) between the two heterogeneous systems through the Soap/restful interface, so as to realize the underlying data flow of the whole cloud computing system, Control level Business attribute flow.

A summary of the comparison of three models

In terms of current domestic user applications, users of computing, network, storage separation management has formed a good experience, which is still a good reference in the cloud computing environment; In the architecture of the transformation/evolution of cloud computing, service delivery and IT operations may be independent of each other, but the former are dependent on the latter, The latter is the target of the former business approach, which requires the cloud management of the operating framework to have a lot of flexibility, but also to the basic level of precision control. Model one is a very natural structure that many users are currently thinking of, because this model obscures the difference between cloud services and cloud infrastructure, the model two and model three are the framework requirements of cloud computing and the traditional IT Operation management mode, which enables the user's it mode to migrate to cloud service in a gradual way.

3 concluding remarks

The applicable data center management operation model can not only make the business model clear and reliable, but also greatly improve the operation ability of the traditional data center, and make the operation mechanism of the conventional datacenter be reused. However, the different cloud computing service model has its own characteristics, based on its own operational capabilities, the requirements of the existing system, select and evolve to fit each cloud computing data center applicable mode, need users, manufacturers, service providers, continuous adaptation, adjustment to optimize the formation.

"Edit Recommendation"