Dealing with cloud management, starting with the data center

As data centers, Grid computing, supercomputing, cloud computing and other technologies and concepts rise, IT industry is also moving towards business model, technology architecture to management operations and other aspects of the direction of change. At the same time, cloud management technology has gradually entered our vision, the topic of cloud management also more and more hot. In terms of user requirements, technical features and functional composition, cloud management is mainly the management of data centers. The management focuses on the integration of key resources and business, visualization and virtualization, while Cloud management focuses on on-demand resources and cloud charging operations; At present, although the experience and technology of data center management are more mature, the related technology of cloud management is still not well developed. The future direction and goal of data center management will be cloud management.

How to manage Cloud? What should we focus on at this stage? From Taobao, Tencent and other domestic cloud computing applications pioneer it construction and management, may be able to gain enlightenment.

Cloud management in the eyes of the pioneer

Daily average of 400 million pages visited, daily turnover of 600 million yuan, the annual turnover of 200 billion yuan, ..., this is Asia's largest online trading platform-Taobao. The IT infrastructure behind these stunning numbers is the tens of thousands of servers, thousands of network devices, and hundreds of applications running in 8 data centers across Hangzhou and across the country. For Taobao, the future cloud computing service model is "b2c+c2c+ Network Marketing + Cloud Leasing Service", is the inheritance and development of existing business, so first of all to the existing IT infrastructure (especially the data center) to consolidate, and the corresponding cloud management is the data center and the underlying infrastructure for integrated management. Specifically, there are three levels:

Equipment level. The management of large capacity equipment (tens of thousands of servers and network devices) needs to be realized, while the distributed and logically unified management needs should be considered.

Business level. It needs to realize the integration of it and IP devices in the same platform, can manage the network from the business point of view, and can monitor and optimize the business from the angle of performance and flow.

Service level. Support for operational services is needed to help the IT department shift to a standardized, auditable service operation Center.

In general, Taobao is currently involved in the cloud management is actually data center management, in accordance with infrastructure management-> upper level business and flow analysis->it service operation order, integrate a variety of resources, including equipment, applications, traffic, services, for the future to build a virtual pool of resources, Provide a foundation for the external provision of cloud services.

Like Taobao, Tencent's cloud management is also focused on managing the infrastructure of the underlying data center. In addition to focusing on resource consolidation, Tencent is further focused on virtualization and automation of resources. This includes two aspects: first, the management of virtualized resources (including virtual network devices, virtual hosts, etc.), the ability to view the status of these virtual resources, and then the automated management of resource pools and the ability to configure physical resources and virtual resources. All in all, the first integration of resources, and then the virtualization of resources and automation, these are Tencent's current cloud management requirements.

Cloud management starts with data center--Data center management solution

Taobao and Tencent from these two cases can be seen, the current cloud management is still in the initial stage, is essentially a data center management, its main needs for the integration of resources, virtualization, automation and so on. However, traditional network management adopts the Fcaps network management model which is based on the equipment administration, and it is difficult to fuse the various administrative tools, which can not meet the flexible and changeable business model and management requirement of the data center. The new data Center management platform should adopt a service-oriented architecture (SOA) design idea, integration and unified management of resources, business, operations, such as the three major data center components, by on-demand assembly of functional components and corresponding hardware equipment, to form a direct customer application requirements of a series of integrated solutions, This provides support for a variety of key business systems in the data center.

Figure 1 Data Center management Solution Model

Figure 1 shows a general overview of the data center management solution, which consists mainly of four parts.

First, data center management needs to provide end-to-end, high-capacity, and visual infrastructure consolidation management scenarios.

Data center In addition to the traditional network, security equipment, there are storage, server and other equipment, which requires the common network management functions to redesign, including topology, Alarm, performance, panel, configuration, etc., to achieve the integration of infrastructure management. In the context of underlying protocols, traditional SNMP network management protocols and other management protocols, such as WMI and JMX, need to be consolidated to support the management of IP devices and it devices.

In software architecture, we need to consider the impact of tens of thousands of devices on the performance of management platform, so we must adopt distributed architecture design, so that the management platform can run on multiple physical servers at the same time to achieve management load sharing.

In addition, the data center in the room, rack, etc. also need to manage, these rely on traditional physical topology search is not out, need to consider adding new visual topology management functions, so that administrators can view such as partitions, floors, rooms, racks, equipment panels and other views, It is convenient for administrators to manage various resources in the data center from various dimensions.

Figure 2 Data Center visual topology view (room, rack, etc.)

Second, data center management needs to provide virtualization, automated management scenarios.

The traditional management software only considers the management of the physical equipment, and the virtual resources such as virtual machines and virtual network devices cannot be recognized, let alone configure these resources. However, data center virtualization and automation are the trend, virtual resource monitoring, deployment and migration needs, will promote the data center management platform for new changes.

For virtual resources, it is necessary to consider adding technical support to the topology, equipment, etc. to enable administrators to manage both physical and virtualized resources on the topology map, view the panels of virtual network devices, and the CPU, memory, and disk space of the virtual machine. Secondly, it is the ability to configure and manage various resources, the ability to distribute network configuration to physical and virtual devices, establish configuration baseline templates, automate regular backups, and support the migration and deployment of virtual network environments (VLANs, ACLs, QoS, etc.) to meet the needs of different scenarios such as rapid deployment, business migration, and new system testing.

Figure 3 Data Center Virtualization Resource Management

Again, data center management needs to provide business-oriented application management and traffic analysis solutions.

Data center has a variety of key business and applications, such as servers, operating systems, databases, Web services, middleware, mail and so on, the management of these business systems should follow the principle of high reliability, the use of agentless without monitoring agents to monitor, as far as possible does not affect the operation of the business system.

In visualization, in order to facilitate the integration of IP and IT management, network management and business management needs to be docking, the topology map can not only display device information, can also display the server menu running business and detailed performance parameters. In addition, the data center brings new business models, such as 1:N (a server running multiple services), N:1 (multiple servers running the same business) and n:m (traffic model between different services), these services for data center traffic has brought a great impact, there may be traffic bottlenecks, affecting business operations.

Therefore, such as flow analysis software can be improved to provide analysis based on Netflow/netstream/sflow and other traffic analysis technology, and through a variety of visual flow view, the business traffic in the interface, applications, host, session, IP Group, 7-tier applications, etc. are analyzed, To identify bottlenecks, programming interface bandwidth, to meet the user's internal business continuous monitoring and improved flow analysis needs.

Figure 4 Data Center traffic model

In addition, data center management also needs to provide controllable, auditable, measurable operation and maintenance management program.

The enterprise IT department that is responsible for running the data center often encounters the following issues:

The workload of IT department is difficult to measure and evaluate;

Fault handling has a greater randomness, it is difficult to find the responsible person and treatment methods;

The flow of technical personnel to increase IT management difficulty, only rely on experienced old managers, new people can not take over management;

IT department cost is not good control, the effect of input and output is not obvious.

Therefore, we must consider the introduction of operation and maintenance management, refer to the best practice--itil management model of IT service management, combing and curing the common fault handling process and configuration change process through the tools of User Service platform, Asset Library, knowledge base etc. Strengthen service responsiveness, summarize relevant experience in time, improve service delivery capability and service support capability of IT department.

Concluding

Cloud computing is the combination of IP technology and IT technology, so cloud management needs not only to protect the business and performance from the perspective of the underlying resources, but also to optimize the network from the perspective of business and performance. This means that the management of the cloud requires a new management model and a flexible functional architecture, and fully consider the infrastructure, technology trends, business operations, operation and maintenance services and other management elements, the establishment of a standardized, open, easy to expand, can be linked to the unified intelligent management platform to achieve resources, business, Operation Dimension Fusion linkage of the fine refinement management.

As cloud construction focuses on the transition from the data center to the operation of different types of cloud, such as the public cloud, the private cloud, and the mixed cloud, the corresponding management tasks also change from the management of the data center to the management of the cloud. From the present "see Cloud is not cloud", to the future "see Cloud or Cloud", this is a process. The best path to cloud management is to start with data center management, consolidate the underlying resources, and deploy through virtualization and automation, eventually transitioning to cloud services. As long as from the reality, in the practice of data center management constantly improve, it is natural to usher in a truly practical cloud management solutions.