"Cloud," changing it

law is everywhere, and in the IT world, we advocate such a law: in it, every change in the application will lead to a new round of management innovation. Obviously, in the current trend of cloud computing, it is also changing the management model of it.





the trend of virtualization leads to the visualization of resources and the integration of management





before the advent of cloud computing, the idea of integrated management already existed, but cloud computing made integration management inevitable.





a few years ago, IT administrators found that as it is becoming more and more mature, its host business system is increasingly complex, in this case, single, functional-oriented management system because there is no uniform standards, no agreement of cooperation, in the enterprise IT facilities change or user core business adjustment, It is often difficult to play its role normally. As a result, the integration of IT infrastructure management is becoming more and more important, on this basis, IT administrators can not only on a platform to understand the status and utilization of various IT resources, but also to the provision of resources on demand.





in the cloud-era data center, virtualization technology is heavily used and management objects are changing. The previous equipment is real, location is also fixed, now become virtual, can be flexible migration. The visibility of the resource's physical location in the datacenter becomes difficult.





in the network equipment domain application N:1 Virtualization technology, will the N network equipment virtual into 1 network devices, for example: H3C IRF Intelligent Elasticity Architecture Technology, Cisco's VPC, the VSS technology, requests the management system not only to be able "from the multiplication into the simplicity", the shielding interior detail, IRF as a whole to show the external, simplified management, but also to be able to "by simple", once the IRF of a physical member node failure or internal connectivity problems can be accurately positioned.





while in the server domain, there is a large application of another virtualization technology, 1:N virtualization to improve data center reliability, reduce data center energy consumption and hardware costs. Virtualization software providers VMware, Microsoft, Citrix, Redhat, and so on to launch their own server virtualization solutions, which bring about the management of virtual machines, server performance and capacity management more and more important.





In addition, the application of virtualization technology, network and server boundaries become blurred, but also caused a new problem, namely the network and computing resources of the collaborative scheduling problem. When you create a virtual machine or virtual machine migration, the VM host can run normally, not only need the resources on the server reasonable dispatch, network connection reasonable dispatch is also necessary. The key to realize the data center based on the business scheduling and finally realize the automation is to realize the gap between the network and the computation, and achieve the resource fusion management and intelligent Dispatch.





second, the "business" as the core resource scheduling





in terms of network resources control and scheduling, ACL, VLAN and QoS management technology has become increasingly mature, but when the administrator faced by different manufacturers, different performance, different operating system of network equipment composed of data centers, it will be found to achieve the business based resource scheduling is a difficult proposition, More new ideas need to be introduced.





in the era of cloud computing, from the end-user point of view, the data center is like a dedicated network, but for the administrator perspective, a host of multiple services/multi-user cloud data center, which corresponds to each business/user is a network slice (note: Slice, extended from medical terminology, A thin slice of tissue or mineral of an organism, which divides the network into logical units of different business services according to demand. The management of the data center network needs to be managed from every slice of the network, including business deployment, monitoring and evaluation, and the management from the perspective of Network Service.





from the deployment of the business, data center network by a large number of devices connected, these equipment models, different capabilities, to achieve network slicing, you need to use the concept of business choreography (orchestration). First, the device capability set is categorized and abstracted into a service model, the actual function of the device is the instantiation of the service model, taking the common server two-layer link as an example (shown in table 1), for server bandwidth, ACL, q0s configuration, some devices support the command line control, some through the MIB control, Even with command-line management, there are differences in command-line formatting. For the users of the service, should not feel the difference, can be considered an abstract service. However, in actual deployment, you need to translate the service policy into specific vendors, the actual capabilities of the device.





　　


Secondly, based on the equipment capability, the equipment function is abstracted into the common network Management Service, and the Open Service management interface is finally provided. From the automated management process for a terminal enterprise/user to the business services provided by the data center, need to provide end-to-end deployment of network slicing access server from user side, such as access side, need to provide access control, user information verification services, data center side need to provide bandwidth, QoS and other network services, ACL Firewall, LB and other network services are required in the server access area. According to the requirements of the network, the upper-level operation system can arrange the network management service and other business services based on the business process to the upper level business process, so as to realize the automation management of the business and network integration.











, opening to the outside, providing support for automation operation





Cloud Computing The ultimate goal is to achieve the system on-demand operations, operating systems can be based on user requests to perform the service opened. The opening of any service requires calculation, storage, scheduling and distribution of network resources, which puts forward new requirements for the openness of management system.





's new demand for openness is mainly embodied in two aspects, the first is the type of interface. Traditional management system and upper system docking, pay attention to the failure of the report and information query, and the cloud era of management system more attention to how to implement automated deployment, in the interface more attention to resource scheduling and allocation, need management system in the business scheduling to achieve openness.





on the other hand, interfaces are provided in such a way that interfaces must meet the requirements of standardization and generalization. So interfaces should not bind to a particular language and should be supported by an open standard system. The Web service standards based on the SOA architecture can support a service based architecture that does not bind to specific computer languages, and is the first choice for cloud management open interfaces such as cloud resources, cloud scheduling, cloud provisioning, and cloud operations.





Four, load network resource dispatching and access security can not be ignored





Cloud Computing system, the traffic model of the bearer network presents two major changes. First, with the construction of the data center, data to the Cloud collection, the original horizontal flow gradually reduced, and from the terminal to the data center of the longitudinal flow increased greatly; second, with the multimedia technology and terminal intelligent, a large number of video services to carry the network bandwidth demand dramatically.





To cope with the change, the higher performance of the load network network equipment is important, but how based on the existing hosting network, through performance management and resource scheduling to better provide support for cloud computing, it is necessary to establish a set from analysis to execution, and then audit monitoring of closed-loop performance management scheme.





for terminal access, in addition to the traditional PC, intelligent terminals have led to the rapid expansion of terminal management category. While providing a large number of terminal access capabilities, it is also necessary to consider the reliable, secure and controllable access terminals.





v. Concluding remarks





the traditional IT management model, administrators focus on resource management, focusing on how to protect business and performance from underlying resources. In the cloud mode, the management is more focused on the automatic provisioning and scheduling of resources, as well as the service depth assurance based on network services, that is, how to adjust and optimize the resource supply scheme according to the change of business performance requirements.




The data center management in the
cloud mode is more important in the data center network, and it also puts forward new challenges to the management software, automation, process, openness, integration, intelligence become the necessary requirements of the next Generation data center management.