Design and Implementation of cloud computing-based university data centers
At present, the application and development of Informatization Construction in military colleges puts forward new demands for further enhancing the capability of basic network platform Service Assurance teaching. Compared with the traditional IT architecture, "cloud computing" represents the development direction of the network service model in the future. Her unique advantages include high cost performance, easy scalability, high reliability, modularization, convenient management, it can better meet the information network construction needs of military colleges and universities, and is a good choice for the Information Network Application Service Model of military colleges and universities in the future. At the same time, we should see that the cloud computing application construction of military colleges and universities must be based on their own reality and accurately grasp their own needs to solve the actual problems encountered during the development and construction of education informatization in current colleges and universities, so that cloud computing applications can be "implemented" to achieve better benefits.
I. purpose of building a cloud computing data center
Solving the practical problems encountered by existing data centers in the service guarantee teaching process and breaking through the bottleneck of traditional computing centers is an urgent demand for the development of data centers. Traditional data center construction mainly adopts a distributed architecture, that is, each application uses an independent server, and Web, Mail, OA system, data backup, and other applications are mutually independent. With the development and growth of information construction scale, more and more equipment, such as business systems, servers, and storage, are available for data centers, and some problems are also emerging, mainly in four aspects: low hardware resource efficiency, high total cost of ownership, high space environment, energy consumption, and other O & M management costs, multiple business deployment processes, long launch cycle, and poor flexibility; poor Business Continuity assurance and weak data reliability.
1. Low hardware resource efficiency and high total cost of ownership
The evaluation results of the Use Efficiency of the main servers in our data center show that the efficiency of most servers is not high. Taking CPU usage as an example, a representative school portal website, a digital campus Portal, a subject website, and other servers are selected for evaluation. sampling is performed every 5 minutes during working hours, data is collected for five consecutive working days. The average CPU resource consumption is 3.739%, the average CPU usage is 9.916%, and the lowest is 0.339%.
According to the statistical analysis results, under the existing conditions, the resource utilization rate of a single server is low, usually below 10%, and the chimney development of each application system fails to achieve effective resource utilization.
2. High O & M management costs such as space environment and energy consumption
The School Information Management Center is monitored through the dynamic environment monitoring system of the data center. Currently, the power consumption of more than 80 IT devices, such as servers and storage, in the data center is about 20 KW, and the peak value is close to 26 KW, the average daily power consumption is about 480 degrees. Currently, most of the central data centers are deployed on traditional 2 to 4 inch rack-mounted servers. The overall heat dissipation is large. In summer, three 30 KW precision air conditioners can work simultaneously to meet the needs. In addition to UPS, fresh air systems, and human resource consumption, the overall overhead of data center maintenance is relatively large. Currently, the central data center has 35 departments and 35 managed devices, accounting for about 43% of all devices in the data center, and the number of managed devices is constantly expanding. With the increasing number of equipment, the IDC Room Conditions and O & M personnel are all put forward with higher requirements. energy conservation, environmental protection, and easy management will be a concern for the development of data centers in the next phase, traditional construction and O & M models must be changed.
3. There are many business deployment procedures, long launch cycles, and poor flexibility
In the traditional data center construction mode, one device is required for each added application according to the "Application Isolation" principle, and the service application has a long online period, this method is gradually unable to meet the rapidly growing demands of colleges and universities for information application. The main reason is that in recent years, with the development of educational information applications, many application system support tasks with short cycle and compact working pace have emerged. For example, the "2110 project" project application system, the deployment of the online examination system for Tianjin National Defense students' political theory basic examination undertaken by our institute, and the selection and awards of excellent military training digital media resources. It usually takes weeks or less than a month to prepare for task release and system deployment. In the current mode, it is very time-consuming to apply for funds, purchase bidding according to regulations, and install and debug the system. In addition, some systems, such as the Basic Political Theory Examination System of National Defense students, have high requirements on server configuration, but the system is only used for two days. After the assurance task is completed, most of the time in a year is idle. It is difficult to cope with the existing data center environment.
4. Poor business application continuity and weak data reliability
Traditional data center backup methods have the following defects: long recovery time, high failure risk, hard device protection, and long data loss time. According to the investigation by relevant organizations, 47% of all kinds of disasters that information systems may encounter are hardware-level faults, such as motherboard, RAID Controller, Nic, and built-in/external storage, up to 53% of the disasters are caused by "soft" errors, such as accidental data deletion, hacker/virus attacks, system crashes, and unstable patches. Tape backup, disk-based backup, Cluster/HA, and other traditional data protection mechanisms can only provide limited protection for the above disasters. Compared with OA, MAIL, and other systems, the organizational schedule of online teaching applications is fixed and the length of time is limited. The course schedule has certain requirements for business application continuity and data recovery time in case of faults. Poor business application continuity and weak data reliability limit the overall development of network teaching to a certain extent. Currently, it is mainly used for self-study. To promote the further development of network teaching applications, it is necessary to comprehensively cover all the resource usage, scheduling and backup mechanisms for "soft" errors and "hard" faults.
Solving the above four problems is the primary task of the next cloud computing platform construction of the college. On this basis, the network distribution service achieves the pooling of self-service and resources, enabling flexible resource scheduling, to better ensure the different needs of the College's informatization construction.
Ii. Cloud Computing Platform Construction Scheme Design
1. Cloud computing platform physical resource construction
The high-end blade server is used as the computing Basis of the cloud computing platform. The blade server features higher computing density, lower computer room requirements, low power dissipation, simple wiring, high reliability, convenient management, and flexible scalability. It is more in line with the management service concepts of virtualization and cloud computing.
Consider the needs of upgrading some old servers and predict the computing resource scale of the cloud computing platform based on the needs of new applications in the future. Generally, the CPU clock speed is 2 GHz, and the memory is 4 GB. For special applications, the CPU clock speed is 8 GHz, and the memory is 64 GB. The computing capability of 20% peak bursts is reserved. It is estimated that the computing capacity needs will double in the next five years. We plan to adopt a group of 10 high-performance blade servers as the core of the cloud computing hardware computing platform. A single blade server is configured with 2 CPUs and 8 cores, with a clock speed of more than GHz per core, the memory is 256 GB. Each server is configured with a dual 10-Gigabit Nic and dual 8GHBA card. All components in the blade chassis must be redundant, and all modules can achieve centralized remote management configuration. It is estimated that more than 100 virtual applications will be supported after completion.
Use virtualization technology to build a centralized storage system. Storage resources are divided by service objects into two parts: one is the high-speed storage service for business application systems, this part of storage should be able to meet high performance requirements such as operating systems and databases for deployment. Sharing with all blade servers is the basis for redundant hot backup, load balancing, and virtual machine migration. Based on 2 TB computing per blade, the estimated high-speed storage capacity is 20 TB. In addition, based on Virtualization, the storage space is expanded to provide resource storage services for business applications and common users. With 10 Gb of actual available network storage space per user, the storage space is expected to be 50 TB in the future to meet the computing requirements of 5000 users. 30% redundant extensions are reserved, and the expected centralized storage capacity is about 100 TB. In order to work With virtualization software to achieve the best application results, an optical fiber storage array product with an interface type of 8 GFC interface is configured to form a standard SAN centralized storage architecture. With the shared SAN Storage Architecture, You can maximize the advantages of cloud computing solutions, achieve online migration of running virtual machines, dynamic resource management, snapshot-based backup, and so on, it provides excellent storage resources for cloud computing platforms.
2. virtualization and cloud service software resource construction
In the Infrastructure Section, VMwareVSphere5 is used for virtualization and integration of some existing physical servers through virtualization. The vMotion function enables online migration of running virtual machines and applications to different physical servers. It allocates resource pools for application resources in the entire data center, ensure that the computing and I/O resources of each application running on the platform, especially the core applications, are effectively guaranteed. The DRS function enables Dynamic Load Balancing and automatic online management of resource pools, and elastic expansion of the data center infrastructure; The DRSHA feature enables high-availability environments; and vStorage enables integrated backup of centralized virtual machines.
The vCenterServer enables centralized and unified management and monitoring of the Virtual Machine cluster environment.
The cloud computing management platform vmwarevclouddireprovides the infrastructure as a service to school users. The cloud computing management platform portal provides a unified interface for allocating and managing virtual machines to implement interactive management functions, allowing departments or branches to log on to create, manage, configure, and use virtual servers and applications on their own, monitors the usage of resources in a virtual data center. Management Personnel of information applications and website construction units can use self-help portals to access their own virtual data centers and freely organize their own virtual machines and application resources. Authorized departments can easily access resources through policy-based resource allocation and deployment. Each user organization has its own virtual resources, an independent LDAP authentication mechanism, a specific policy control mechanism, and a unique directory.
The information management center uses the cloud computing management platform to add virtual machines on the cloud computing platform without downtime, including virtual CPU, virtual memory, virtual network card, and virtual storage capacity, dynamic Expansion of virtual servers without service interruption. It can manage multiple virtualized data centers through the cloud computing management platform, and manage different levels of resource pools based on performance and reliability to meet the requirements of different levels of application servers.
Iii. Practical Application Results
Through practical tests, the cloud computing platform is an effective solution to a number of problems encountered during the construction of the network data center of the College and the guarantee of teaching services. Combined with the upgrading and replacing of old servers, the existing application systems are integrated and optimized during the construction of the cloud computing platform. In addition to performance improvement, the device utilization rate is improved, and the resource loads of multiple applications are combined on one server to reduce the number of devices and greatly reduce energy consumption, reduces management and O & M costs, implements ECs risk management, and enhances reliability and availability.
1. Greatly improve application management of data centers
We have implemented data migration for some information application system resources and adjusted and deployed them to the cloud computing platform, including the stable operation of various application systems of the new network teaching platform. In the cloud computing platform, the average CPU usage rate of a single server is increased to about 23%. As the platform's applications increase and the application scope expands, this rate will be further increased, it is much higher than the average CPU usage of 3.7% in the old mode. The efficiency of hardware resources has been significantly improved.
Virtualization technology is used to standardize computing platform resources and build a unified architecture. By establishing resource pools, computing resources are dynamically allocated as needed and managed in a unified manner, it implements Application Deployment, permission management, upgrade and maintenance, performance analysis, diagnosis and warning, data backup/recovery, data synchronization and Disaster Tolerance. This makes the entire data center operate and maintain professionally and with standardized procedures. Automatic data migration is implemented when a hardware fault occurs, which ensures the continuous and stable operation of network teaching applications at a low cost and is not affected by a single hardware fault.
Analyze the test run results to support application systems of the same scale and use existing cloud computing platform service assurance methods, compared with the traditional data center operation mode, the maintenance cost is reduced by more than 50%, the energy consumption is reduced by more than 40%, and the space is saved by more than 60%.
2. Promote the informatization level of education to meet personalized information application needs
Through the establishment of the cloud platform of the Information Management Center, it provides teachers and students with a better information-based teaching environment, enhances the ability to serve teaching, guarantee teaching, and assist teaching, and meets the increasing needs of information-based teaching.
First, through targeted settings, flexible provision of a variety of information environments, to better ensure the personalized needs of teaching resources. In this way, users can apply for virtual servers and storage space on their own online and on-demand basis, breaking through the traditional network service platform. due to limited resources and technical conditions, users can only provide a unified technical framework, limits on the personalized development of users.
Second, it effectively allocates resources rationally, improves the efficiency of using basic hardware resources, highlights the construction of high-quality information resources, and ensures key teaching applications. Cloud computing uses the network distribution service to achieve self-service and resource pooling, as well as real-time monitoring of server CPU and memory resource consumption, and dynamic resource allocation. By rationally allocating resources, you can improve the efficiency of using basic hardware resources, highlight high-quality products, and ensure the key points. This eliminates the fixed restrictions on the configuration of hardware and software resources on traditional IT network service platforms.
Third, it is easy to expand, shorten the time for deploying and using new application systems, flexibly cope with all kinds of emergency protection tasks, and better adapt to the rapid development of education informatization construction in military colleges, and ensure the fast pace of teaching. Using the cloud computing service platform, the activities such as "Basic Political Theory Examination for National Defense students in Tianjin" and "network teaching monitoring system testing" were efficiently guaranteed, it achieves network computing resource emergency scheduling, rapid system deployment, and resource recycling, and achieves satisfactory practical application results.
Iv. Summary
The infrastructure cloud computing platform effectively solves the main problems encountered by the data center in our hospital and greatly improves the application management of the data center. IT is the future development trend of IT resource service models and applications, with the further development and application of cloud computing technology, it will greatly promote the improvement and development of education informatization level.