Shulou(Shulou.com)06/03 Report--

With the development of new IT technologies, such as virtualization, cloud computing and big data are no longer unfamiliar to everyone. When the cloud implementation is carried out in the enterprise, while the advantages of all kinds of new technologies appear, there are many problems that follow. After getting on the cloud, sometimes we don't feel relaxed but more tired. So what's the problem? Has our IT become simple or complex? Is our thinking still confined to the traditional IT mode? The following is to share our thinking and practice in the process of cloud.

I. stage analysis of enterprise cloud evolution

Cloud computing has been developed for more than 10 years since its emergence in 2006, during which hardware resource virtualization and its management technology flourished and promoted people's concept and understanding of cloud: KVM entered the Linux kernel in 2007; LXC released its first version in 2008; VMware launched VSphere;HyperV in 2009 and its driver was submitted to the Linux kernel in 2009; and CloudStack opened source in 2010. OpenStack, which later became the most important infrastructure for the private cloud, was launched by NASA and Rackspace in 2010. A large number of service models have emerged, and a large number of XaaS concepts have emerged. Many important cloud computing technologies are released in open source mode, open source has gradually become an important choice of cloud infrastructure, and the concept of cloud computing is becoming clear.

Since 2012, domestic cloud computing has developed rapidly, and private cloud has begun to rise in enterprises. Open source Openstack has been widely used and promoted at the right time. Open source technology ecology has become an important force in the technology development of the industry, and cloud computing has entered a new period of prosperity and development.

Looking at the evolution of enterprise cloud, it has been four years since we started to build our own private cloud in 2013. we sum up the past four years in six words, that is, from decentralization to centralization and then to integration. Before the cloud, each business system is scattered in various departments, by different departments to plan, manage and operate. With the cloud, the enterprise is forcing all business systems to migrate to the cloud, this time is to achieve a simple centralization. After centralization, another road that must be taken is integration, because after all your systems are on the cloud, how can you make each business system run better and achieve the best efficiency in the cloud environment? then you need to carry out unified scheduling and unified management, which must be faced with an integration. Including the integration of your computer room, the integration of IT equipment. After the integration of these infrastructure, we set up our own cloud platform on it, and further manage all kinds of devices and resource pools. in addition, we need to fully integrate some resource billing systems, process systems and other visual operation and maintenance within the enterprise, including some data acquisition and analysis systems. This is called the integration of our CT capabilities. Go further is the business platform integration, unified portal, virtual desktop, OA, ERP, CRM and HR systems should be integrated on the entire cloud platform to achieve integration and integration.

The process from decentralization to centralization to integration is called cloud 1.0. Reviewing this stage, we have summed up four points: 1. Unified planning, we have achieved unified planning for software and hardware resources; 2. Distribution according to demand. We use virtualization technology, through the cloud platform intelligent scheduling and balancing algorithm to achieve the allocation of resources on demand. 3. The interconnection of resources, which is a very important point. All business systems are connected with each other in the cloud. After the interconnection of resources is realized, it lays a good foundation for enterprises to realize the integration of big data and applications in the future. 4. Unified management is of great value. Achieve unified management, unified monitoring, unified scheduling, unified process and event management, and finally unified services. The unified management of data on the cloud provides a basis for enterprise data backup and disaster recovery, and is more suitable for our management and expansion requirements. This is also the main achievement of Yunhua 1.0.

Second, the problems and thoughts faced by enterprises in the stage of cloud 2.0.

What will the enterprise achieve next in the future cloud 2.0? In a nutshell, it is "green operation-data intelligence-open architecture". Why mention green operation? As we all know, the annual operation and maintenance cost of the IT infrastructure in the data center is also very high. Let's not say anything else, but its electricity is a very big consumption. The average annual cost of refrigeration and electricity for a server is about 10, 000 yuan. If you have 5, 000 units, that's 50 million. What a big expense it is. In the next few years, driven by cost pressure, regulatory compliance and corporate social responsibility, corporate IT is bound to move towards the ranks of energy conservation and emission reduction. And with the general trend of decarbonization, it will move towards clean energy and solar energy in the future.

Let's look at the power consumption statistics of a data center and the corresponding PUE. In July and August, in the hot season, we can see that its PUE value is relatively high, all about 1.76, and decreases in September and October, and 1.35 to 1.33 when it is colder in November and December. We can see that the average daily electricity consumption is more than 10,000 kilowatt-hours. This is a great expense for the enterprise.

Industry explorations include Microsoft's undersea data center, Facebook's Arctic Circle data center, and Ali's water-cooled data center in Qiandao Lake. Recently, we have seen that Ali also has a cluster of immersed liquid-cooled servers, and the PUE value can approach the theoretical limit of 1.0. These are very good explorations, and if they can be successfully realized, they will be of great value to the energy conservation of the data center.

The second is data intelligence, which means to process, analyze and mine massive data through large-scale machine learning and deep learning technology based on big data engine, and extract valuable information and knowledge contained in the data. make the data "intelligent", and through the establishment of models to find solutions to existing problems and achieve prediction and so on. Generally speaking, the initial enterprises positioned IT to build a simple automation system, but with the development of business and the gradual improvement of their position in the market, enterprises gradually began to need some large systems, such as ERP, MRP, CRM and so on. Some enterprises have built an integrated information work platform, and the standardization, informationization, integration and intensification of enterprise management have been promoted by leaps and bounds. At the same time, with the increasing maturity and in-depth application of information systems, enterprise data show the characteristics of massive growth and dynamic changes, and users put forward higher requirements for in-depth mining and real-time reflection of massive data. through big data and deep learning and other new technologies to better mine and extract valuable information and knowledge to form better problem solutions and achieve prediction. This aspect provides strong technical support for the needs of users. Through the intelligent analysis of data, we can also better make our business system run on our cloud platform.

Another point is that it is open and compatible, there are a lot of good things in open source, and the technology is developing very fast. The traditional business software used by enterprises in the past has made slow progress and is relatively closed. The enterprise cloud platform adopts an open architecture, which can be compatible with common devices in the industry and mainstream operating systems, virtualization software and applications, thus greatly reducing the development, operation and maintenance costs of the cloud platform. Now many traditional enterprises actively embrace open source and move towards open source, which is a good thing. Only by being open and compatible can we absorb more and better things.

For enterprises, what kind of functions your cloud platform should need, we have summed up a principle after years of exploration and practice: the enterprise's business needs and specific use scenarios are the key to determine what kind of functions a cloud platform should have. Many times you find that when talking about their products, many manufacturers always emphasize how awesome and powerful their products are, but they do not understand the specific business and actual needs of the enterprise. because it does not conduct research on the actual needs and business of the relevant enterprises, so what you produce is a high-end product, but the result is not suitable for the enterprise. We all know that there is an idiom called cutting feet to fit shoes. Can your products allow enterprises to cut feet to fit shoes? It must be impossible. You must be tailored to the foot of the enterprise to make your products, so that your products can best match the needs of the enterprise, can be truly valuable.

Clouding is a continuous process, and we can't achieve it overnight. In the cloud order of business, we generally follow the order of periphery before core, new before old, small before big. The key points of cloud are summarized as follows: 1, the system can scale flexibly; 2, application cluster deployment; 3, application and data separation; 4, data distributed deployment; 5, data platform. When going to the cloud, how to transform the business is mainly to sort out the applications, classify different applications, and take the corresponding cluster support at the bottom, such as computing-intensive, IO-intensive, etc., while comprehensively considering the peaks and troughs and business characteristics to configure.

Building a private cloud of an enterprise is actually a complex proposition. Four years after the private cloud began to be built, we look back and find that there are still some places that are not well done and the solution is not complete and straightforward. Share a few thoughts: 1, architecture. In the process of building a private cloud, different products will be used, including commercial and open source, but an elegant and reasonable private cloud architecture is not a simple accumulation of basic equipment, virtualization, disaster recovery, cloud services and other products. You need to combine business and applications to constantly evolve the architecture to find the most suitable one. Otherwise, many enterprises will face the situation that they can not be removed, changed or added, which will eventually lead to the abandonment and reconstruction of the platform. 2. Management. It is mainly about the thinking about CMDB. CMDB is very important in large-scale private cloud construction. Many times you need to have a reasonable location and evaluation of resources. At this time, like the CMDB done in many enterprises, it is very static, and it is troublesome for you to update it, many of which are manual. We hope to achieve an automatic collection and dynamic update, including your bare metal, virtual machine and container, resource scheduling and dynamic change, which can be well integrated into the cloud platform. 3. Operation and maintenance point of view. It is mainly automation and standardization. In the construction of private cloud, it is necessary to build a powerful automated standard operation and maintenance platform for cloud operation and maintenance, so that all data of operation and maintenance can be circulated and analyzed on this platform, so as to provide the best data support for operation and maintenance personnel to be efficient and move on demand. 4. Development point of view. Mainly because of your corporate culture and process reengineering, we need to give developers more freedom. For example, Docker gives you more new ideas here, so that technicians can slowly integrate into the technology culture of devops through the construction of cloud. 5. Safety angle. Information security and self-control is a very important strong demand for traditional enterprises, there are many typical events will no longer be said, people who often pay attention to technology will understand. In recent years, VPC network isolation, hearthstone legends, as well as AWS outages, April Fool's Day Beiyou computer room fire, have all taught security lessons. We need to strengthen the response to security risks in the cloud environment, and the construction of private clouds in large enterprises also needs comprehensive information security solutions and services.

The summary of enterprise cloud is simply summed up in one sentence: an architecture or a cloud product cannot solve all problems. Each product has its own applicable scenario, and there is no best but the most suitable one. Building a perfect enterprise cloud platform requires multi-dimensional and multi-angle consideration, which can not be solved overnight.

Third, the future prospect of enterprise cloud.

With the deepening of enterprise cloud, we can see: 1. Infrastructure cloud computing has entered a critical stage of business application. The high availability of cloud computing, the integration of container technology and cloud network end is the direction of development in the future. 2, the digitization and intelligence of traditional enterprises rise to the national level. At some important meetings, state leaders have repeatedly stressed the need to speed up the construction of information technology, promote the deep integration of the Internet and the real economy, and speed up the intelligence and digitization of traditional enterprises. The anxiety and embarrassment of enterprises in this era undoubtedly comes from the various business and technological revolutions we are experiencing. The past is no longer established and the future is not clear. What's worse than being conquered is that you don't know who you're being conquered.

The recently released three-year Action Plan for the Development of Cloud Computing (2017-2019) fully affirms the basic position of cloud computing in the development of emerging fields such as big data, Internet of things and artificial intelligence. it is considered that cloud computing is the key element to promote the integration of manufacturing and Internet, is an important driving force to promote the strategy of manufacturing power and network power, and launched a series of supporting policies and measures. It provides a strong guarantee for the rapid development of cloud computing. Therefore, there is no time to wait, enterprises need to strengthen internal and external cooperation, and constantly improve the ability and level of private cloud construction.

Building a private cloud is a long-term iterative process, and we are all on our way. Look to the distance, look to the future, the future has come, foresee can meet, realize can youdao!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.