Virtualization, cloud computing, open source code, and more

A, Virtualization

Virtualization refers to the ability to simulate multiple virtual machines on the same physical machine. Each virtual machine has a separate processor, memory, hard disk, and network interface logically. The use of virtualization technology can improve the utilization of hardware resources, so that multiple applications can run on the same physical machine with each other isolated operating environment.

There are also different levels of virtualization, such as virtualization at the hardware level and virtualization at the software level. Hardware virtualization refers to the simulation of hardware to obtain a similar to the real computer environment, you can run a complete operating system. At the level of hardware virtualization, there is full virtualization (all virtualization, almost complete simulation of a real hardware device.) Most operating systems run directly in a fully virtualized environment without any modification. , Partial Virtualization (partial virtualization, only provides simulations of critical computing components or instruction sets.) The operating system may need to make some modifications to run in a partial virtualized environment. and paravirtualization (semi-virtualization, does not simulate hardware devices, virtual machines have an independent operating environment, through virtual machine management program to share the underlying hardware resources. Most operating systems need to be modified to run in a semi virtualized environment. ) and other implementation methods. Software level virtualization, often refers to the same operating system instance on the basis of the provision of multiple isolated virtual operating environment, but also often referred to as container technology.

At the level of hardware virtualization, modern virtualization technology is often a hybrid of full virtualization and semi virtualization. Common virtualization technologies such as VMware, Xen, and KVM provide support for both full virtualization and semi virtualization. Virtual machines provided in the form of hardware virtualization typically run a complete operating system, with a large number of identical or similar processes and memory pages on the same host, leading to significant performance losses. At present, the memory pages containing the same content can be identified and merged by KSM Technology, but there is no effective means to optimize the processing of the same or similar processes. As a result, hardware virtualization is often called heavyweight virtualization, and the number of virtual machines that can run concurrently on the same host computer is quite limited. At the level of software virtualization, all virtual machines on the same host share the same operating system instance, and there is no performance loss due to running multiple operating system instances. As a result, software virtualization is often referred to as lightweight virtualization, and the number of virtual running environments that can run concurrently on the same host is relatively loose. Taking container on the Solaris operating system as an example, an instance of a Solaris operating system can theoretically support up to 8,000 container (the number of container that can actually be run depends on system resources and load). Similarly, LXC on a Linux operating system can easily support a significant number of virtual running environments on the same host.

In the field of virtualization, domestic companies have a large interest in hardware virtualization, and in the research and production environment, the hardware virtualization technology is mostly adopted. Taobao is the early domestic research and application of software virtualization technology, their experience in the main Taobao site shows that using Cgroup to replace Xen can improve resource utilization. As to whether hardware virtualization or software virtualization should be selected in a real-world scenario, focus should be given to the end user's need to have full control over the operating system (for example, upgrading the kernel version). Software virtualization can be more cost-effective if end users only need control over the running environment (such as the various app engine services at the PAAs level). Software virtualization is also a good choice for applications that provide horizontal scalability for the same application.

For technicians who need insight into virtualization technology, VMware's White paper "Understanding full virtualization, paravirtualization, and hardware Assist" is a good reference.

Generally speaking, the number of users who can directly use virtualization technology is relatively small. In the case of Linux operating systems, users who are able to perform virtual machine lifecycle management are generally those who have access to libvirt permissions. In a company or other entity, these users are usually system administrators.

B, Virtualization Management

Early virtualization technology solves the problem of providing multiple independent operating environments on the same physical computer. When the number of physical machines that need to be managed is less than an hour, system administrators can manually log on to different physical machines for virtual Machine Lifecycle management (resource configuration, startup, shutdown, and so on). When there is a large number of physical machines to manage, it is necessary to write scripts/programs to improve the automation of virtual machine lifecycle management. Software that manages and dispatches a large number of physical/virtual computing resources, called Virtualization management tools. The Virtualization management tool enables system administrators to perform the following tasks from the same location: (1) The life cycle management of virtual machines on different physical machines; (2) querying and even monitoring all physical machines and virtual machines; (3) establishing a direct mapping relationship between virtual machine naming and virtual machine instances, makes it easier to identify and manage virtual machines. The Virtmanager on the Linux operating system is a simple virtualization management tool. VMware vsphere is a powerful virtualization management tool in the VMware product family.

Virtualization management tools are a natural extension of virtualization technology. A simple virtualization management tool solves the problem of the complexity of the work content caused by the increasing number of physical machines. At this level, virtualization management is often accompanied by the concept of clustering. A virtualization management tool that often requires virtual machine lifecycle management rights on each physical machine (for example, a username and password with access to Libvirt permissions). In the same cluster, for convenience, it may be necessary to set up a common management user across the cluster level. Virtualization management can be considered to be a convenience for system administrators, but it does not delegate the permissions of virtual machine lifecycle management to other users.

C, Data center virtualization

At the data center level, system administrators need to face a large number of different types of hardware and applications. Compared with small clusters, the system complexity of data center is greatly improved. At this time, simple virtualization management tools can not meet the requirements of the system administrator, so the virtualization management tools on the basis of the development of a variety of data center virtualization management system. At the hardware level, the data Center virtualization Management System organizes hardware resources in a way that divides resource pools, which are usually a cluster, and exposes computing resources to users in the form of virtual infrastructure. At the software level, the data Center Virtualization management system introduces two different roles of system administrator and ordinary user, even based on the need of the application scenario to set finer granularity role-based permission control (role Based access Control,rbac). System administrators have administrative rights over the physical and virtual machines of the entire data center, but generally do not interfere with normal virtual machines. Ordinary users can only perform virtual machine lifecycle management operations within their own pool of resources, and do not have permission to control the physical machine. In extreme cases, ordinary users can only see the resource pool assigned to them, without knowing the details of the physical machine that makes up the resource pool.

Before Data Center virtualization, the act of creating a virtual machine requires a system administrator to do so. In the Data Center virtualization management System, the authority of virtual machine lifecycle management is delegated to the so-called "ordinary user" through role-based privilege control, which can reduce the burden of the system administrator to some extent. However, for system security reasons, not all employees in the company can have such "ordinary user" accounts. In general, this "ordinary account" can only be assigned to the person responsible for a particular team. It can be argued that the lifecycle of a virtual machine is centrally managed until the level of data center virtualization.

The Data center virtualization management System is a further extension of the Virtualization management tool, which solves the problem of system complexity caused by the increase of hardware and application scale. After a specific physical device is abstracted into a resource pool, the company's executives only need to understand the size, load, and health of each resource pool, and end users need only understand the size, load, and health of the resource pool assigned to them. Only the system administrator needs to be familiar with the configuration, load, and faults of each physical device, but the concept of a resource pool also logically collates and categorizes all physical devices, making it easier for system administrators to work.

Modern data center virtualization management system, often provides a large number of functions to facilitate operation and maintenance automation. These features include (1) rapid deployment of a series of identical or similar operating environments based on templates, (2) Monitoring, reporting, early warning, accounting functions, and (3) high availability, dynamic load balancing, backup and recovery, and more. Some relatively open data center virtualization management systems, even in the form of open APIs, enable system administrators to develop additional extensions based on their own scenarios and processes.

In the VMware product family, VMware vcenter is a data center virtualization management software. Other recommended data center virtualization management software includes Convirt, XenServer, Oracle VMs, OPENQRM, and more.

D, Cloud Computing

Cloud computing is a further encapsulation of datacenter virtualization. In cloud management software, there are two (or more) different roles and different permissions for the cloud administrator and the average user. Administrators have administrative rights over the physical and virtual machines of the entire data center, but generally do not interfere with normal virtual machines. Ordinary users can self-service virtual machine lifecycle management through the browser, or they can write programs to automate virtual machine lifecycle management through Web service.

At the cloud level, the power of virtual machine lifecycle management is completely decentralized to real-life users, but it also masks the concepts of resource pools and physical machines, and so on, from the perspective of ordinary users. Ordinary users can get computing resources, but there is no need to have any knowledge of the physical resources behind them. On the surface, cloud computing seems to provide computing resources in a pattern compatible with Amazon EC2/S3. In essence, cloud computing is a pattern of computing resource management that has changed, and end users no longer need the help of a system administrator to get and manage computing resources themselves.

For cloud administrators, delegating virtual machine lifecycle management rights to end users does not reduce their work pressure. Instead, he has a more troubling thing to deal with. In traditional it architectures, it is often an application with a set of computing resources, the application of physical isolation, problem diagnosis is relatively easy. After upgrading to cloud computing mode, multiple applications may share the same set of computing resources, there is resource competition between applications, and problem diagnosis is relatively difficult. Therefore, cloud administrators often want to choose the cloud management software to have a relatively comprehensive data center virtualization management capabilities. Critical features for cloud administrators include (1) monitoring, reporting, alerting, accounting functions, (2) High availability, dynamic load balancing, backup and Recovery, and (3) dynamic migration, which can be used for local load tuning and troubleshooting.

Clearly, from virtualization to cloud computing, the packaging of physical resources continues to improve, the virtual machine Lifecycle Management Authority gradually decentralized.

In the VMware product family, VMware Vcloud is a cloud management software. Other recommended cloud management software include OpenStack, Opennebula, Eucalyptus and Cloudstack. Although OpenStack, Opennebula, Eucalyptus, and Cloudstack are cloud computing management software, their functions vary greatly, because different software has different design concepts. The original design goal of Opennebula and Cloudstack is the data center virtualization management software, therefore has the more comprehensive data center virtualization management function. After the concept of cloud computing arose, Opennebula increased the Occi and EC2 interfaces, and Cloudstack provided additional components called Cloudbridge (Cloudstack included the Cloudbridge component by default from version 4.0). This enables compatibility with Amazon EC2. Eucalyptus and OpenStack The Amazon EC2 as the Top-down design cloud management software, from the outset to consider the compatibility with Amazon EC2 (OpenStack also increased its own extension), However, there is a lack of functionality in data center virtualization management. In both cases, the Eucalyptus project is significantly stronger than the OpenStack project in the data center virtualization management because of its early start.

E, private cloud and public cloud

Cloud computing, as described in D, is just a narrow-sense cloud computing or cloud computing similar to Amazon EC2. Cloud computing in its broadest sense refers to a variety of practices that access physical/virtual computers over the network and utilize their computing resources, including cloud computing as described in D and Data center virtualization as described in C. The common denominator is that cloud computing service providers provide computing resources to users in a virtual machine, without having to understand the actual physical resource conditions behind the virtual machine. If a cloud platform serves only within a group, the cloud platform can also be called a "private cloud", and if a cloud platform serves the public, it can also be called a "public cloud". In general, the private cloud serves different divisions within the group (or applications), emphasizing the flexibility of virtual resource scheduling (e.g., end users can specify the processor, memory, and hard disk configuration of a virtual machine); Public cloud services Emphasize the standard of virtual resources (for example, public cloud service providers provide only a limited number of virtual machine product models, each virtual machine product model of the processor, memory and hard disk configuration is fixed, end users can only choose the most close to their own requirements of the virtual machine product model).

For public cloud service providers, the business model is similar to that of Amazon EC2. Therefore, public cloud service providers should usually choose cloud management software as described in D. For private cloud service providers, the software should be decided according to the management mode of computing resources within the group. If you centrally manage your computing resources and simply delegate the permissions of the virtual machine lifecycle management to the level of the department manager or the team leader, you should choose a Datacenter virtualization management system as described in C. If you want to delegate the rights of virtual machine lifecycle management to end users who really need to compute resources, you should choose cloud management software as described in D.

Traditionally, the private cloud is thought to be based on the enterprise's internal data center and its own hardware. But after the hardware vendors joined the cloud-computing service providers, the boundaries between the private cloud and the public cloud became increasingly blurred. Rackspace launched private cloud services, customers can choose to use their own data center and hardware, you can also choose to rent Rackspace data center and hardware. Oracle has recently further advanced the private cloud services "owned and managed by Oracle" (owned by Oracle, Consolidator by Oracle). In this new business model, the private cloud that customers enjoy is simply a pool of resources that are relatively isolated from other customers in the public cloud of the cloud service provider (you got private cloud in the I-Cloud). For cloud service providers, the infrastructure used to provide public cloud services may simply be a resource pool in their own infrastructure (private cloud), or even a resource pool in the hardware vendor's own infrastructure (private cloud) (You got public cloud in my private Cloud).

For customers, private cloud services that use data centers and hardware based on cloud service providers are financially reasonable. This means that the fixed asset investment (CapEX) of the data center and the purchase of hardware equipment becomes the operating cost of installment (OPEX), and valuable cash can be used as working capital for the expansion of the business. Even if the overall cost of owning such a private cloud over the long term is higher than building a data center and purchasing hardware, the return on using extra cash for business expansion may exceed the cost difference between two scenarios. In extreme cases, even if the enterprise ultimately did not succeed, there is no need to distressed the newly purchased a large pile of hardware equipment. Unless the real estate market is a big improvement in a short period of time, a company that is on the verge of failure will usually not regret not having built a data center. (It's important to note that for a company that can work long hours, earning money through real estate is entirely possible.) Before Sun was bought by Oracle, it had sold ancestral to turn the earnings around. ）

So what is the role of hardware makers in this game? When the user's fixed asset investment (CapEX) becomes the operating cost of installment (OPEX), does the hardware vendor need more time to recover the payment?

Willian jevons,1835-1882, British economist William Jevins, wrote a book titled "The Coal Problem" (the MacArthur Question) in 1865. Jevons describes a seemingly paradoxical phenomenon: improvements in the efficiency of the steam engine increase the energy conversion rate of coal, the increase in energy conversion leads to lower energy prices, and lower energy prices lead to higher coal consumption. This phenomenon is called the Jevons paradox, the core idea is that the increase in resource utilization leads to lower prices, which will eventually increase the use of resources. In the past 150 years, the Jevons paradox has been empirically used in many fields such as industrial raw materials, transportation, energy, food industry and so on.

The core value of the public cloud computing service is to transform the server, storage, network, and other hardware devices from the fixed assets purchased into the public resources. Virtualization technology improves the utilization of computing resources, leads to the decrease of computational resource price, and ultimately increases the usage of computing resources. Knowing this logic, you can see why HP will decisively join the OpenStack camp and take the lead on OpenStack based public cloud services as OpenStack is not yet mature. Of course, doing cloud computing does not necessarily save HP from crumbling, but if it does not do cloud computing, HP may not have much time to do so. Also, with this logic in view, it is possible to see why Oracle will dismiss cloud computing as a practitioner of cloud computing. After acquiring Sun, Oracle became the world's leading hardware provider overnight. The concept of cloud computing was just emerging, and Oracle's dismissive attitude suggests that it has not yet fully adapted to its status. Now that cloud computing has moved from a conceptual hype to a real-life exercise, Oracle, one of the major hardware vendors, is not going to take a slice of the cloud, it's that the true reflection arc is too long.

According to the Jevons paradox, for users, price reduction is the premise of increased dosage. So how do you price cloud computing resources?

At present, most of the virtual machine products of public cloud service providers are priced according to the configuration. In the case of Amazon EC2, the configuration of the type (Medium) virtual machine (3.75 GB memory, 2 ECU, 410 GB storage, 0.16 USD per hour) is a small (SGT) virtual machine (1.7 GB memory, 1 ECU Computing unit, 160 GB storage, 0.08 dollars per hour) twice times the price is twice times the size of a small virtual machine. The newly launched HP Cloud Services, as well as the domestic Sheng and Aliyun, are basically copying Amazon's EC2 pricing method. The problem is that the virtual machine's performance has not improved as the virtual machine is configured to improve. A series of performance test results for Amazon EC2, HP Cloud Services, Sheng, and Aliyun show that, for many types of applications, the price/performance ratio is actually decreasing as the configuration of virtual machines increases. Such pricing strategies clearly do not achieve the goal of encouraging users to use more computing resources.

It may be a more appropriate practice to price a virtual machine in terms of its performance. For example, a brand of soap has a size of two kinds of packaging, a small package with a soap and a large bag containing two pieces of soap. Users are willing to pay twice as much to buy a large package, often because the large package washes twice times more clothes than it does because it looks bigger. Similarly, different virtual machine products from the same public cloud service provider should be kept to the same level as possible. The problem is that different types of applications have significant differences in the demand for computing resources such as processors, memory, and storage, and their performance-configuration curves vary. Therefore, a framework for a comprehensive evaluation of virtual machine performance is needed in the public cloud service area, and the results obtained through this framework represent the integrated processing capabilities of a single virtual machine, not just any of the processors, memory, and storage. Based on such a test framework, not only can the product of the same public cloud service provider be compared, but the products of different public cloud service providers can be compared.

F, open source code

In recent years, we have observed a pattern in the field of information technology. When a closed-source solution succeeds in the marketplace, it will soon be followed by a number of open-source or closed sources that provide similar functionality (or services). (First of all, open source software, and then appear to compete with the closed-source software case is relatively rare.) In the operating system domain, Linux has gradually reached, or even surpassed, Unix's technical level, thereby replacing Unix's market position. In the virtualization arena, Xen and KVM are closely following VMware's technology development and breakthroughs, gradually eroding VMware's market share. In the field of cloud computing, Enomaly pioneered a closed-source solution based on the Amazon EC2, followed by an open source solution that represented eucalyptus and OpenStack. At the same time, the traditional sense of closed-source manufacturers of open source projects and community attitudes are also changing. For example, Microsoft, which has been hostile to open source projects for years, set up a subsidiary of Microsoft Open Technologies in April this year with the goal of promoting Microsoft's investment in open space, including interoperability, open standards and Open-source software.

The business environment we are in today is quite different from the recent rise of the Free Software Movement (Movement) in the last century of the 80. Since 1998, when Netscape first presented the term open source, open source has become a new software development, extension and sales model, not a substitute for commercial software. Compared with the traditional business model of closed-source software, the business model based on open source code has the following characteristics:

(1) Attracting potential customers and partners through keywords such as open source software or free software in the budding phase of the project. For potential customers, the choice of open source software can be free or low-cost access to closed-source software (part of) functionality. For partners, the point of interest may be in selling an enhanced version of the Open-source software (such as an Enterprise Edition), providing a solution based on open source software, or the success of the Open-source software may contribute to the sale of its own products.

(2) In the growth phase of the project, the main research and development personnel from the enterprise initiating the project and the project's corporate partners. There are also a few individual developers who contribute code to open source projects simply out of interest, but the number is relatively small. We often meet in the publicity materials of open source software similar to the description of "developed by So-and-so community". In the last 10 years, the major research and development forces in various "communities" have always come from a very limited number of business partners. But some open source projects often intentionally or unconsciously downplay the importance of corporate partners, or even mislead the audience into thinking that the community's main ingredient is individual developers.

(3) In the project harvesting phase, the project initiator and the main partner can obtain financial returns through the sales enhancement version or the solution offering. While other vendors can offer similar products or services, the main players in open source projects tend to have greater power of voice and authority in the marketplace. As for the profitability of open source projects, Marten Mickos, CEO of Eucalyptus, said during his tenure as CEO of MySQL: "If you want to succeed in open source software, you need to serve: (A) people who are willing to spend time to save money and (B) people who are willing to spend money to save time. "If a company is successful in open source, it should receive at least as much in terms of sales and services from open source software than in research and development." It is clear that some users are free to use open source software, although it is because their participation has reduced the open source software in the development and promotion of investment, on the other hand, because the pay users for open source software to pay more money.

So why is a solution based on open source software often cheaper than its closed-source rivals? In general, closed-source software as a pioneer in the field of market research, product design, research and development testing, marketing and so on are facing great challenges. Open source Software as a follower of closed source software, in the market research has closed source software as a successful case, in product design has closed source software as a reference template, in the promotion of sales also benefited from the closed-source software market expansion. In research and development, open source software appears later than the closed-source software, in this time period of technological progress will significantly reduce open source software into the relevant areas of the threshold. In addition, open source software may be in some features beyond the closed-source software, but at the overall level of its function of completeness, ease of use, stability, reliability will be less than the closed-source software. Therefore, the solution based on open source software will usually take the "30% of closed-source software to provide closed-source software 80% of the function" such marketing ideas. In addition, the customization of solutions based on open source software is particularly appealing to some customers.

In China's business environment, IT companies (or internet companies) are usually willing to spend time saving money rather than it companies (or traditional industries) that are usually willing to spend money to save time. It is important to point out that China's non-IT companies often do not care whether the software is open source, but pay great attention to the customization of open source software.

As a new business model, open source code has no higher moral standard than the traditional closed-source model. Similarly, it is not appropriate to judge different open source practices on a moral level. At the embryonic stage of the OpenStack project, Rackspace's promotional copy claims that OpenStack is "the only truly open source IaaS system in the world." Cloudstack, Eucalyptus and Opennebula, and other Open-source projects with similar capabilities have retained some of the closed-source Enterprise Editions (April 2012 ago, Both the Cloudstack project and the Eucalyptus both publish fully open source community and partially closed source Enterprise editions. After April 2012, the Eucalyptus Project announced full open source, and the Cloudstack project was bought and donated by Citrix to the Apache Foundation. or an automated installation package for paying customers only (Opennebula Pro is an automated installation package that includes enhancements, but all of its components are open source.) and is Rackspace classified as "not a true open source project." Similar propaganda lasted nearly two years, until Rackspace company launched the Rackspace Private cloud software based on the OpenStack project-an automated package similar to Opennebula Pro in nature. Opennebula Pro is a software package provided only to paid users, but any user can download and use Rackspace Private cloud software free of charge. The problem is that when users manage more than 20 servers, they need to seek help from the Rackspace company (to purchase the necessary technical support). Here we will not discuss the issue of limiting the number of nodes to 20 servers, and whether this part of the code is open source. The initiative and major contributors to open source projects have added to their repackaged distributions the ability to limit the scope of the software's application, which is difficult to explain ethically, but is normal at a commercial level. In the past two years, the OpenStack project in the research and development, promotion, community and other areas of the various measures taken, can be called based on open source code of the business model of the classic case.

As we mentioned earlier, there are often multiple competing open source projects in the same area. In the broad sense of cloud computing as an example, in addition to our familiar cloudstack, Eucalyptus, Opennebula, OpenStack, there are Convirt, XenServer, Oracle VM, OPENQRM, and so many other choices. For a specific application scenario, how to choose among the numerous open source solutions? According to my personal experience, the whole project selection process can be divided into three stages: demand analysis, technical analysis and business analysis.

(1) In the requirement analysis stage, the real purpose of using cloud computing technology is to dig into specific application scenarios. In China, many project decision makers ' knowledge of cloud computing often rests on the stage of "increasing resource utilization, reducing operational costs, and providing more convenience", without realizing that the list is already a basic function that most open source software can offer. In addition, many project decision makers default to the full functionality provided by VMware Vcenter as a requirement for open source software, regardless of whether specific projects require these features. Therefore, it is very necessary to research specific application scenarios, explicitly classify them according to the data center virtualization and the narrow sense of cloud computing, and further explore the specific functional requirements of the project. In many cases, data center virtualization and the narrow sense of cloud computing can meet the overall needs of customers, then the task of sales is to the customer's specific needs in favor of their own direction. This technique, we call customer expectations management (expectation Management). You can filter out a subset of the options by using the requirements analysis to define the classification of specific scenarios.

(2) In the technical analysis stage, first compares each open source software The reference structure, the emphasis considers in the specific application scenario according to the reference frame carries on the difficulty which the implementation faces. Second, in the functional level of the various open source software to compare, and will have to have the function (moment-in Have) and the ability to add points (good to Have) differential treatment. In addition, the installation configuration can also be difficult, the specific features of ease of use, reference documentation completeness, two development possibilities and so on to evaluate. Through technical analysis, can give each open source software scoring ranking, on this basis can eliminate the lowest scoring options.

(3) During the business analysis phase, it is important to make sure that decision-makers are willing to pay for open source solutions. If policymakers are unwilling to pay, then the project is "willing to spend time to save money" scenario, and vice versa is "willing to spend money to save time" scenario. Applications that are willing to spend time to save money rely primarily on the open source community for technical support, which can be used as an important reference for open source project community activism. For applications that are willing to spend money to save time, relying primarily on service providers for technical support, focus on the impact of service providers in the industry and local service capabilities, the community activism of open source projects is irrelevant.

In China's (narrowly defined) cloud computing market, cloudstack and eucalyptus are priorities for customers willing to pay for it. These two projects start early, with better stability and reliability, in the industry has a greater impact, and in the domestic team can provide support and services. At the same time, some domestic entrepreneurial teams began to provide OpenStack solutions, but in a short period of time it is difficult to accumulate the necessary actual combat experience, and has a wealth of experience in the Sina SAE team has not opened up to provide technical support business. Although there are some units in the domestic use of opennebula, but in the near future, it is difficult to form the ability to provide technical services to third parties. For customers who are willing to spend time, the advantages of Cloudstack and OpenStack are obvious, because they have a relatively high level of community activity. In both cases, cloudstack is richer and has more enterprise-class customers and success stories, and may be a better choice in the short term. In the long run, OpenStack solutions are becoming more prevalent, but other solutions are also making progress in technology and markets, making it difficult to shape eminence over the next three years. Purely commercially, cloudstack and eucalyptus have a greater chance of succeeding.

G, other

Some friends want me to add some of the cloud computing in China. To be frank, I don't have enough data at the moment to discuss it here. Liu Liming recently published an article entitled "Comment on Aliyun Sheng's cloud computing IaaS Industry", which is worthy of reference.