Sky Cloud software Guo Hong: The future trend and application of cloud Platform

December 12, the world's first large-scale conference to explore the industry Internet, 2014 CVW. The industry Internet conference was held in Beijing and was synchronized through the big screen of New York Times Square. The conference was made by the Asia Letter Group, the cloud base and the Chuang-Zhuang economic and Technological Development zone jointly hosted more than 5,000 global it and traditional industry leaders and elites who are concerned about the development of Internet and traditional industries, and explored the evolution of "Internet access to traditional industries", "traditional industry internet" and the technological model and business innovation of industrial Internet.

At the Cloud Platform @ Internet forum in the afternoon, Mr. Guo Hong, vice president of cloud software research and development, brought the keynote speech on the future trends and applications of the clouds platform to share his understanding of the cloud platform and the future of the cloud platform.

The following is the full text of his speech:

Guo Hong: Ladies and gentlemen, good afternoon! I am Guo Hong, vice president of cloud software research and development, I am responsible for the cloud platform of Sky cloud software products. Share with you this afternoon our understanding of the cloud platform and the future development of the cloud platform.

Today, everyone here to participate in the Cloud Platform forum is also a great interest in the cloud platform, the development of the cloud platform has a few years of history, what is the cloud platform, the cloud platform to solve what problems. The cloud is evolving from resources and virtualization, and we're talking about Amazon AWS based on virtualization, and virtualization is the first step. For the cloud platform, it is to realize the automation management of the resources on the basis of virtualization. So the first stage is IaaS, from the domestic 3 years ago began the construction of IaaS. One of the simplest understandings of the IaaS is the virtual machine, which is not, it includes the entire lifecycle management. Cloud can reflect its characteristics, in the Amazon's public cloud feature is very obvious is its dynamic, a virtual machine can be dynamically created, dynamic deletion, or even change, IaaS is to solve our entire creation process and lifecycle management.

Cloud to achieve the entire system cloud, it must implement the application, so from the application support or rise to a highly applied cloud platform, that is, SaaS cloud platform. In the end it must be a bunch of apps running in the clouds, which are supported by the application of cloud environments. So apps is the first step in the cloud platform.

The cloud platform is actually the operating system of the cloud, just because there are so many applications on the cloud, what are the resource management benchmarks for virtual machines and other storage networks? An operating system is actually the entire management of resources, so the cloud platform is actually the operating system of the cloud. At the same time, the cloud provides a variety of services, the IT resources packaged into services, application and invoke these resources to implement services. On this basis the need for heterogeneous, to support a variety of virtualization, more well-known is VMware, Hen, KVM, IBM. Virtualization technology is based on low load, a little different from the virtual machine, but the basic idea is similar, but the virtual machine does not have an operating system, with the computer system, so today I share with you is the development of the cloud platform is actually the support of the application, how to better support the application, the development of technology is very typical. It is not only a virtual machine, but the application of the environment into its basic consideration of the development of that environment can be migrated throughout the system. So we understand the basic architecture of a cloud platform, which is that I can support all kinds of virtualization or technology.

There are also several popular or open source standards for supporting these virtualized industrial platforms, like OpenStack. Our idea is a cloud platform based on these standards, has done a lot of virtualization work, on the virtualization also opened a virtual machine, including IP allocation, including some HA and storage of the entire management basic functions are actually done. So the cloud platform is actually an operating system to manage these open source middleware and implement the entire support for virtualization through middleware. At the same time, cloud Platform uses OpenStack, Cloudstack virtualization to achieve landing.

Now more and more public cloud, Amazon is the first cloud, we build a private cloud will want to follow this idea to build a private cloud here, but also to consider with my customers docking. From the perspective of the cloud platform, I manage the private cloud and build resources on the public cloud, using AWS or Aliyun interfaces to get resources out of the private cloud and the public cloud.

As a cloud platform we use ready-made, we do not have to do repetitive work, but the open source package is not everyone to take down will be able to apply, it is not a product, just say it has a function. In fact, like Hadoop, you are responsible for developing functionality. Openstack3 years ago is just a frame, it may not be easy to use, but in recent years with the increase of large companies, more and more companies, including IBM, Hewlett-Packard, many of them are based on OpenStack. Like OpenStack's heat, it is an automated deployment that solves the application of a machine and a machine. I have a machine that puts the application in and I can automate the deployment so that one of the machines I get is a machine with applications. This is the standard defined by heat.

In other OpenStack, there are stored, computational, network, Real in the private cloud can be used in the calculation of the main, storage and network platform Many are on the public cloud. The heat can be well integrated on a private cloud.

Docker for the idea and heat is the same, itself is based on the container, the bottom with the CPU and memory isolation, and the virtual machine is the same. At the same time, a file system, each container has its own file system. I can also stack up, packaged into an application is an application server, the server includes both the application system and the application. So you can make a template, templates are templates for applications, middleware, and operating systems, and this template can be replicated on different machines, not only virtual machines, Docker is a container, the concept is not equivalent to the virtual machine, but to the heat equivalent, I am an application container, this container can include the machine, Including the above middleware and applications. The development of this technology makes the application of the support to do better, and more can reflect the idea of the cloud. Docker has an ecological chain and can build Docker clusters. Our idea is that there are already ready-made things, and on this basis I manage it, realize the cloud and apply to the ground, so it is implemented as our infrastructure. We have made it our distribution, but also can tune, according to our needs to develop the function.

This is what the cloud infrastructure can do, but are these features clouds? No, how to achieve the best support for the application. What our cloud platform has to do is to support the application, and I can make the application run the fastest in the cloud. Including the application of automated deployment, the application may have middleware, if the Java word is outsourced, have their own code, these can be automated deployment to achieve my automatic growth. I want to add a machine, I can do it automatically.

Another is the concept of templates. Virtual machine template, I put the application can be used after the plug in. This application of the template can be combined with Docker, for the application of my requirements is to apply to help me turn up on the line. On the application I can have the cluster application or the page number, we can automatically deploy the page number.

Application of the entire management can be based on OpenStack, can also be based on Cloudstack, as Docker development, the underlying resource layer can be deployed through Docker.

On the right is a business system, assuming that there are three tiers, Web servers, application servers, and databases, each layer can be deployed in a virtual machine, or deployed in a Docker, Docker is a lightweight virtual machine. Such a physical machine is isomorphic, it can be a bunch of Docker to run, more efficient.

For application management, it's actually how I define an application that, through the template, allows an application to finally be deployed on the cloud.

To define an application with a multi-tiered architecture, so using a graphical editor to edit the structure of each layer, dragging the way, such as the left is a good application template, you can drag these templates into the editor to define my network level, it defines the entire application of the architecture, landing can be a docker, It can also be a virtual machine. Application management can be configured to generate what you want.

The application of the template or the application of the container benefits is the ability to take advantage of the development environment, test environment, production environment, in fact, to solve a large number of applications online are manually deployed.

The application system is not a set of deployment, but the deployment of multiple sets, the university is an example. A teacher in a class, this course is a pile of machines to form an application, the application into a template, so that each student to deploy a more set, he can do his application development. Through the template way can let my development, test and production is a set, this environment and Docker application environment is coincide.

Now Docker is very hot, even Microsoft says he wants to support Docker,docker is based on Linux, he wants to support, Amazon also support. So if a variety of applications are isomorphic or require multiple cluster applications, it may be better to choose Docker than virtualization. Large cluster applications are more suitable for Docker, virtual machines are small applications, because each is different, so you need the virtual machine operating system.

With the concept of applying a template, the cloud platform is actually an all-in-one machine that accommodates a wide variety of applications, this goes up to a height, starts just to apply for the virtual machine, now the cloud platform to the application, I want to let the application on the cloud platform landing, an application as a whole, I can publish this application for more people to use this application. So his idea is to encapsulate all kinds of ability, provide platform application level service, a variety of services on the platform to form a store.

Here's an example where we work with an operator can be his ICG capabilities, such as location services, mobile positioning services, are packaged into the ability to allow Third-party applications can run on these capabilities, but its size can be changed, the template has only one set, the whole logic is exactly the same. Through this platform I have different capabilities of encapsulation, the application can be better supported, using the application of services also through this platform to achieve its operation. At the same time through the development of management so that more applications through the standard application of the release of the interface can be published to the Cloud Platform, which formed a lot of applications on the cloud platform landing.

The first three years in the Yunping platform open virtual machine is the focus of cloud platforms, now focus on the application of the cloud platform how to apply and manage. As applications or cloud platforms increase in size, scheduling is becoming more and more important. Our scheduling is the core of the cloud platform, the larger the size you need to schedule. So when we first built the cloud platform, said I directly open the virtual machine is over, the virtual machine is also a few, people can tube, but when you have thousands of units, tens of thousands of units, and the number of machines in the change, people are certainly not to do, this time need intelligence, this intelligence is the operating system to schedule. So the scheduling is the core of the cloud platform, the future who will do a good job of the cloud platform can be a foothold. In other words, how to maximize the public cloud platform to improve its application, but also to meet the needs of users of various services.

The operating system is a machine, the hardware of the machine and the software running on it are the operating system, so a machine must have an operating system. From UNIX is mainly the process scheduling, it is matching demand, application, through the middleware Java program, operating system is a process, you have to apply for CPU, application memory, the application of hard disk, operating system to help you to achieve, this is a stand-alone operating system concept. In the cloud, the cloud is the management of all resources, can have computing, network, storage, in the future many soft functions can be made into resources, users or applications, but the application is not a stand-alone application, may be cluster applications. Therefore, the cloud platform is the core of its scheduling, it is responsible for the cloud management of resources and applications between the scheduling, this scheduling is the core of the cloud platform.

The resources below are as good as possible and no problem, users do not feel that my service is different, here refers to the concept of SLA, different services may have different SLA requirements. In addition, different applications may affect each other, whether I take the CPU to occupy the other people can not do the work, each other is not affected, this is the problem of scheduling to solve.

The most basic scheduling may be based on load scheduling, which means that, depending on the resource usage, where should I put my resources when I apply for resources. The load will change, the lowest is not to say that the moment is the lowest, but to make the whole system how the most effective, so to consider synthetically.

For a variety of services to the cloud, different services are different needs, and some very key, even if the utilization rate is not high also to leave the resources to it, some may be tested, I am not anxious, no resources are OK. So different SLAs have different requirements, requiring different transfers to manage different SLAs, and satisfying different services can run on one platform. So based on SLA standards I want to know its priority, while cloud computing is to use resources on demand, your resources will face an unlimited amount, so there is a limit to the SLA, but also the actual use of tracking.

With the increasing of cloud system, scheduling may become a bottleneck, so a distributed scheduling system is needed. The cloud itself is already a distributed, it is a dispatch, the dispatch needs more distributed.

Before we share is that we believe that the focus of the cloud platform is an application of support and application of the landing, or later will be called Application platform, scheduling is its core. From our product development, we think these two are our development direction, do cloud platform when you can think about this problem. Through cloud services you can provide different services, but it is the most fundamental resources, the previous release of the SkyForm3.0 can support OpenStack, but also to support the Cloudstack, mainly support the virtual machine services, the end of the year we added physical machines and storage services and so on. At the same time we have added scalability, support the scale to improve, currently our version can support to 2000 physical machines and 10,000 virtual machines. Within half a year users have this requirement to reach 100,000 virtual machines, there will be three-dimensional distributed technology, maybe I am not a set of systems, but more than a set of systems distributed everyone work together, so like Amazon their number of machines will also be layered partition.

We hope that it will be used to run on the cloud, such as Docker we are looking for very valuable applications, with Docker to achieve this cloud and the entire scheduling. Operators now large applications in small machines, requiring x86,x86 not a single machine can be done, it must be a cluster, may be a virtual machine, may also be Docker, and its application itself is a large cluster application, so Docker will be better. We will work with customers to make the latest technology in the cloud to achieve the entire cloud.

Later we will support the network SDN, which is related to hardware, through the VPC, we are currently building public cloud, private cloud, to the two through, so that they realize sharing between. The system is more and more big, people more and more wide, rely on people have not understand, need an intelligent system can tell me exactly how, in the future, so to do intelligent analysis and prediction.

(Responsible editor: Mengyishan)