Eight typical problems commonly encountered on the road of private cloud construction

Problem one: On the basis of virtualization construction, the choice of three private cloud construction direction is discussed.

As we all know, virtualization and cloud computing are different, when the scale and day of virtualization increase, to the urgency of automation, standardization, process and service quality requirements to a certain extent, we will choose to carry out enterprise private cloud construction.

But private cloud construction currently has a roughly three-way direction:

1. On the basis of the original virtualization, we use all levels (IAAS, PAAS, SAAS) cloud management platform products from ready-made vendors to provide unified access, unified management and unified process for virtualization.

2. On the basis of the original virtualization, the use of standard open source framework, such as Openstack,kubernetes, according to the needs of enterprises, tailored to their own cloud computing needs.

3. On the basis of the original virtualization, from the basic framework to the requirements of the software all according to the needs of the enterprise, tailored development, more relevant to the actual enterprise, high safety factor, strong reliability.

What are the three directions for the above private cloud building? How to choose the direction of the enterprise.

Reply:

The recommendations are as follows:

The commercial bank chooses that route, still needs to look at its own strategic plan and the personnel quality ability. The main differences between the two routes are as follows:

Open Source Private Cloud: Code self-control, platform compatibility, customization ability, but need to have a large number of personnel and financial input, and is a continuous investment, personnel quality and financial must keep up with, at the same time, open source product version Iteration fast, robustness is not enough, the direction is not clear (maybe we have a brain change a frame \\\\ The cost of error correction is high. In short, their own use of open source to build and develop a private cloud, to their own ability to bring great challenges, require enterprises to play a protracted war, and can continue to enrich and absorb nutrients in the community. There is also a layer, is the development, maintenance has its own completion, no third-party risk-shifting.

Commercial software: Shortcomings we are very clear, easy to be vendor binding, poor compatibility, customized poor, with the expansion of the scale of the increase in cost is obvious, but the characteristics of the implementation cycle is short, for the enterprise's own personnel quality than the situation of self-development requirements are much lower, mainly the role of product managers and users. And for the maintenance of the system, OLA can pass the part risk through the business way. The comparison applies to the enterprise scale is small, asks the cloud as soon as possible situation.

Finally, a lot of commercial software is built on the basis of the open source private cloud, you can consider the combination of the two advantages, open-source way to enhance openness, through the commercial way to reduce their own construction costs.

The goal is the 3rd, but requires the enterprise's own ability to strengthen, can consider based on open source commercial products, and demand enough open, gradually accumulate experience, slowly do 2 coexistence, gradually replaced.

Question two: What is the difference between cloud and virtualization? The clouds are much more.

1: To achieve what kind of hardware scale, only on the need for private cloud.

There are hundreds of thousand that is on, some 10 units 20 is also on. Each enterprise may measure a different standard. How to judge whether to go on.

2: In the case of virtualization, what kind of help the cloud can bring.

Clouds are much more than virtualization.

Reply:

More cloud services, an Infrastructure encapsulation service model. Virtualization does not provide Saas,paas, nor does it provide billing, metering, service development, and self-service customization.

Virtualization is just one of the more convenient technologies to support cloud computing, not cloud computing.

Then go back to the question and don't go on the private cloud. In fact, we are here to analyze the private cloud is what the demand, whether these appeals are worth investing, input-output how.

Oberyun, Cloud Management Appeal 1: There is a strong demand for infrastructure services, many times this demand from the resource scale, organizational structure of the Division, infrastructure departments for a number of development environments, the urgent need through the encapsulation of their own services, the service capabilities of their own clear, service operating costs, the responsibility to distinguish, A good way to provide cloud services. Claim 2: The demand for efficient use and batch management of large-scale resources. Cloud computing is economies of scale, scale does not go out, and therefore, according to their own ability and scale, to see if it is worthwhile to go on, whether it can continue to invest. Appeal 3, is also the most important demand, whether the supply contradiction is outstanding, decided whether to take this road, the virtual environment is now tube is very good, very convenient, to meet business needs, then, the individual think that this contradiction is not to change productivity when, when change. Infrastructure requirements of the speed of the existing means to keep up, can not meet the application of rapid expansion of flexibility, can not meet the application of higher levels, such as the demand for Paas\\\\saas, or the need to pass the standard framework of control, mass operation. This is when the contradiction is prominent, is also the cloud day.

Therefore, in the case of financial and human support, the infrastructure can use existing virtualization to further the cloud computing environment (also consider network, storage, process input), is possible, but not necessary, need to consider their own cost, scale, tailored.

Question three: How to realize the upgrade and transformation of private cloud in the situation of not big change. Should pay attention to those aspects of the problem.

Our existing virtualized environments include Powervm,vmware two, and our application environment requires that we only do private cloud considerations, but according to our understanding of the requirements of private cloud construction, it is necessary to involve the transformation of network and storage in the existing environment. Ask the experts, how to achieve or partially realize the transformation of the private cloud function, and improve the new requirements of the application environment through the application of new technology in the case of not big change. can also improve operations and maintenance management and other aspects of support.

Reply:

If the server as a point of application in the data center, the network, storage often belong to the surface, belonging to the underlying infrastructure, the difficulty and risk of transformation is higher than the computational resources. In the traditional industry, there is no opportunity for a new environment, the gradual transformation, is also difficult. Inevitably, the network and storage in the private cloud to build some technology to keep up.

Before talking about the private cloud understanding, not necessarily the private cloud must go with new technologies, such as SDN, storage virtualization, the focus of the private cloud is to achieve customer-oriented service model support, resource resilience and rapid service capabilities. At this point, the general enterprise through the use of the existing environment, to standardize the allocation of efforts to promote the development and construction of automation capabilities, and cloud platform to integrate, while enhancing the way of pre-capacity planning, but also to gradually realize the cloud environment, at the same time, combined with the new opportunity, Gradually replace the original infrastructure with technologies and devices that better support the characteristics of cloud computing. ，

Question four: What is the disaster-tolerant implementation of a private cloud?

Private cloud local disaster recovery, remote disaster recovery, disaster tolerance in different data centers, OpenStack \ Distributed storage, etc., what are the specific scenarios, success stories, and precautions in the disaster-tolerant scenarios.

Reply:

Here's a concept, first of all to see if you understand the consensus, I think disaster preparedness matters, in the cloud environment, divided into two, one is the disaster preparedness cloud, one is the cloud disaster preparedness. Disaster preparedness Cloud refers to the primary provision of services in this cloud environment for disaster preparedness, cloud disaster preparedness, referring to the full backup of the cloud environment to the disaster preparedness Center.

Feel the landlord wants to ask is the second how to achieve, in OpenStack and distributed storage, do disaster preparedness I understand that there is no mature solution in the industry, and in the cloud computing industry, especially in the private cloud, we basically use disaster preparedness cloud to solve the problem of disaster tolerance. The main realization is to build disaster preparedness cloud environment in a remote place, the application uses the traditional way, carries on the data level, the application level disaster recovery, because the disaster preparation cloud environment elasticity, often these resources daily is used and the development test multiplexing, manifests the disaster preparation cloud benefit, this in the past, the disaster preparation test multiplexing often is very complex, and is only applied in some low-grade systems. The benefit is that the more the application layer is implemented, the less the isomorphism requirements for the underlying infrastructure can be as much as possible to save the cost burden of disaster preparedness. It also realizes the loose coupling between application reliability requirements and infrastructure, which is obviously in line with the characteristics of distributed, Internet architecture and cloud computing.

If you want to do cloud disaster preparedness, basically still involves data files, virtual machine file replication, the requirements of the underlying architecture is also high, distributed storage, cloud management platform backup, but also brings complexity. Often, such a disaster can lead to a slight change in infrastructure, a change in the end, and an impact on the application. As a result, most of the vendors that provide the public cloud are mostly using distributed deployment applications and data, rather than taking the entire cloud in a way that is fully disaster-tolerant. For reference

Question five: How server virtualization Transitions to the private cloud.

At present, the company has several sets of virtualization systems, all using VMware for deployment management, each system has its own separate San network and storage. How can the next step be developed to the private cloud transition, technical how to achieve. Now I can think of is the use of SVC or similar devices to integrate all storage resources, build a unified cloud storage, as to how the host computing resources deployment is not well planned, are all the host to be added to a large San network, I hope experts give guidance.

Reply:

Personal opinion, biased forgive. You may understand the concept of a private cloud some deviation, the private cloud does not represent unlimited resources, no boundaries, does not mean that the entire data center into a resource pool.

The resource pool must have a boundary. Resource pools have their own boundaries, no matter how large or small the data center is. There is no need to get all the sans through, and it is not necessary to add all the hosts to the large San network.

Proposed scenario: the boundary of a resource pool is an organic combination of the capacity boundaries of each domain. For example, according to the storage capacity of cloud services, computing capacity, the two-tier network needs to design a modular resource pool. 4c8g500g Services, plan how many of this module to plan the size of the SAN, and plan the ratio of compute resources. Capacity on the basis of performance prediction, leaving the relevant elastic resources. In this way, the construction of several resource pools, will be more than the entire data center as an integrated resource loss of some capacity, but for all areas of comprehensive strength and integration is economic. At the same time, it is suggested that the resource pool, to do related resources to open up technical means, this means may not be elastic heat transfer, but to have this ability, when needed, through the technical means can realize some of the resource migration to solve some of the capacity hot issues.

Inquiry: At present, according to the company's internal network of different or different business deployment of the virtualization system, so whether it is computing resources or storage resources should be considered to have boundaries. How to further upgrade to a private cloud, what more needs to be strengthened.

Reply: Then should be the overall management of these virtual pools of the cloud platform, the demand through the cloud platform to provide and inward supply, distinguish between different service requests, in different areas of the implementation of resource pool automation

Question six: How to determine whether a system is deployed on a private cloud, and what the indicators are.

For example: Credit scoring system, billing system, etc.

Reply:

The most typical word to describe a "cloud native" application. Give you a reference, agile development of the 12 principles, to meet the application of the 12 principles, basically in the cloud computing in this distributed, virtualized environment can be well run. Here, personally, the most important thing is to cluster, support the application of compensation mechanism, modularization. Refer to Http://12factor.net.

Use only one copy of the benchmark code, but can be deployed to multiple environments

Dependencies between apps if specified, such as profile descriptions, do not use implicit code associations

The configuration is provided to the application in the same way as the environment variables, rather than in constant or code-related ways in the code.

Code uses resources such as databases, message queues, Distributed file systems, and so on, to be available as a attach resource, and not to write dead code. Resources are provided as resource strings that can be injected from environment variables to the application and immediately served

Strictly will compile, release, production and other environments to isolate, even if you want to change the production configuration, you need to use a continuous release from the compilation to build and automatically publish to the production system, do not directly change the production system

Stateless processes provide services in a way that applications need to be stateless without sharing. If you need to maintain a state or share, you need to use an external service, such as external session manager, etc.

Use address and port binding to provide services, such as the consumer of an application service only need to know the URI address and the corresponding port, after binding will be able to consume the service

Scale-out through the process model, where applications or micro-services are scaled horizontally and linearly to extend support capabilities through multiple instances

The usual application process is designed to be quick to start and gracefully terminate the destruction mode, which facilitates fault recovery and scale-out

The development environment is equivalent to the online environment, as much as possible to maintain the same development, pre-release, online production environment. To be able to continue publishing requires minimizing the difference between local and online production environments. Try to oppose the use of different backend services in different environments

The log is treated as an event stream, summarizing the entire log to monitor the application and environment of the platform, so that the root cause of the problem can be discovered through the comprehensive analysis of big data.

Management or other tasks are treated as a one-time process, such as performing a disposable system check, a snapshot of a health state, and so on.

Question seven: How to achieve unified resources in a private cloud environment.

Customer's actual operating environment is very complex, how to achieve the different platform resources of the unified cloud, from the host, storage, network at different levels can be automated management, to achieve self-service.

Reply:

The key test is the openness of the cloud platform.

Enterprises in their choice of cloud platform to fully consider their own environment, how many types to be involved, how many platforms to be satisfied with the tube, whether to support heterogeneous, whether the support of modular access.

It is recommended that the cloud platform is positioned as a tool framework to achieve sufficient openness, standardized interface and unified format access, all areas in accordance with standards to complete their own automated packaging, self-configuration acquisition, unified view display on the cloud platform, cloud platform through the process engine scheduling domain modules, implementation of operations, nanotubes.

At the implementation level, basically is the domain responsibility system, each wants to be managed the platform (calculates, the storage, the network and so on), completes own development and the service registration.

Question eight: What are the requirements on the private cloud for the enterprise?

1: What are the requirements for operation and maintenance level? or the direction of the quality or effort required after the cloud.

What are the typical requirements for the OPS team? System norms, technical level, role personnel.

2: Standardization degree requirement.

System norms, the level of enterprise overall it governance stage.

Reply:

The most important requirements of operation and maintenance level are the use of automated operation and maintenance tools, cross-domain synergy, operation and maintenance organization structure adjustment, and the transformation of operation and maintenance culture.

Automated operation and maintenance tools to deal with the massive operation and maintenance, the most basic requirement is the accuracy of configuration management, or who dares to go to the automation.

Cross-domain collaboration is very important in the construction of private cloud, in large enterprises, network, computing, storage, middleware and other fields, often are independent departments, there is an independent change and implementation process, but in the private cloud design, Hinves, this must be one, even if there is a virtual team to undertake. This means that, in general, the organizational structure should be adjusted.

The personnel must have the capacity to face the massive infrastructure operation and maintenance, must have the structure team, the demand Analysis team. Personnel to have the operation and maintenance tools development capabilities, this suggestion Baidu Google's SRE team, is a very good location.

Talking about standardization, is the most serious, but also the private cloud is the most characteristic and the most advantageous one link, the enterprise structure control does well, the standardization degree is high, decides the cloud computing level, the SaaS service provides, relies on the high standardization. From the physical hardware layer, os\\net\\db\\storage all areas of standardization, and then continue to standardize the deployment mode of application, gradually standardize the formation of standards, in favor of the provision of PAAs, true to the specific application of standard words, to complete the transformation of SaaS.