Three related technical details of virtual machines in cloud environment

In order to improve the deployment speed of virtual machines in a cloud environment, we first need to consider parallel deployments and collaborative deployments. First look at the parallel deployment, which is to deploy the virtual machines to multiple physical machines at the same time, ideally, parallel deployments can multiply the time required for deployment, but this approach is susceptible to network bandwidth and cloud deployment server read and write capabilities. For example, when network bandwidth is limited, the cloud deployment server runs multiple deployment tasks that compete for network bandwidth, and deployment speed cannot be further enhanced when network bandwidth is fully occupied. In this case, the collaborative deployment technology can be used to further improve deployment speed. The following describes cloud computing.

The idea of a collaborative deployment is to increase the speed of deployment by mirroring the virtual machine across a network between multiple target physical machines, rather than just transferring between the deployment server and the target physical machine. Cooperative deployment technology can greatly improve the speed of system deployment, because there is a lot of shared bandwidth between physical machines, so the scheme is not affected by the bandwidth of competitive networks.

1. Resource Scheduling

The huge scale of cloud computing brings new challenges to resource scheduling. Resource scheduling needs to take into account the real-time use of resources, which requires real-time monitoring and management of the resources of cloud computing environment. Cloud computing environment in a variety of resources, large-scale, real-time monitoring and management of resources become very difficult. In addition, a cloud computing environment may have thousands of computing tasks, which challenges the complexity and effectiveness of scheduling algorithms. From the granularity of scheduling, the scheduling of internal application of virtual machine is more concerned by cloud users. How to dispatch resources to meet the service-level agreements within the virtual machine is also a problem to be solved at present.

2. Multi-tenant technology

Compared with traditional software operation and maintenance mode, cloud computing requires hardware resources and software resources to be better shared, with good scalability, and any enterprise user can configure SaaS software according to their requirements without affecting the use of other users. Multi-tenant technology is the key technology to meet the above requirements in the cloud computing environment at present.

It is now widely accepted that SaaS with Multi-Tenant technology should have two basic characteristics: The 1th is that SaaS applications are web-based, can serve a large number of tenants and can be easily scaled 2nd requires that the SaaS platform provide additional business logic that enables tenants to extend the SaaS platform itself to meet the needs of larger enterprises. The technical challenges faced by Multi-Tenant technologies include data isolation, customized configuration, schema expansion, and performance customization.

Data isolation means that when multiple tenants use a system, the tenant's business data is isolated from each other, and the business data of different tenants do not interfere with each other. There are three ways to manage data management for multiple tenants: Create a separate database for each tenant, data from multiple tenants are stored in the same database, differentiated by different schemas, and multiple tenants not only deposit the same database and use the same schema, That is, the data is stored in a table and is differentiated by the tenant's identification code.

A customized configuration is a SaaS application that enables different customers to customize the configuration of SaaS applications. Schema expansion is the ability of multiple tenant services to provide a flexible, highly scalable infrastructure that guarantees the performance of multiple tenant platforms under different workloads. Performance customization means that for a SaaS application, the performance requirements of different customers may be different, and how to flexibly configure performance for different customers on this set of shared resources is a difficulty in multi-tenant technology.

3. Mass data processing

The most popular programming model for mass data processing is the MapReduce programming model designed by Google. The MapReduce programming model divides a task into fine-grained subtasks that can be scheduled between idle processing nodes, allowing the nodes to handle the more tasks as the processing speed is handled, thus avoiding the process of slow nodes prolonging the completion time of the entire task. The above describes cloud computing.