Cloud computing and grid computing

Cloud computing and grid computing-general Linux technology-Linux technology and application information. The following is a detailed description. Introduction

You may be very concerned about the comparison between cloud computing and grid computing. This article introduces the cloud computing service types, and similarities and differences between cloud computing and grid computing. At the same time, this article discusses the advantages of cloud computing over grid computing. The two face common problems and some security issues. This document uses Amazon Web Services as an example.

The implementation of cloud computing requires three parts: thin clients (or clients capable of switching between fat and thin), grid computing, and utility computing. Grid computing connects independent computers into a large infrastructure to make full use of idle resources. Utility computing is the service used to pay for shared servers, just like paying for public utilities (such as electricity and natural gas ).

Through GRID Computing, computing resources can be provided as utilities that can be turned off. Cloud computing goes further, providing computing resources as needed. In this way, we can avoid excessive supply when using public pricing. This eliminates the need for over-supply when meeting the needs of millions of users.

Infrastructure as a service

Consumers can obtain services from a complete computer infrastructure through the Internet. Such services are called Infrastructure as a Service (IaaS ). Internet-based services (such as storage and databases) are part of IaaS. Other types of services on the Internet include Platform as a Service (PaaS) and Software as a Service (SaaS ). PaaS provides complete or partial application development that users can access. SaaS provides complete and directly usable applications, such as managing enterprise resources over the Internet.

As an example of Infrastructure as a Service (IaaS) in practical applications, The New York Times uses hundreds of Amazon EC2 instances to process terabytes of document data within 36 hours. Without EC2, The New York Times will take several days or months to process The data.

IaaS has two usage types: public and private. Amazon EC2 uses a public server pool in the infrastructure cloud. More private services use a group of public or private server pools in the enterprise's internal data centers. If software is developed in an enterprise data center environment, both types can be used, and the cost of temporary resource scaling using EC2 is also low-for example, testing. By combining the two methods, you can develop applications and services faster and shorten the development and test cycles.

Amazon Web Services

With EC2, you can create your own Amazon Machine Images (AMI), including the operating system, applications, and data, and control how many instances of each AMI run at a given time. The customer pays for the instance time (and bandwidth) used, increases computing resources during peak hours, and reduces computing resources when not needed. EC2, Simple Storage Service (S3), and other Amazon products can provide services to millions of users over the Internet.

Amazon provides five different types of servers, from single-core x86 servers to eight-core x86_64 servers. You do not need to know which server is used to provide the service instance. You can place instances in different geographic locations or zones. Amazon allows elastic IP addresses and can be dynamically allocated to instances.

Cloud computing

With cloud computing, enterprises can greatly improve their computing capabilities without investing in new infrastructure, carrying out new training or purchasing new software licenses. Cloud computing is most suitable for small and medium-sized enterprises that want to outsource all their data center infrastructure, or large enterprises that want to achieve higher load capabilities without having to build larger data centers at high costs. In either case, service consumers use the services they need on the Internet and only pay for the services they use.

Service consumers no longer need to stay next to the PC to use applications on the PC, or purchase versions for specific smart phones, PDAs, and other devices. Consumers do not have to own cloud infrastructure, software, or platforms, thus reducing upfront costs, capital expenditures, and operating costs. Consumers do not need to worry about how to maintain servers and networks in the cloud. Consumers can access multiple servers anywhere without knowing which server is used and where they are located.

Grid computing

Cloud computing evolved from grid computing and can provide resources as needed. Grid computing can or may not be in the cloud, depending on what type of users are using it. If users are system administrators and integrators, they will be concerned about how to maintain the cloud. They upgrade, install, and Virtualize servers and applications. If you are a consumer, you do not have to worry about how the system runs.

Grid computing requires that the use of software can be divided into multiple parts, the program segment as a large system image to pass to thousands of computers. One problem with the grid is that if the software segment on a node fails, it may affect the software segment on other nodes. If this segment can use failover components on other nodes, the problem can be mitigated. However, if the software segment relies on other software segments to complete one or more grid computing tasks, the problem persists. Large System Images and hardware for operation and maintenance may cause high capital and operating expenses.

Similarities and Differences

Cloud computing and grid computing are both scalable. Scalability is achieved through load balancing of Application Instances running independently on various operating systems connected through Web Services. CPU and network bandwidth are allocated and recycled as needed. The system storage capacity is adjusted based on the number of users, the number of instances, and the amount of data transferred at a specific time.

Both types of computing involve multitenancy and multi-task, that is, many users can execute different tasks and access one or more application instances. Large User pools are used to share resources to reduce infrastructure costs and improve peak load capabilities. Both cloud computing and grid computing provide service level agreements (SLAs) to ensure availability, such as 99%. If the service does not reach the promised normal running time, the consumer will receive service compensation due to data delay.

Amazon S3 provides a Web Service for storage and data retrieval in the cloud. Set the maximum number of objects that can be stored in S3. Objects of only one byte can be stored, and objects of 5 GB or even TB level can be stored. S3 uses the bucket as the container for each storage location of the object. These data are stored securely using the same data storage infrastructure as Amazon e-commerce websites.

Although storage computing in the grid is very suitable for data-intensive storage, it is economically inappropriate to store a Byte object. In the data grid, the number of distributed data must be large enough to maximize the benefits.

Computing Grid focuses on operations with a large amount of computing. Amazon Web Services in cloud computing provides two types of instances: Standard and high CPU.

Issues to consider

Cloud computing and grid computing face four problems: Threshold policies, interoperability, hidden costs, and abnormal behavior.

Threshold Policy

Suppose there is a program that performs credit card verification on the cloud and hits the peak sales season in March. So we found a higher requirement, so we created more instances to meet this requirement. As the sales season passes, this demand will decrease, and resource instances will be recycled and re-allocated to other applications.

To verify whether a program can work, a threshold strategy needs to be developed, improved, and implemented in the pilot study phase before being transferred to a real production environment. Check whether this policy can discover a sudden increase in demand, so as to create more instances to meet these needs. See how to reclaim idle resources and transfer them to other jobs.

Interoperability problems

If enterprises outsource or create applications together with a cloud computing vendor, it may be difficult to switch to other providers that use private APIs or have different import and export data formats. This results in interoperability between two cloud computing vendors. You may need to change the data format or application logic. Although there are no cloud computing industry standards for API or data import and export, IBM has partnered with Amazon Web Services to achieve interoperability.

Implied cost

Cloud computing does not tell you what the hidden costs are. For example, if an enterprise wants to use the cloud storage service provided by the service provider and database applications that contain TB-level data, it may have to pay a high network fee. This cost is higher than the cost of purchasing new infrastructure, training new employees, or purchasing new software licenses. Another example of network costs is that if an enterprise is far away from a cloud provider, there may be a long delay, especially when traffic is high.

Unpredictable behavior

Assume that the credit card verification application runs well in the enterprise's internal data center. Applications in the cloud need to be tested through a feasibility test to check for unpredictable behavior. For example, check how the app verifies the credit card, how to allocate resources and release idle resources during peak sales in March, and switch to other tasks. If an exception occurs during credit card verification or the release of idle resources, you need to resolve these problems before placing them in the cloud.

Security Questions

In February 2008, Amazon S3 and EC2 experienced three hours of downtime. Although the SLA provides data recovery and compensation for this situation, during this period, the user loses the sales opportunity and the executives do not have the necessary important business information.

Do not passively wait for downtime. Users should check their own security-to see how much data the supplier can restore. Testing is very simple. No special tools are required. You only need to request the data stored in the past to see how long it takes for the supplier to restore it. If it takes a long time, ask the supplier how much service compensation can be obtained under different circumstances. Check whether the checksum matches the original data.

One aspect of security testing is to use a trusted algorithm to encrypt data on a local machine and then use a decryption key to access data on a remote server in the cloud. If the data that has been accessed cannot be read, the encryption key is damaged or the supplier uses its own encryption algorithm. You may need to know the algorithm from the supplier.

Another is potential cloud data problems. To protect data, you may need to manage your own private keys. Ask the supplier about how to manage the private key. If you sign the certificate, Amazon will provide a certificate.

Software Development in the cloud

When developing software using high-end databases, the most likely option is to use the ECS pool of the enterprise's internal data center. during testing, you can use Amazon Web Services to temporarily expand resources. In this way, project management personnel can better control costs, manage security issues, and allocate resources. Project management personnel can allocate different hardware resources for different cloud types: Web development cloud, test cloud, and product cloud. Different types of cloud costs are also different. The unit time cost of cloud development may be lower than that of the product cloud, because additional features such as SLA and security are distributed in the product cloud.

A project can be restricted to a specific cloud. For example, some services on the product cloud can be used for product configuration. Development cloud services are only used for development. To optimize the assets at different stages of a software development project, managers can track usage by project and user to obtain cost data. If the cost is high, the management staff can use Amazon EC2 to temporarily expand resources at a low cost, as long as the security and data recovery problems have been solved.

Environmentally friendly cloud computing

One of the motivations of cloud computing is its greater environmental friendliness. First, the hardware required to run applications in the enterprise's internal data center is reduced. Replacing the hardware with cloud computing can reduce the power required to run the hardware and reduce the temperature. These systems can be integrated into the remote center for more effective management.

Second, the cloud computing technology has improved the telecom technology, such as remote printing and file transmission, which may reduce the demands for office space, new furniture purchase, old furniture elimination, and office cleaning. It also reduces the need to drive to work and carbon dioxide release.

Conclusion

This article compares the similarities and differences between cloud computing and grid computing, and introduces how to solve the problems in cloud computing and grid computing, recover data in pay-as-you-go environments, and manage the security of private keys, this helps you prepare for using cloud computing. Users' potential demands for larger Internet capacity pose challenges to developers and other members of the project team. Handle Web Applications