The relationship between cloud computing and parallel computing

Cloud computing is a computing model that represents a way to share resources to design, develop, deploy, run applications, and expand the resources to shrink and support application continuity. Most of today's cloud computing infrastructure is made up of user terminals delivering service-ready networks to data centers, as well as different levels of virtualization technology built on servers, so that people can use them wherever they have a network infrastructure. Therefore, the basic concept of the cloud is to automatically split the vast computational processing program into countless smaller sub-programs through the network, and then the large system composed of multiple servers to search, calculate and analyze the results back to the user terminal. With this technology, remote service providers can reach tens of millions of trillion or even billions of messages in seconds to achieve the same powerful Web services as "supercomputers".

You might say, isn't that the concept of distributed computing? Yes, cloud computing can be said to be distributed computing, parallel computing, grid computing and other same strain technical routes, and even that their genes are the same. But they are different in their appearance, and they differ in technical architecture. Simply put, cloud computing is a commercially packaged term that finds a profit model for distributed storage and distribution computing. Conceptually, cloud computing can be seen as an organic combination of "storage cloud + compute cloud", i.e. "cloud computing = Storage Cloud + Compute cloud". The underlying technology of storage cloud is distributed storage, and the basic technology of computing cloud is distributed computing, more accurately in parallel computing.

Therefore, the infrastructure of cloud computing is first and foremost to ensure that parallel computing can be implemented. The function of parallel computing is to split the large computing tasks, then distribute them to the various nodes in the cloud for distributed parallel computing, and finally collect the results after the unified processing. If cloud computing is a distillation of parallel computing, then cloud computing infrastructure is more advanced at another level: computational resource virtualization. Simply put, all computing resources in the cloud are considered as a pool of computing resources that can be allocated and recycled, and users can request the appropriate computing resources based on their actual needs. This resource virtualization benefits from the recent emergence of virtual machine technology, virtual machine to achieve the virtualization of resources, to avoid the characteristics of heterogeneous hardware. That is, no matter what kind of hardware machine is saved together, its computing resources can be quantified into the compute resource pool and dynamically allocated and dynamically adjusted. Dynamic adjustment is one of the greatest uses of virtualization technology to adjust the size of a resource without restarting the system. This virtualization is much the same as the virtualization technology we install on our own machines, and the difference is that the use of individual users is to virtualize the resources of a physical machine into multiple parts so that it can start more than one operating system at the same time, and the virtualization technology in the cloud is to virtualize the resources of multiple physical machines into A large pool of resources that makes the user feel like a machine using a huge resource.

From this point of view, cloud computing is not a new technology, but a novel resource utilization model. We can think of cloud computing as a phase upgrade of previous generation computing patterns, such as grid computing, transparent computing, and utility computing. The essence of cloud computing is a methodology of infrastructure management, which is to make a large amount of computing resources into an IT resource pool, which is used to dynamically create highly virtualized resources for users. In a cloud computing environment, all computing resources, such as servers, can be dynamically added and subtracted from the hardware infrastructure to suit the needs of the task. So, the most real and complete infrastructure for cloud computing is an end-to-end solution that integrates multiple products and services. The essence of cloud computing infrastructure is to maximize the utilization of IT investments through consolidation, sharing, and dynamic provisioning of hardware devices. Therefore, the nature of the computing cloud infrastructure can be seen as: Parallel computing + resource virtualization.

Intel divides the infrastructure of cloud computing into cloud data centers (bureaus) and applications (access terminals). For the local side, Intel's strategy is to optimize and improve the parallel computing power of the data center, because simply expanding scale is inefficient. At present, Intel's strategy to enhance cloud computing infrastructure is based on the Intel 5600 platform, enabling the next generation of servers to be more scalable, enabling more processor cores in the same cabinet space, significantly increasing memory capacity and built-in storage capacity, resulting in a significant increase in compute density and storage density for the product unit space , which is more beneficial for applications such as parallel computing, virtualization, server consolidation, database, and enterprise on-demand expansion, which will undoubtedly be ideal for enterprise deployment virtualization and cloud environments.

The relationship between cloud computing and parallel computing