A deep comparison between cloud computing and grid computing

Different goals

In general, when it comes to grid computing, everyone will think about Rage's search for aliens, that is, by installing a screen saver on this machine, you can take advantage of everyone's PC's spare time computing power to participate in the search for extraterrestrial computing.

This also illustrates the goal of the grid, which is to use as many resources as possible. Through specific grid software, a large project is decomposed into countless independent and unrelated subtasks, which are then computed by each compute node. Even if a node has problems, not able to return the results in time, does not affect the entire project process, even even if a certain computing node suddenly collapse, the task can be assigned to the tasks of scheduling system to other nodes to continue to complete. It should be said that, from this point of view, job scheduling is the core value of grid computing.

Now when it comes to cloud computing, we can immediately think of using the Internet to package all of the data center's resources into service delivery. In general, Grid computing, like cloud computing, constructs all of the resources into a huge pool of resources, but the grid calculates a resource that is provided out of the way to accomplish a particular task. For example, a user might need to request a certain amount of resources from a resource pool to deploy their applications without submitting their tasks to the entire grid.

From this point of view, most of the grid construction to complete a specific task needs, which is the bio-grid, geographic grid, the national education grid and other different grid projects appear. Cloud computing is generally designed for general-purpose applications, and there is no specific grid named for a particular application.

Different ways of allocating resources

For Grid computing, although its resources have been pooled, in the outside world seems to be a huge resource pool. For the user who is submitting a particular task, he does not know which grid's physical nodes will run on. He simply submitted the job task to the grid system in a specific format and waited for the grid to return the result.

Grid job scheduling system automatically find the resources matching the task, and then find out the free physical node, the task assigned to the past until completion. Although the grid can realize the parallel operation of the cross physical machine, it needs the user to write the parallel algorithm first, and to decompose the job into various physical nodes through the dispatch system, which is a relatively complicated process, which is also the reason that many grid computing is built to fulfill the specific requirements.

Professor Jinhai, head of the National education Grid project, said to reporters the current education grid can only be implemented to assign a particular task to a specific node, that is, the grid can monitor a node is idle, and then assign tasks, but if there is no suitable node to achieve the task required to complete the configuration, Then the task must wait for other tasks to release the resource and not take advantage of the distributed configuration.

And cloud computing is through the virtualization of physical machine resources to cut, from this point of view to achieve the resources on demand allocation and automatic growth, and its automatic allocation of resources and increase or decrease can not exceed the physical node itself physical limit. Despite the control, cloud computing sees all IT resources as a resource pool, but the physical opportunities for different chips are grouped into different resource pools.

For example, to echo an application request, and to allocate a x86 CPU or power CPU, Itanium CPU, allocate memory, hard disk space, and then install Linux system, related applications, but not at the same time allocate a x86 CPU and a different CPU to form a heterogeneous environment. Moreover, if the physical machine in the node is the highest number of CPUs is 4, then even if 10 such nodes constitute a 40CPU resource pool, you can not assign a single application 8 CPU virtual node.

From this perspective, Amazon's 2006 launch of the EC2 project (elastic Compute Cloud, the elastic Computing cloud) does count as a cloud computing project, but the concept of cloud computing is not emerging, and the concept of grid computing is in the ascendant, Amazon still uses the concept of grid to push the project out.

Roads

In fact, as many people generally understand, grid computing and cloud computing have a lot in common, both of which can be seen as a derived concept of distributed computing, both to make it resources transparent to users, in order to enable IT resources to achieve better usage.

From the point of view of increasing resource utilization, the concept of Web service is gradually born, then the network company makes a huge computing resource by deploying tens of thousands of servers, which can provide new services that could not be completed before. Companies or individuals can use the computing resources released by large network companies through the Internet to deploy applications or provide services outside. This is the historical process from grid computing to cloud computing.

The significance of both is that both users and enterprise developers can access data through the Internet or to compute, although local resources are limited, but can be complex operations over the network, and its data calculation process for users like the Internet for the local network users, As you can remember the network cloud, the backend implementation is transparent.