Peng: Grid computing and cloud computing

Grid is the next generation of Internet core technology developed in the mid 90. The founder of Grid technology, Ian Foster, defines it as "collaborative sharing of resources and solving problems in a virtual organization that is dynamic and involved in multiple organizations". Grid is on the basis of the network, based on service-oriented architecture (SOA, service-oriented architecture), using interoperability, on-demand integration and other technical means, will be dispersed in different geographical resources virtual into an organic whole, to achieve computing, storage, data, Software and equipment, such as the sharing of resources, thereby greatly increasing the utilization of resources, so that users get unprecedented computing and information capabilities.
The international grid community is committed to grid middleware, grid platform and grid application construction. As far as grid middleware is concerned, the well-known foreign grid middleware has Globus Toolkit, Unicore, Condor, glite and so on, in which Globus Toolkit has been widely adopted; on the grid platform, the internationally renowned grid platform has TeraGrid, Egee , Coregrid, D-grid, Apgrid, GRID3, gig, etc. The American TeraGrid is an ultra large open scientific research environment funded by the National Science Foundation program. TeraGrid integrates high-performance computers, data resources, tools, and high-end experimental facilities. Currently, TeraGrid has integrated more than 750 trillion computing power per second, 30PB of data, and has more than 100 domain-oriented grid application environments. The European Union e-science facilitates grid egee (enabling Grids for e-science), and is another super large, grid-oriented computing infrastructure for many domains. More than 120 institutions have been involved, including 250 grid sites in 48 countries, 68,000 CPUs, 20PB of data resources, 8,000 users per day, and an average of 30,000 operations daily, with a peak of more than 150,000 jobs; in the case of grid applications, There are hundreds of well-known grid applications, including atmospheric science, Forestry, marine science, environmental Science, bioinformatics, medicine, physics, astrophysics, Earth Science, astronomy, engineering, social behavior, etc.
China has 863 supported China National Grid (cngrid,863-10 theme) and China spatial Information grid (SIG,863-13 theme), China Education Research Grid (CHINAGRID), supported by the Ministry of Education in the 15 period, Shanghai supported Shanghai Grid (Shanghaigrid) and so on. China's National Grid has 10 nodes, including Hong Kong, with an aggregate computing capacity of 18 trillion, and currently has 408 users and 360 applications. The Chinese educational research grid Chinagrid connects the computing facilities of 20 colleges and universities,More than 3 trillion operations per second, the development and implementation of biological information, fluid dynamics, such as five of the grid of the typical applications. During the 25 period, the state supported the grid more strongly, and supported the grid technology through 973 and 863, Natural Science Fund and other ways. 973 plan has "the basic theory of the semantic grid, the model and the method research" and so on, 863 plan has "the high efficiency computer and the grid service environment", "the grid geographical information system software and the important application" and so on, the National Natural Science Fund major research plan has "the network computation application support middleware
Just as cloud computing can be divided into IaaS, PAAs, and SaaS three types, grid computing can be divided into three types: computational grids, information grids, and knowledge grids. The computational grid goal is to provide a virtualized computing infrastructure that integrates a variety of computing resources. The goal of information grid is to provide an integrated intelligent information processing platform, integrate various information systems and information resources, and eliminate the islands of information, so that users can obtain the precise information after integration, that is, the service on Demand and step services (one Click is enough). Knowledge Grid Research integrates intelligent knowledge processing and understanding platform, so that users can easily publish, process and acquire knowledge.
What needs to be explained is that the current understanding of the grid exists a misconception that only the use of well-known grid middleware such as Globus Toolkit is the grid. We believe that as long as the grid concept, the distribution of a certain range of heterogeneous resources into an organic whole, to provide resources sharing and collaborative work services platform, can be considered as a grid. This is because, because the grid technology is very complex, there must be a process from norm to normalization, should recognize the objectivity of the existence of difference. Although the grid community has been working from the outset to construct a fully interoperable environment, however, due to the grid in the forefront of information technology, many areas have not yet stereotypes, published individual specifications too complex to cause ease of use, the existing grid system for specific applications to apply, personalized framework design and implementation technology, The difficulty of interoperability between grid systems is the reason why the Open Grid forum OGF (open Forum) proposes a plan gin (grid interoperation now) to establish interoperability mechanisms for different grid systems. On the other hand, while the plan to build a global unified grid platform has a long way to go, it does not preclude grid technology from playing an important role in a variety of specific application systems. The relationship between grid computing and cloud computing is shown in the following table.

Table 1 comparison between grid computing and cloud computing

Grid computing

Cloud

Target

Sharing high performance computing power and data resources, realizing resource sharing and collaborative work

Provide common computing platform and storage space, provide various software services

Source of resources

Different institutions

Same institution

Resource type

Heterogeneous resources

Homogeneous resources

Resource node

Computer

Server pc

Virtualization View

Virtual Organization

Virtual machine

Calculation type

The main problem of tight coupling

Loose coupling problem

Type of application

Scientific calculation is the main, computation is dense

Data processing mainly

User type

Scientific

Commercial Society

Payment method

Free (government funded)

Billing by volume

Standardization

There is a unified international standard OGSA/WSRF

There is no standard yet, but there is already an open cloud computing Alliance OCC

Grid computing goes in the way of the College school: having debated the concept for many years, three bones in architecture, and a great deal of effort on the standard specification, the goal is very ambitious--to share resources and collaborate to solve problems across platforms, across organizations, across trusted domains in extremely complex heterogeneous environments, The resources to be shared are also diverse-from high-performance computers, databases, devices to software, and even knowledge; cloud computing is a reality: for the moment, regardless of the concept, regardless of the standard, Google Cloud computing and Amazon Cloud computing is very different, Cloud computing is just a new common buzzword for what they've done before; the shared storage and computing resources are temporarily limited to a single enterprise, eliminating many issues of organization-wide coordination; Cloud computing, represented by Google, is as simple as its interface in its internal management, and it saves its functions, Google's file system does not even allow the modification of existing files, greatly reducing the difficulty of implementation, but with its unparalleled scale effect to release unprecedented energy.

The relationship between grid computing and cloud computing is like the relationship between OSI and TCP/IP: The ISO-developed OSI (Open Systems Interconnection) network standard is thoughtful and complex, taking into account the conversation layer and presentation layer issues for many years. Very far-sighted, but too spring, the difficulty and the cost of implementation is very large. When a simplified version of OSI--TCP/IP out, the seven-layer protocol was simplified to four layers, and the content was greatly streamlined, making it a quick success. After TCP/IP eminence for many years, the semantic web and other issues have been put on the agenda, beginning to make up for TCP/IP lessons, increase its ability to speak and express. Therefore, the OSI is the academic school, TCP/IP is the reality faction. OSI is the basis of TCP/IP, and TCP/IP drives the development of OSI. It is not the question of being King and loser, but the problem of rolling development.

As early as 2002, Peng, author of the book, pointed out the practical problem of traditional grid computing, and proposed the concept of Grid computing pool (Computing pool) [1][2], which is consistent with today's cloud computing. "Traditionally, people want to" accumulate "the computational power of several supercomputers in a grid to create an ' unprecedented ' virtual supercomputer. So far, however, these projects have been experimental, proprietary, and unlikely to engage in long-term, practical operations against common tasks. The reason is that most of the computational tasks cannot be divided into sub tasks that do not communicate or communicate with each other, so there is frequent communication between subtasks. Regardless of how well the network conditions are, communication bandwidth and latency between remote supercomputers (caused by long-distance transport and TCP/IP protocols) cannot be compared to the super computer internal bus and the system area network SAN. ...... Since the current stage of network conditions and research level can not effectively support the cross node distribution operations (this is the internationally conceived computing grid), limited to three points: (1) computing resource sharing (n local or off-site high-performance computer to join this system). Can greatly improve the utilization of resources and quality of service. (2) Do not break down a task into n subtasks, but simply schedule it to run on one of the appropriate machines. (3) Submit tasks and view results through the web. "In other words, grid computing pool" to the decentralized High-performance computers with high-speed network connectivity, with specially designed middleware software organically bonded together, the Web interface to accept the scientific workers from all over the calculation request, and assign it to the appropriate node to run. The computing pool can greatly improve the service quality and utilization of resources, and avoid the inefficiency and complexity of dividing the application across nodes, and can meet the practical requirements under the current conditions. ”

Without the foundation of Grid computing, cloud computing will not come so soon. Cloud computing is a simplified, practical version of Grid computing, unlike Grid 2.0, which is like grid 0.2. Grid 0.1 refers to the previous implementation of scientific research-oriented grid, attaches great importance to standard norms, but also very complex, but lack of successful business model. Cloud computing is a simplified form of grid computing, and the success of cloud computing is the success of the grid. Grid not only to integrate heterogeneous resources, but also to solve a lot of technical coordination problems, and not like cloud computing has a successful business model, so it is much more difficult to achieve than the cloud. But for many high-end scientific or military applications, cloud computing is unable to meet demand, and must rely on the grid to solve.

At present, many people claim that grid computing has failed, and cloud computing has replaced it, which is an illusion. Grid computing has been more than 10 years old, and it is normal to be as compelling as it was when it first arose. In fact, some government-led, narrowly-targeted grids have achieved decisive victories. Typical: TeraGrid in the United States and egee in Europe, these grids have hundreds of thousands of jobs each day. The main battlefield of future scientific research will be based on grid computing. Militarily, the U.S. military's global information grid gig has included more than 7 million computers, larger than any existing cloud computing hub.

It is believed that in the near future, the Business 2.0 built on cloud computing and the Science 2.0 built on grid computing will be successful.

[1] Peng. This paper presents a practical method of grid implementation-Grid computing pool model, 2002-11-27, http://www.chinagrid.net/show.aspx?id=1672&cid=57

[2] Peng Liu, Yao Shi, San-li Li, Computing pool--a simplified and practical Grid Model, computational Second ional Workshop on Grid and cooperative Computing (GCC 2003), Full-circle, Dec 7-10, 2003, published into lecture Notes in Computer Science (LNCS), vol. 3032, Heidelberg:springer-verlag, 2004.http://www.chinagrid.net/show.aspx?id=1915&cid=48

Author: Peng, Source: China Cloud Computing (http://www.chinacloud.cn)

(This article belongs to Freedocs, welcome reprint, but need to include the above hint information)