What Is Grid Computing
What Is Grid Computing?
The term “Grid” was coined in the mid-nineties. The real and specific problem underlying the Grid concept is coordinated resource sharing within a dynamic and multi-institutional virtual organization (Virtual Organization, briefly denoted by VO). The sharing is not limited to the exchange of files, but extends to direct access to computer software, in general all the necessary hardware to solve a scientific problem, or industrial engineering.
Individuals and institutions that provide their resources in the grid for the same purpose, are part of the same VO. Common feature of Grid projects is the need for a data-intensive computing environment within which applications have the need to access large amounts of geographically distributed data quickly and reliably and it is precisely the burden of Grid, such applications to operate in the best possible way.
It is easy to see that no computer currently on the market would be able, alone, to draw up similar amounts of data in a reasonable time, but the sharing of resources such as CPUs and disks properly coordinated may give the user access to a supercomputer virtual, with an incredible computing power and storage capacity capable of handling large workloads. From idea to portray the whole architecture of a Grid as a single virtual supercomputer, hiding all the complexity inside the user and showing only the benefits, the need arose to design and build a resource scheduler Resource Broker.
Resource Broker: It is one of the critical components of the system of resource management. It is responsible for allocating resources to the job (Gridlet) to meet the needs of application and system. The resources it needs to track and manage include: computing and data storage systems (using the Storage Broker, interconnection network, and through the Network Monitor).
The scheduling is a traditional field of information technology, but despite many techniques have been studied for many types of systems (from uniprocessor to multiprocessor distributed systems), the typical characteristics of data grids make many of these approaches inadequate.
In fact, while in traditional systems of resources and jobs are under the direct control of the scheduler, the resources are geographically distributed grids. These are heterogeneous in nature and belong to different individuals or organizations, each with their own scheduling policies, different cost models of access, workload and availability of resources varies dynamically over time.
The lack of centralized control, with the presence of users that generate jobs (Gridlet), very different from each other, make scheduling more complicated than that of traditional computing systems.
Grid Computing: History
The sharing of computing resources allows us to divide the history of computing in 4 eras. The first was characterized by the idea of a single computer for many users. This was the purchase of computing resources has cost so large that the problem is that the concurrent use of these resources by different users. Since the ’80s hardware costs have suffered declines that enabled a computer to each user. Born in this period, the first personal computers and computing infrastructure is evolving toward the SIMD.
Since the end of the ’80s are starting to spread the idea of sharing the hardware architecture even said thanks to falling prices, leading to the birth of the first virtual parallel machines. The 90s are the ones during which you apply Moore’s Law in its entirety and affirm the computer networks and the Internet (basic concepts for the Grid)
Evolution of Grid computing
The SETI @ home project, launched in 1999 by Dan Werthimer, is a well-known example of a project, albeit simple, Grid computing. SETI @ Home has been followed by many other similar projects in mathematics and microbiology.
Currently, the most important European grid is that CERN is now called EGEE (gLite middleware is the name of producing; previously before DataGrid and LCG), developed by – among others – by a team of Italian-Czech and mainly INFN, National Institute of Nuclear Physics.
Unlike that used by SETI @ Home, a grid is currently conceived providing a level of middleware between the memory and computing resources (CE – Computing Element and SE – Storage Element) and users of the grid itself.
The main purpose of middleware is to make the so-called match-making, namely the coupling between the resources required and those available to ensure the distribution of jobs (the term used to describe a batch system or a part thereof) in better condition, having always been the vision of the entire grid.
Another important phenomenon to be noted is the emergence, alongside the great national and international Grid, multiple implementations at local or metropolitan distributed systems that maintain the characteristics of a GRID. These systems are indicated by the term Local Area Grid (LAG) and Metropolitan Area Grid (MAG) or, more simply, Metropolitan Grid with clear reference to the classification introduced in the network.
As the coordination of national grid provides further set up a World Wide Grid, local implementations or Metro Grid approach to the world of the Intranet. Indeed, they provide a kind of utility that can be used simply to introduce the Internet distributed computing across the enterprise.
The body of reference for the development of uniform standards and protocols used by the grid is GGF (Global Grid Forum), which created the standard OGSA (Open Grid Services Architecture). In 2004 he was issued WSRF (Web Services Resource Framework), which is a set of specifications to help programmers write applications able to access the Grid resources.
Today the most famous and used software is BOINC, a software for grid computing developed by the University of California (Berkeley). The acronym stands for Berkeley Open BOINC Indeed Infrastructure for Network Computing. This software is open source.
Applications of Grid computing
An example of the paradigm of Grid computing is neuGRID, a project of the 7th Framework Program involving the development of infrastructure for the study of neuro-degenerative diseases.
GridSim
It was prepared a graphical interface that allows the user to input the characteristics of the Grid system, which analyzes the behavior, and present each time the graphic reconstruction.
At the end of the first phase, namely the incorporation of features, you start the second phase on the simulation. During the simulation data are processed and the user is presented with the report all information and system responses. Simulation uses the mortgage GridSim while the graphical representation of system is used JUNG. JUNG (Java Universal Network / Graph Framework) is a library of open source modeling and visualization of graphs, written in Java.
Study: From Wikipedia, the free encyclopedia. The text is available under the Creative Commons.
