Grid computing is a new IT architecture that produces more resilient and lower cost enterprise information systems. With grid computing, groups of independent, modular hardware and software components can be connected and rejoined on demand to meet the changing needs of businesses.
The grid style of computing aims to solve some common problems with enterprise IT: the problem of application silos that lead to under utilized, dedicated hardware resources, the problem of monolithic, unwieldy systems that are expensive to maintain and difficult to change, and the problem of fragmented and disintegrated information that cannot be fully exploited by the enterprise as a whole.
Grid Computing Defined
The grid style of computing treats collections of similar IT resources holistically as a single pool, while exploiting the distinct nature of individual resources within the pool. To address simultaneously the problems of monolithic systems and fragmented resources, grid computing achieves a balance between the benefits of holistic resource management and flexible independent resource control. IT resources managed in a grid include:
Infrastructure: the hardware and software that create a data storage and program execution environment
Applications: the program logic and flow that define specific business processes
Information: the meanings inherent in all different types of data used to conduct business
Core Tenets of Grid Computing Two core tenets uniquely distinguish grid computing from other styles of computing, such as mainframe, client-server, or multi-tier: virtualization and provisioning.
With virtualization, individual resources (e.g. computers, disks, application components and information sources) are pooled together by type then made available to consumers (e.g. people or software programs) through an abstraction. Virtualization means breaking hard-coded connections between providers and consumers of resources, and preparing a resource to serve a particular need without the consumer caring how that is accomplished.
With provisioning, when consumers request resources through a virtualization layer, behind the scenes a specific resource is identified to fulfill the request and then it is allocated to the consumer. Provisioning as part of grid computing means that the system determines how to meet the specific need of the consumer, while optimizing operation of the system as a whole.
The specific ways in which information, application or infrastructure resources are virtualized and provisioned are specific to the type of resource, but the concepts apply universally. Similarly, the specific benefits derived from grid computing are particular to each type of resource, but all share the characteristics of better quality, lower costs and increased flexibility.
Infrastructure Grid Infrastructure grid resources include hardware resources such as storage, processors, memory, and networks as well as software designed to manage this hardware, such as databases, storage management, system management, application servers, and operating systems.
Virtualization and provisioning of infrastructure resources mean pooling resources together and allocating to the appropriate consumers based on policies. For example, one policy might be to dedicate enough processing power to a web server that it can always provide sub-second response time. That rule could be fulfilled in different ways by the provisioning software in order to balance the requests of all consumers.
Treating infrastructure resources as a single pool and allocating those resources on demand saves money by eliminating under utilized capacity and redundant capabilities. Managing hardware and software resources holistically reduces the cost of labor and the opportunity for human error.
Spreading computing capacity among many different computers and spreading storage capacity across multiple disks and disk groups removes single points of failure so that if any individual component fails, the system as a whole remains available. Furthermore, grid computing affords the option to use smaller individual hardware components, such as blade servers and low cost storage, which enables incremental scaling and reduces the cost of each individual component, thereby giving companies more flexibility and lower cost.
Infrastructure is the dimension of grid computing that is most familiar and easy to understand, but the same concepts apply to applications and information.
Applications Grid Application resources in the grid are the encodings of business logic and process flow within application software. These may be packaged applications or custom applications, written in any programming language, reflecting any level of complexity. For example, the software that takes an order from a customer and sends an acknowledgement, the process that prints payroll checks, and the logic that routes a particular customer call to a particular agent are all application resources.
Historically, application logic has been intertwined with user interface code, data management code, and process or page flow and has lacked well-defined interfaces, which has resulted in monolithic applications that are difficult to change and difficult to integrate.
Service oriented architecture has emerged as a superior model for building applications, and service oriented architecture concepts align exactly with the core tenets of grid computing. Virtualization and provisioning of application resources involves publishing application components as services for use by multiple consumers, which may be people or processes, then orchestrating those services into more powerful business flows.
In the same way that grid computing enables better reuse and more flexibility of IT infrastructure resources, grid computing also treats bits of application logic as a resource, and enables greater reuse of application functionality and more flexibility in changing and building new composite applications.
Furthermore, applications that are orchestrated from published services are able to view activities in a business as a single whole, so that processes are standardized across geography and business units and processes are automated end-to-end. This generates more reliable business processes and lowers cost through increased automation and reduced variability.
Information Grid The third dimension to grid computing, after infrastructure and applications, is information. Today, information tends to be fragmented across a company, making it difficult to see the business as a whole or answer basic questions.about customers. Without information about who the customer is, and what they want to buy, information assets go underexploited.
In contrast, grid computing treats information holistically as a resource, similar to infrastructure and applications resources, and thus extracts more of its latent value. Information grid resources include all data in the enterprise and all metadata required to make that data meaningful. This data may be structured, semi-structured, or unstructured, stored in any location, such as databases, local file systems, or e-mail servers, and created by any application.
The core tenets of grid computing apply similarly to information as they do to infrastructure and applications. The infrastructure grid exploits the power of the network to allow multiple servers or storage devices to be combined toward a single task, then easily reconfigured as needs change. A service oriented architecture, or an applications grid, enables independently developed services, or application resources, to be combined into larger business processes, then adapted as needs change without breaking other parts of the composite application. Similarly, the information grid provides a way for information resources to be joined with related information resources to greater exploit the value of the inherent relationships among information, then for new connections to be made as situations change.
The relational database, for example, was an early information virtualization technology. Unlike its predecessors, the network database and hierarchical database models, in which all relationships between data had to be predetermined, relational database enabled flexible access to a general-purpose information resource. Today, XML furthers information virtualization by providing a standard way to represent information along with metadata, which breaks the hard link between information and a specific application used to create and view that information.
Information provisioning technologies include message queuing, data propagation, replication, extract-transform-load, as well as mapping and cleansing tools to ensure data quality. Data hubs, in which a central operational data store continually syncs with multiple live data sources, are emerging as a preferred model for establishing a single source of truth while maintaining the flexibility of distributed control.
Grid Resources Work Well Independently and Best Together By managing any single IT resource – infrastructure, applications, or information - using grid computing, regardless of how the other resources are treated, enterprises can realize higher quality, more flexibility, and lower costs. For example, there is no need to rewrite applications to benefit from an infrastructure grid. It is also possible to deploy an applications grid, or a service oriented architecture, without changing the way information is managed or the way hardware is configured.
It is possible, however, to derive even greater benefit by using grid computing for all resources. For example, the applications grid becomes even more valuable when you can set policies regarding resource requirements at the level of individual services and have execution of different services in the same composite application handled differently by the infrastructure - something that can only be done by an application grid in combination with an infrastructure grid. In addition, building an information grid by integrating more information into a single source of truth becomes tenable only when the infrastructure is configured as a grid, so it can scale beyond the boundary of a single computer.