0 оценок0% нашли этот документ полезным (0 голосов)
17 просмотров8 страниц
"Grid computing" has emerged as an important new field, distinguished from conventional computing. Authors define 'grid problem' as flexible, secure, coordinated resource sharing among Virtual Organizations. They present an extensible and open grid Architecture, in which protocols, services, application programming interfaces, and software development kits are categorized according to their roles in enabling resource sharing.
"Grid computing" has emerged as an important new field, distinguished from conventional computing. Authors define 'grid problem' as flexible, secure, coordinated resource sharing among Virtual Organizations. They present an extensible and open grid Architecture, in which protocols, services, application programming interfaces, and software development kits are categorized according to their roles in enabling resource sharing.
Авторское право:
Attribution Non-Commercial (BY-NC)
Доступные форматы
Скачайте в формате DOC, PDF, TXT или читайте онлайн в Scribd
"Grid computing" has emerged as an important new field, distinguished from conventional computing. Authors define 'grid problem' as flexible, secure, coordinated resource sharing among Virtual Organizations. They present an extensible and open grid Architecture, in which protocols, services, application programming interfaces, and software development kits are categorized according to their roles in enabling resource sharing.
Авторское право:
Attribution Non-Commercial (BY-NC)
Доступные форматы
Скачайте в формате DOC, PDF, TXT или читайте онлайн в Scribd
“Grid Computing” has emerged as an construction of such an infrastructure, important new field, distinguished from but the “Grid” has also been conflated, at conventional distributed computing by least in popular perception, to embrace its focus on large scale resource sharing, everything from advanced networking to innovative applications, and, in some artificial intelligence. One might wonder cases, high performance orientation. We whether the term has any really a distinct define ‘Grid Problem’ as flexible, “Grid problem” and hence a need for secure, coordinated resource sharing new “Grid Technologies”? among dynamic collections of individuals, institutions, and resources- Grid concept is indeed motivated by a what we refer to as Virtual real and specific problem and that there Organizations. In such settings, we is an emerging, well-defined grid encounter unique authentication, technology base that addresses authorization, resource discovery, and significant aspects of this problem. In other challenges. It is the class of the process, we develop a detailed problem that is addressed by Grid architecture and roadmap for current and technologies. Also we present an future grid technologies. Furthermore, extensible and open Grid Architecture, we assert that while grid technologies in which protocols, services, application are currently distinct from other major programming interfaces, and software technology trends, such as internet, development kits are categorized enterprise, distributed, and peer to peer according to their roles in enabling computing, these other trends can benefit resource sharing. We describe the significantly from growing into the importance of defining a compact set of problem space addressed by green inter grid protocols to enable technologies. interoperability among different grid systems. Finally, we discuss row grid The real and specific problem solving technologies related to other hat underlies the grid concept is contemporary technologies, including coordinated resource sharing and enterprise integration, application problem, multi-institutional virtual service provider, storage service organizations. The sharing that we are provider, and peer to peer computing. concerned with is not primarily file exchange but rather direct access to 1. INTRODUCTION computers, software, data, and other The tern “GRID” was coined in the mid resources, as is required by a range of 1990s to denote a proposed distributed collaborative problem-solving and computing infrastructure for advanced resource brokering strategies emerging science and engineering. Considerable in industry, science, and engineering. This sharing is, necessarily, highly protocols and services that support controlled, with resource providers and secure remote access to computing and consumers defining clearly and carefully data resources and the co-allocation of just what is shared, who is allowed to multiple resources; information query shared , and the conditions under which protocols and services that provide sharing occurs. A set of individuals configuration and status information and/or institutions defined by such about resources, organizations, and sharing rules from what we call a services; and data management services VIRTUAL ORGANIZATION (VO). that locate and transport dataset between storage systems and applications. The following are examples of VOs: the Because of their focus on dynamic, application service providers, storage cross-organizational sharing, Grid service providers, cycle providers technologies complement rather than etching particular, we see a need for compete with existing distributed highly flexible sharing relationships, computing systems can use Grid ranging from client-server to peer-peer; technologies to achieve resource sharing for sophisticated and precise levels of across institutional boundaries; in the control over how shared resources are ASP/SSP space, Grid technologies can used, including fine-grained and multi- be used to establish dynamic markets for stakeholder access control, delegation, computing and storage resources, hence and application of local and global overcoming the limitations of current policies; for sharing of varied resources, static configurations. ranging from programs, files, and data to computers, sensors, and networks; and 2. THE EMERGENCE OF VIRTUAL for diverse usage modes, ranging from ORGANIZATIONS single user to multi-user and from Consider the following three scenarios: performance sensitive to cost-sensitive 1. A company needing to reach a and hence embracing issues from of decision on the placement of a service, scheduling, co-allocations, and new factory invokes a accounting. Current distributed sophisticated financial computing technologies do not address forecasting model from an ASP, the concerns and requirements just providing with access to listed. appropriate proprietary historical data from a corporate database It is here that Grid technologies enter the on storage systems operated by a picture. Over the past five years, SSP. During the decision-making research and development efforts within meeting, what-if scenarios are the Grid community have produced run collaboratively and protocols, services, and tools that interactively, even though the address precisely the challenges that division heads participating in arise when we seek to build scalable the decision are located in VOs. These technologies include different cities. The ASP itself security solutions that support contract with a cycle provider for management of credentials and policies additional “oomph” during when computations span multiple particularly demanding institutions; resource management scenarios, requiring of course that cycles meet desired security resources. For example, members of and performance requirements. a consortium may provide access to 2. An industrial consortium formed specialize software and data and/or to develop a feasibility study for pool their computational resources. a next generation supersonic aircrafts undertakes a high 3. THE NATURE OF GRID accurate multi-disciplinary ARCHITECTURE simulation of the entire aircraft. It identifies fundamental system This simulation integrates components, specifies the purpose proprietary software component and functions of these components, developed by different and indicates how these components participants, with each interact with one another. In defining component operating on that Grid architecture, we start from the participant’s computer and perspective that effective VO access to appropriate design operation requires that we able to databases and other data made establish sharing relationships available to the consortium by its among any potential participants. member. Interoperability is thus the central 3. A crises management team issue to be addressed. In a networked responds to a chemical spill by environment, interoperability means using local weather and soil common protocols. Hence, grid model to estimate the spread of architecture is first and foremost the spill, determining the impact Protocol architecture, with protocols based on population location as defining the basic mechanisms by well as geographic features such which VO users and resource as rivers and water supplies negotiate, establish, manage, and creating a short term migration exploit sharing relationships. A plan, and tasking emergency standard –based open architecture response personnel by planning facilitates extensibility, and coordinating evacuation, interoperability, portability, and code notifying hospital and so forth. sharing: standard protocols make it These examples differ in many easy to define standard services that aspects, the number and type of provide enhanced capabilities. We participants, the type of activities, can also construct Application duration and scale of interaction, and Programming Interfaces and sources being shared. In each case, a Software Development Kits to number of mutually distrustful provide the programming participants with varying degrees of abstractions required to create a prior relationships want to share usable Grid. Together, this resources in order to perform some technology and architecture task. Furthermore, sharing is about constitute what is often termed more than simply document middleware, although we avoid that exchange (as in “virtual term here due to its vagueness. enterprises”): it can involve direct access to remote software, 4. GRID ARCHITECTURE computers, data, sensors, and other DESCRIPTION In specifying the various layers of • Catalogs: This specialized the grid architecture, we follow the form of storage resource principles of the “hourglass model”. requires mechanisms for The narrow neck of the hourglass implementing catalog query defines a small set of core and update operations. abstractions and protocols, onto which many different high level behaviors can be mapped, and which themselves can be mapped onto many different underlying technologies. By definition, the APPLICATION APPLICATION number of protocols defined at the neck must be small. In our architecture, the neck of the hourglass consists of Resource and COLLECTIVE Connectivity protocols, which facilitate the sharing of individual support for advance reservations RESOURCE make it possible for higher level services to aggregate resources in interesting ways that would otherwise be impossible to achieve. TRANSPORT CONNECTIVITY The following list provides resources: INTERNET • Computational Resources: Mechanisms are required for FEBRIC starting programs and for LINK monitoring and controlling the execution of the resulting processes. Fig: The layered architecture and its • Storage Resources: relationship to the internet protocol Mechanisms are required for architecture. putting and getting files. • Network Resources: 1. Fabric: Interfaces to Local Control Management mechanisms Fabric components implement the local, that provide control over the resource-specific operations that occur resources allocated to on specific resources (whether physical network transfers can be or logical) as a result of sharing useful. operations at higher levels. There is thus • Code Repositories: This a tight and subtle interdependence specialized form of storage between the functions implemented at resource requires the Fabric level, on the one hand, and the mechanisms for managing sharing operations supported, on the versioned source and object other. Richer Fabric functionally enables code. more sophisticated sharing operations; at the same time’ if we place few demands on Fabric elements, then deployment of Two primary classes of Resource layer Grid infrastructure is simplified. protocols can be distinguished: 1. Information protocols 2. Connectivity: Communicating Easily 2. Management protocols and Securely A small and mostly standards-based set The connectivity layer defines core of protocols is adopted communication and authentication • A Grid Resource protocols required for Grid-specific Information Protocol network transactions. Communication (GRIP, Currently Based protocols enable the exchange of data On The Lightweight between Fabric layer resources. Directory Access Authentication protocols build on Protocol: LDAP) is used communication services to provide to define a standard cryptographically secure mechanisms for resource information verifying the identity of users and protocol and associated resources. Communication requirements information model. include transport, routing and naming. • The HTTP-based Grid While alternatives certainly exist, we Resource Access and assume here that these protocols are Management (GRAM) drawn from the TCP/IP protocol stack: protocol is used for specifically, the Internet (IP and ICMP), allocation of transport (TCP, UDP), and application computational resources (DNS, OSPF, RSVP, etc.) layers of the and for monitoring and Internet layered protocol architecture. control of computation on Authentication solution for VO those resources. environments should have the following • An extended version of characteristics: File Transfer Protocol, 1. Single sign on GridFTP, is a 2. Delegation management protocol for 3. Integration with various local security data access; extensions solutions include use of 4. User based trust relationships Connectivity layer security protocols, partial file access, and 3. Resource: Sharing Single Resources management of The resource layer build on Connectivity parallelism for high speed layer communication and authentication transfers. protocols to define protocols (and APIs • LDAP is also used as a and SDKs) for the secure negotiations, catalog access protocol. initiation, monitoring, control, accounting, and payment of sharing operations on individual resources. 4. Collective: Coordinating Multiple Resource layer implementations of these Resources protocols call Fabric layer functions to While the Resource layer is focused access and control local resources. on interactions with a single resource, the next layer in the architecture contains protocols and organizations to interoperate and services (APIs and SDKs) that are exchange or share resources. Resources not associated with any one specific that speak these protocols can be said to resource but rather are global in be “on the grid nature and capture interactions across collections of resources. 6. RELATIONSHIPS WITH OTHER 5. Applications TECHNOLOGIES The final layer in our Grid The concept of controlled, dynamic architecture comprises the user sharing within VOs is so fundamental applications that operate within a VO that we might assume that Grid-like environment. Applications are technologies must surely already be constructed in terms of, and by widely deployed. In practice, however, calling upon, services defined at any while the need for these technologies is layer. At each layer, we have well- indeed widespread, in a wide variety of defined protocols that provide access different areas we find only primitive to some useful services: resource and inadequate solutions to VO management, data access, resource problems. Current distributed computing discovery, and so forth. At each approaches do not provide a general layer, APIs may also be defined resource-sharing framework addresses whose implementation (ideally VO requirements. Grid technologies provided by third-party SDKs) distinguish themselves by providing this exchange protocol messages with the generic approach to resources sharing. appropriate service(s) to perform This solution points to numerous desired actions. opportunities for the application of Grid Technologies. 5. “ON THE GRID”: THE NEED FOR INTEGRID PROTOCOLS 1. World Wide Web Grid Architecture establishes The ubiquity of web technologies makes requirements for the protocols and APIs them attractive as a platform for that enable sharing of resources, constructing VO systems and services, and code. It does not otherwise applications. However, while these constrain the technologies that might be technologies do an excellent job of used to implement these protocols and supporting the browser-client-to-web- APIs. In fact, it is quite feasible to define server interactions that are the multiple instantiations of key Grid foundation of today’s Web, they lack architecture elements. The long-term features required for the richer success of Grid computing requires that interaction models that occur in VOs. we select and achieve widespread deployment of one set of protocols at the 2. Application and Storage Service connectivity and resource layers-and, to Providers a lesser extent, at the collective layer. Application service Providers, storage Much as the core Internet protocols service providers, and similar hosting enable different computer networks to companies typically offer to outsource interoperate and exchange information, specific business and engineering these Intergrid protocols enable different applications(in case of ASPs) and storage(in case of SSPs).A customer computing is about controlled negotiates a service level agreement that sharing. Resource owners will defines access to a specific combination typically want to enforce policies of hardware and software. that constrain access according to group membership, ability to pay, 3. Enterprise Computing Systems and so forth.Hence, accounting is Enterprise development technologies important, and Grid architecture such as CORBA, Enterprise Java Beans, must incorporate resource and Java 2 Enterprise Edition, and DCOM collective protocols for exchanging are all systems designed to enable the usage and cost information, as well construction of distributed applications. as for exploiting this information They provide standard resource when deciding whether to enable interfaces, remote invocation sharing. mechanisms, and trading services for discovery and hence make it easy to • The Grid requires a distributed share resources within a single operating system. organization. However, these In this view, Grid software should mechanisms address none of the specific define the operating system services VO requirements. to be installed on every participating system, with these services providing 4. Internet and Peer-To –Peer for the Grid what an operating Computing system provides for a single Peer-To-Peer Computing and Internet computer: namely, transparency with computing is an example of the more respect to location, naming, security, general(“beyond client-server”) sharing and so forth. modalities and computational structures that we referred to in our • The Grid requires new characterization of VOs.As such they programming models. much in common with Grid Programming in Grid environments technologies. introduces challenges that are not encountered in sequential(or 7. OTHER PERSPECTIVE ON parallel)computers, such as multiple GRIDS administrative domains, new failure The Grid is a next generation Internet.” modes, and large variations in The Grid” is not an alternative to “the performance.However,we argue that Internet”: it is rather a set of additional these are incidental, not central, protocols and services that build on issues and that the basic Internet protocols and services to programming problem is not support the creation and use of fundamentally different. computation-and data-enriched environments. • The Grid makes high- performance computers • The Grid is a source of free superfluous. cycles of free cycles. The hundreds, thousands, or even Grid computing does not imply millions of processors that may be unrestricted access to resources. Grid accessible within a VO represent a significant source of computational also presented both and a framework for power, if they can be harnessed in a a Grid architecture, identifying the useful fashion. This does not imply, principal functions required to enable however, that traditional high- sharing within VOs and defining key performance computers are obsolete. relationships among these different functions.Finally, we have discussed 8. SUMMARY how Grid technologies relate to other This paper has concise statement of the important technologies. “Grid problem”, which we define as a controlled and coordinated resource sharing and resource use in dynamic, scalable virtual organizations. We has