📄 rfc1210.txt
字号:
Cerf, Kirstein, & Randell [Page 16]RFC 1210 Network and Infrastructure User Requirements March 1991 provide intercontinental packet video, multimedia conferencing and computer supported collaborative group technology facilities. The workshop should propose actions which could be made the basis of a future harmonised ESPRIT and DARPA/NSF work programme.6.9 Access to Unique Resources A number of resources can be labelled unique in the scope of ESPRIT/DARPA/NSF or even on a worldwide basis. Their uniqueness may derive from their nature (e.g., large test facilities or a focus point of knowledge in a discipline) or be such in a transitory phase. In the spirit of the future EC/US cooperation, it is clear that there should be agreed access to some such resources. This will require: (i) Provision of appropriate access and usage information; (ii) Physical access for visitors; (iii) Continued non-local access. The third point has clear networking implication. Appropriate remote access to the resources, connectivity to the users and adequate access speeds have to be provided, possibly together with access control facilities. The most demanding cases are those of newly developed products; their transitory uniqueness does not allow one to amortise costs over substantial periods as would be reasonable for large scale centres like NCAR or CERN.6.9.1 One Year Target (i) Identify appropriate unique transitory resources (e.g., Touchstone); (ii) Specify the provisions needed to make at least one such resource available.6.9.2 Three Year Target Set up one or more significant transatlantic pilots demonstrating remote, secured access.6.9.3 Recommended Actions Organise a workshop dedicated to analysing the needs and defining the steps required to provide pilot access to one or more specific such resources. The workshop may need to address networking needs,Cerf, Kirstein, & Randell [Page 17]RFC 1210 Network and Infrastructure User Requirements March 1991 security provisions, documentation and advisory requirements, modification of current access capabilities, and usage policies.6.10 Distributed Visualization Scientific visualization applications often involve multiple resources. These resources can span a complete range of sophistication, from simple hardcopy at one end to elaborate rendering at the other end. Interactive graphics workstations, supercomputers and specialized scientific databases may all be involved in a single application. The scientist at a workstation should be able to view all of these resources as a single network resource, although they may be physically distributed over considerable distances. A typical example is a high performance graphics workstation, a supercomputer and a network to connect them together, all with appropriate software. The workstation may be close to the supercomputer or distant from it. Currently there are efforts underway at several installations - including ones funded by NSF/DARPA and ESPRIT - to develop techniques, interfaces and software necessary to create this environment. In limited instances it already exists. Better coordination of these efforts on both sides of the Atlantic would be desirable. Coordinating such efforts across the Atlantic will be necessary for effective collaboration in end-user visualization applications in a variety of disciplines to take place in the future.6.10.1 One Year Targets Identify the significant current development efforts in these areas and determine which ones to support. Identify the areas requiring standards. Minimize duplication of effort and begin to distribute the techniques and software.6.10.2 Three Year Targets Establish mutually agreed upon standards. Demonstrate transatlantic distributed visualization applications.6.10.3 Recommended Actions Establish a working group to further refine and to implement the one year and three year targets and to identify additional distributed visualization topics that would benefit from coordinated efforts. Determine the appropriate mechanisms for supporting such collaborations.Cerf, Kirstein, & Randell [Page 18]RFC 1210 Network and Infrastructure User Requirements March 1991UNDERLYING SERVICES Most of the services described below are required to achieve the goals of reliability, availability and transparency of the user services.6.11 Network Management Current network management technology and practice are not adequate to support large scale, international research networks. Time-zone differences and lack of organizational operational network management agreements combine to make international network management a serious challenge. To be effective, network management must operate on a campus-to-campus basis, since the campuses are the sources and sinks of traffic in the system.6.11.1 One Year Target Put in place an administrative structure to coordinate existing facilities manually and to plan technical solutions.6.11.2 Three Year Target Develop and deploy technology for automating international network management.6.11.3 Recommended Actions (i) Convene an international research network operations, planning and management team to develop and apply procedural and technical recommendations for international network management; (ii) Organize a set of international network operations centres devoted to configuration management, fault detection, isolation and repair of network problems; (iii) Form one or more intercontinental Computer Emergency Response Teams to coordinate response to attacks against hosts and networks and to develop procedures for collecting actionable evidence.6.12 Multi-protocol Support Users depend on a variety of protocols to support their research. The international network infrastructure does not uniformly support the use of multiple protocols (e.g., DECNET, TCP/IP/ST, OSI) on an end-to-end basis. The use of various portions of the internationalCerf, Kirstein, & Randell [Page 19]RFC 1210 Network and Infrastructure User Requirements March 1991 network also may be restricted by policy, and this must be accommodated in implementing routing for campus-to-campus protocols. Support for campus-to-campus multi-protocol transmission and routing is needed at a minimum of 64 Kb/s end-to-end - higher for the support of some of the services. Where the end-users have adopted similar protocols, the intervening networks should not impede the full exploitation of the facilities available in the chosen protocol suite. Where different protocol suites are used, high quality application-level gateways which can translate among protocols are needed also; to the greatest extent possible, these should allow people to use their own procedures, even though they are communicating with services which use different ones. For some services, this will lead to a requirement to upgrade access, and possibly even transparent access (including protocol conversion), to at least 1.5 Mb/s between individual campuses in the US and Europe.6.12.1 One Year Targets (i) Support campus-to-campus communication for a subset of coexisting protocol suites (at least OSI and TCP/IP) at a minimum of 64 Kb/s; (ii) Deploy internationally supported versions of existing application level (protocol-translating) gateways.6.12.2 Three Year Targets (i) Improve management and resource allocation for multi-protocol routers (e.g., to achieve service guarantees); (ii) Support campus-to-campus communication at a minimum of 1.5 Mb/s.6.12.3 Recommended Actions (i) Validate current multi-protocol solutions for intercontinental, and indeed campus-to-campus use; (ii) Collaborate on research and experimentation with multi-protocol routing and resource allocation; (iii) Make recommendations, to funders and national research network service providers, on technical solutions and standards for multi-protocol support.Cerf, Kirstein, & Randell [Page 20]RFC 1210 Network and Infrastructure User Requirements March 19916.13 Client-Server Technology Among the more important computer communications techniques emerging on a widespread basis during the last decade is the client-server model of interprocess communication. This notion was actually developed during the earliest stages of packet network exploration and dramatically enhanced with the invention of local area networks (such as Ethernet) which could support very high speed, low delay inter-computer exchanges. Applications of this concept range from remote procedure calls to remote file access and support for remote, bit-mapped graphics. At present, these techniques work best in a high bandwidth, low delay environment; they are generally not well-supported in wide-area, intercontinental networks. Collaborative efforts between the US and Europe could be enhanced substantially by support for client-server services on an intercontinental basis. Such facilities would permit collaborative use of distributed filing systems, X-windows applications and other distributed computing applications. High capacity, low-delay channels will be needed on an intercontinental basis to support serious use of this technology. In addition, agreement must be reached on which protocols should be used to support this technology.6.13.1 One Year Targets (i) Provide limited bandwidth intercontinental X-Windows support for graphical user interfaces; (ii) Achieve agreements on intercontinental Remote Procedure Call and Distributed File System protocols; (iii) Validate support of X-Windows under OSI and through protocol translating gateways.6.13.2 Three Year Targets (i) Achieve selective support for intercontinental remote visualization; (ii) Achieve support for intercontinental RPC and Distributed File Systems.6.13.3 Recommended Actions (i) Convene workshops to achieve agreements on intercontinental Remote Procedure Call and Distributed File System protocols;Cerf, Kirstein, & Randell [Page 21]RFC 1210 Network and Infrastructure User Requirements March 1991 (ii) Form working group on support for X-Windows in OSI and to validate performance through TCP/TPn protocol translating gateways; (iii) Initiate collaboration on implementation and test of intercontinental RPC and distributed file systems.Section 6.14 Archival Storage for Distributed Computing Environments There are several major issues that must be addressed by distributed computing environments (DCEs) containing supercomputers. Resolution of these issues is likely to evolve over the next five to ten years. One such issue is archival storage and bitfile management for the
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -