RAMS: A Resource Allocation Management System
A Case for Support under the LINK HPIP Programme
Contributors:
J. Crowcroft, Z. Wang, J. Cowan, D. Lewis (University College London)
Keith Shipton, Tim Carmichael (Communic8 Software Europe)
Greg Brougham (Tadlon)
Ken Brown (Texaco)
In the last few years the Internet has experienced phenomenal growth. Commercial companies now account for more than half the users of the network and increasingly they are using the Internet protocols in their internal networks. Demand for network resources is outstripping availability. This raises questions about how the finite resources in networks should be allocated. A major problem is ensuring that critical business applications receive appropriate resources so that they can operate reliably and efficiently.
This case for support from the LINK scheme outlines a novel programme of research aimed at addressing just such a problem. It will first analyse the data traffic patterns of applications in operational commercial networks that use the Internet protocols. The programme will then use the results of this analysis to develop a network management system that will distribute resources in such a way that key applications can receive priority without impairing the performance of other applications.
The consortium assembled for this proposal provides the appropriate skill set required both to perform the research, field trials and subsequently to exploit the results of the project. University College London's Computer Science Department, a leading UK Internet research centre, will design and implement a Resource Allocation Management System prototype. This development will be performed in conjunction with Communic8, a distributed applications developer, who will enhance their products so that they interact with the RAMS prototype. Tadlon, a small software consultancy, will, with assistance from UCL, perform an analysis of three networks, to determine how the applications use the network both before and after the prototype is deployed. Communic8, Tadlon and UCL will collaborate in exploiting the results of the project.
The consortium consists of the following organisations:
University College London, Computer Science Department: A centre of excellence in Internet, network management and distributed multimedia research.
Communic8 Software Europe: A medium sized developer of distributed applications.
Tadlon: A small consultancy specialising in providing network and distributed application support to City firms.
Texaco: A large oil company with an extensive corporate network.
This consortium provides a set of complementary skills appropriate both for tackling the research problems and exploiting the research at the end of the project. Tadlon has the expertise to determine how effectively the Resource Allocation Management System improves the performance of key real-time applications. UCL and Communic8 will combine their expertise in distributed system development to develop a prototype to demonstrate the System. Texaco have a large corporate network, which will be analysed to provide a picture of traffic conditions of a large and complex network. Communic8 will also make their own corporate network available for analysis and integration work.
UCL will be the coordinating partner in the project taking advantage of the considerable experience of its personnel in managing collaborative projects between academia and industry. Project coordination meetings between all partners will be held every month for the duration of the project.
As distributed applications become more crucial to corporate competitiveness, network congestion becomes a more serious issue. The proposal addresses this problem, but does so in a way that accounts for commercial requirements from the outset, giving network administrators the degree of visibility and control over their installations that they demand.
The aim of this project is to examine the interaction between distributed applications and the network, and to develop a resource allocation management system which can be used to ensure that critical applications have the network resources they need to operate efficiently. This system will operate by providing sources of network traffic information on the state of the network using standard network management protocols. Two methods of controlling the sources of network traffic will be investigated. One method will be through the manipulation of the protocol stack software on individual machines, controlled to a resolution that affects the traffic injected into the network by different applications. The other approach will involve controlling the application directly in order to change the volume of network traffic it generates. The resource allocation management system will be demonstrated by the development of a prototype. This development will be performed jointly by UCL and Communic8, UCL applying its expertise in the Internet protocol suite and network management and Communic8 integrating the system with its Z/ftp product.
Since the applicability of this system to a commercial environment is seen as essential by the partners, the prototype development will be preceded by an analysis of the relationship between network load and application usage at operational corporate networks. This analysis will require the application of Tadlon's knowledge of distributed network applications and will be performed over Texaco's and Communic8's corporate networks and possibly those of some Tadlon clients. The applicability of the system to commercial environments will be assessed further towards the end of the project by trials at UCL and Communic8 that will monitor the performance of the prototype and which will then be compared to the network characteristics derived from the initial analysis. The results of the initial and final analyses will be presented to project partners both in written form and as a seminar aimed at user site personnel. Further dissemination of results will be made by UCL, subject to IPR and commercial confidentiality agreement, through publication of papers. UCL is a regular contributor to the Internet Enginnering Task Force and hopes to use this experience in forwarding the results of this project as input to Internet standards.
The emphasis on the commercial applicability of this research will place the project members in an excellent position to exploit the results of the project. It is currently planned to develop the enhancements to Z/ftp that enable it to interact with the prototype to product level through collaboration between Communic8 and UCL. The level of contributions from Communic8 indicates their view of the potential of this approach.
A plan has been established that divides the project work into five activities each with its own deliverable as follows:
1. The initial analysis of user site's network traffic and application usage. This will be conducted by Tadlon with assistance from UCL and Communic8 over a 2 month period and will result in a written report outlining the findings and requirements for the resource allocation management system prototype.
2. The development of the resource allocation management system prototype in parallel with the enhancements to Z/ftp. This will be done by UCL and Communic8 over a six month period and will result in separate demonstrations of the two parts of the prototype.
3. The integration of the UCL prototype and the Communic8 Z/ftp enhancements. This will be conducted over a two month period and will result in a demonstration of the integrated system in operation.
4. The trials of the integrated prototype over the UCL and Communic8 networks which will allow quantifiable assessments of its performance to be made. This will be performed over two months by UCL and Communic8 and will result in a written report comparing this analysis with the initial one.
5. The assessment of the prototype performance and the steps required to exploit it. This will be performed over two months by UCL, Communic8 and Tadlon and will culminate in a seminar and demonstration.
It is strongly felt by the project partners that in this highly competitive field the total time to market should be minimised. The project plan therefore extends over a single twelve month period, resulting in some overlap between the above activities. The total effort required is 41.5 man months: 26 for UCL, 29 for Communic8, 4 for Tadlon and 0.5 for Texaco.
The following table outlines the total costs that will be incurred by each partner during the project, the contribution of each partner to these costs and the public support required.
Partner Total Total Support Support
eligible contributions required from required from
costs DTI EPSRC
UCL 62000 0 0 62000
Communic8 284000 168000 116000 0
Tadlon 40000 20000 20000 0
Texaco 10000 10000 0 0
396000 198000 136000 62000
Total
The public funding required is 50% of total costs.
The importance of data networking is now seen as central to the survival of national economies world-wide. A major phenomenon in recent years has been the explosive growth in the use of the Internet and its TCP/IP suite of protocols by all types of organisations. Research into and exploitation of the Internet has been traditionally led by the United States. This proposal presents an ideal opportunity for public funding to stimulate the transfer of technologies in this field from a UK centre of excellence in Internet research to UK companies.
The Internet is currently undergoing a fundamental transformation as it moves from being a playground for academics into being a commercial service. With more than 50% of the network commercially provided, and more than 50% of the subscribers being businesses, the Internet is now a very different place from what it was in the 1980s. The UK already has a number of UK companies that provide Internet access (Demon, Pipex etc.), but the commercial opportunities for providing software products and services have not yet been fully taken up by UK companies.
The advent of new and innovative applications such as real-time video conferencing and information access tools have enabled the delivery of services previously not possible. The growth rate for the new services are phenomenal by any standards. For example, the MBONE, an Internet-wide multicast testbed is now connected to over 1000 sites and there are often video conferences going on round the clock. The UCL-CS WWW server was only set up a few months ago. There are now 9500 accesses per day, and 240,000 GIF images were shipped to date to nearly 12,000 hosts (0.5% of the entire Internet!) on between 4000 and 5000 networks (more than 10% of the entire Internet!). Increasingly large commercial corporate users are also starting to use TCP/IP for their business and internal communications.
The characteristics of the new applications are different from that of traditional data applications. Some applications are resource-thirsty, and can potentially generate traffic at a very high rate. For example, a video source may easily overwhelm a bottleneck and cause congestion problems for other data applications. Some are highly interactive. The long delay caused by congestion may severely affect the usability of many such applications. Many applications require Quality of Service (QoS) guarantees.
This project will examine the traffic characteristics of applications and develop a resource allocation management system to make efficient use of network resources.
The problem of resource management has attracted considerable attention in the Internet community. In the past several years, there has been much discussion and research in the area of resource reservation, admission control and packet scheduling. The Internet Engineering Task Forces (IETF) has two Working Groups (WGs); The INT-SERV WG is working on a new network service and the RSVP WG is developing a resource reservation protocol. Most of the work, however, has focused on the network side.
In this project, we will focus on the interaction between applications, the network and network management system, and develop a system that can make the most efficient use of network resources.
The project consortium consists of one University (University College London), one software products company (Communic8 Group), one small consultancy firm (Tadlon Ltd) and one large end user company (Texaco Ltd.) Both Communic8 and Tadlon are SMEs.
University College London, Computer Science Department
UCL CS department has been actively involved in Internet research from the early 70's. UCL-CS was the first node outside USA on ARPANET, and is the only place which has received continual US DARPA funding for the last 8 years. There are currently 8 projects in networking and communications: Multimedia Video Conferencing (US DARPA), MICE: Multimedia Integrated Communications Pilots (CEC Esprit), PREPARE: Pre-pilot in Advanced Resources Management (CEC RACE), Multicast and Next Generation Internet Protocol (US Bellcore), EMMA: Enhancing Multicast For Multimedia Applications (UK JISC), FUME: Framework for a Unified Multicast Engine (UK SERC),MMN: Management of Multiservice Networks (UK BT), and ATM Pilot over Superjanet (UK JNT). We are also very active in the Internet Engineering Task Force (IETF), and collaborate with researchers and companies in the US (MIT, LBL, XeroxPARC, Sun and Bellcore). The following personnel will be involved in this project:
Jon Crowcroft is a senior lecturer in the Department of Computer Science, University College London, where he is responsible for a number of European and US funded research projects in Multi-media Communications. A recent project just completed worked on protocol migration (Internet and OSI). He has been working in these areas for over 10 years. For 3 years he was also consulting to the Bloomsbury Computing Consortium as a Senior Systems Analyst on the installation of a multi-campus distributed system. He graduated in Physics from Trinity College, Cambridge University in 1979, and gained his MSc in Computing in 1981, and PhD in 1993. He is a member of the ACM, the British Computer Society and the IEE. He is general chair for the ACM SIGCOMM 94 symposium. He is also on the editorial teams for the IEEE/ACM Transactions on Networking and the Journal of Internetworking Research & Experience.
Current Projects which reflect his interests include: an ARPA funded project building Multimedia over Multiservice Networks, RACE PREPARE Project Piloting ATM for Conferencing Systems and the management of the resultant services, the EPSRC FUME project investigating advanced IP multicast algorithms and an ESA Project investigating Load Balancing for Space Craft Control Systems
David Lewis gained a degree electronic engineering at Southampton university in 1987 before working in that field in industry for two years. In 1989 he came to UCL were he gained a degree in Computer Science before joining as a researcher in 1990. While at UCL he has conducted research in ISDN hardware and IP traffic analysis. Since 1992 he has been working in the PREPARE project in which he is leading technical work-packages in broadband network specification and teleservice management development. He is also actively involved in project management, being members of both the project and technical coordination committees.
Ian Wakeman gained degrees in electrical engineering from Cambridge and Stanford in 1987 and 1988, then worked at the GEC Hirst Research Centre on High Speed Networks and their protocols. In 1991 he joined UCL as a researcher, and he is currently interested in the problems of transmitting video over packet switched networks and other networking area research.
Communic8 Software Europe Plc
Communic8 Software Europe is part of the K2 Group which includes, bespoke business software development (K2 Systems), UNIX software distribution (Xanadu), Help Desk Support (Actiontrac). The group is targeted to achieve a 10 million turnover in 1994 with a projected profit of 1 million. Some 65 staff are employed in two offices in Watford and London. The customer base is primarily blue chip corporations who use a cross section of group services.
Communic8 Software Europe specialises in developing and marketing network software. Areas of specialisation are TSSnet (DECnet network integration for UNIX systems), Open RS/36 (System/36 emulation for UNIX systems) and PKZIP ( a data compression solution for Mainframes, Mini's and UNIX systems). The latest product just being launched is Z/ftp (to be called StreamZ) a high performance Enterprise File Transfer system designed to replace the ftp component in the TCP/IP suite. Principle staff involved in this project are:
Keith Shipton is Managing Director of Communic8 Software Europe. Keith has 15 years in the computer industry working in both direct sales of computer systems as well as management of successful computer operations. Keith has been involved in several recent projects for customers such as MAFF, MOD and Barclays Bank.
Tim Carmichael has been in the IT industry for 13 years and has worked for Communic8 for the last 2 years. Tim is responsible for new product development and marketing at Communic8 where he works closely with Keith Shipton. Previously Tim worked for Computacenter, the largest PC dealer in the UK in varying roles including project management and was also UK Manager of the Installation Engineering Division (65 staff) for 3 years.
Mike Charnock has worked for Communic8 for 3 years and is responsible for new product research. Mike has worked in the IT industry for 5 years and has a wide range of skills in networking and programming.
Brian Dorricott is Communic8's lead Software Engineer for networking products. Currently Brian is working on a new file transfer product called Z/ftp, soon to be launched in the UK and Europe. Brian has worked in the IT industry for 6 years and is highly proficient in industry standard development environments such as the 'C' Language and X-Windows.
Tadlon Ltd
Tadlon is a small software house with a number of clients in the City of London. Tadlon has extensive expertise in the integration of real-time prices and information delivery systems from the major information vendors. Principle staff involved are:
Greg Brougham is a principal consultant with more than ten years experience of implementing dealing and dealing support systems in the City; he has worked on large real-time dealing systems across multi-site heterogeneous environments. He will be involved in the specifying the site characteristics and approach to be adopted in analysing the network requirements of applications.
Huon Butterworth is a principal consultant with over twelve years experience of open systems. He has considerable experience in the implementation of applications in large multi-user heterogeneous environments and has experience of object oriented design and implementation. Huon will be involved both in the specification of site characteristics and in the analysis of the selected applications.
Texaco Ltd.
Texaco is a large multinational oil company with several sites in the UK. They have a complex corporate network an make heavy use of distributed applications. They will be represented in the project by Ken Brown, network administrator at a Texaco site.
Though there are several opportunities for exploitation of the results form this project, the one that will
be most actively pursued is the integration of the RAMS technology into the Z/ftp product set. The enhancements to Z/ftp will add significant functionality and improve its prospects in the market place. The SNMP agent controlling the operation of Z/ftp will be marketable after the project. Integrating Z/ftp into RAMS will create an early lead in the management of distributed network applications. We would expect that the approach adopted will also be widely applicable in other products.
It is difficult to accurately access the true commercial potential of an integrated Z/ftp and RAMS product sets, however Communic8 are budgeting in the region of £1 million in revenues from Z/ftp sales over the next 12 months. They are looking to develop commercial relationships with European partners as well as development of the USA and Pacific Rim markets. They would anticipate revenues from Z/ftp in the second 12 month period being in the region of £5 million. Communic8 would anticipate additional revenue opportunities from a properly integrated Z/ftp and RAMS product set but are unable at this moment to assess the likely revenue potential. One obvious concern is that we are in a position to jointly proceed in product integration as soon as possible so that we are able to maximise potential at the earliest opportunity.
We are not currently aware of any commercially available products that will address the networking problem in the comprehensive manner that Z/ftp and RAMS will be able to do so once integrated, so clearly we appear to have a competitive edge that we should take full advantage of. Our view is that as small but dynamic commercial organisation working with a centre of network research excellence we offers a rare combination of skills well capable of fully exploiting the competitive edge the integrated system will offer.
The aim of the project is to examine the interaction between applications and the network, and to develop a Resource Allocation Management System which can be used to ensure that critical applications have the network resources they need to operate effectively. Currently, there is a lack of information flow between applications and the network. An application has little knowledge of what is available in the network and thus can only hope that its traffic can be accommodated. On the other hand, the network can do little with excessive traffic except discard packets. In this project, we will try to establish a two-way information flow between the applications and the network, so that both applications and the network can make sensible decisions as to how to make the best use of resources.
We will use the standard Internet network management protocol, SNMP, to develop the system. RAMS will work by managing the protocol stack on each workstation attached to the network; on each workstation there will be a network management agent which controls how many packets the protocol stack can inject into the network. Packets will be discarded before they enter the network; in this way, we hope to improve network performance substantially.
We will also investigate how applications will interact with RAMS. Most applications will be unaware of its presence; it will be up to the human managers to configure RAMS so that these applications receive appropriate network resources. Tadlon will investigate how the performance of such applications can be improved using RAMS.
One application, developed by Communic8 Software Europe, will be enhanced to work intelligently with RAMS. Z/ftp is a sophisticated file transfer system which incorporates data compression, reliable transfer and security features. This will provide a reference point against which other applications can be measured, as Z/ftp replaces a major TCP/IP component called ftp (file transfer Protocol). Since ftp is an application under constant and heavily loaded use, we will be able to compare how effective the incorporation of RAMS awareness in Z/ftp is. Communic8 Software Europe will use Hewlett Packard's OpenView, the rapidly emerging market leader in network management, to develop an SNMP agent which will both control Z/ftp and interact with RAMS.
A major part of the project will involve examining the traffic on networks to ensure that the solution is widely applicable. We will conduct a traffic analysis which shows what percentage of network resources is used by all applications, including Z/ftp. After RAMS is deployed, we will investigate what impact giving more resources to Z/ftp has on other applications such as Network File Systems.
Explanation of techniques
Traditional network management systems like the ones defined in the OSI (CMIS/P) and Internet (SNMP) standards have a limited role to play in congestion control. A network management system cannot gather statistics about network conditions in a timely way to manage individual connections in a network; there is also the problem that traffic injected by the network management system itself will add to the congestion in a network.
UCL's research into the management of network traffic has been guided by the view that traffic should be managed by the protocol stack in host computers in a network. The TCP/IP protocol stack monitors network conditions via round trip timers and when it senses congestion, reduces the amount of data that is injected into the network. The problem with this approach is that the protocol stack can only manage individual connections and does not know about overall network conditions.
The novel way in which we wish to apply traditional network management is to manage the protocol stack itself . We will implement hooks inside the protocol stack and then link them to the management system. On each computer a (network management) agent will control the amount of data that the protocol stack can inject into the network. A central manager will determine the behaviour of the agent on each host, how aggressively it can increase the number of packets it sends and how rapidly it should back off when maximum network capacity is reached.
There are various ways in which requests for resources could be processed. One way would be for protocol agents to send requests to the central manager, which would calculate the impact of the request and then reply to the protocol agent. This would probably introduce far too much delay, and in such a tightly coupled system failure of the central manager is likely to cause serious problems. Alternatively, the resource allocation scheme could work in a similar way to the TCP congestion control algorithm, though it would apply to the whole stack not just to individual connections. Information about network conditions will come from regular transmissions from the central manager rather than from the packet acknowledgements of individual connections. As it receives reports about network conditions, each protocol agent will increase or decrease its flow, taking into account the number of connections to which it needs to give priority. This approach will result in a far less tightly coupled system, which should be more resilient to failure; if the central manager fails, the stack would continue to operate as an unmanaged stack.
There are numerous problems to be addressed. We will investigate how to incorporate the traffic control into the protocol stack. The protocol stack was not designed to be managed so we need to modify it and install hooks to the management system. The control will have some CPU and communication overhead. We will look at mechanisms as to how the overhead can be reduced to minimum. We will also experiment with various schemes for dealing with congestion and priority.
We will look at what level the stack can be managed and what level of granularity can be achieved. We will also examine the interaction between the management agents and the traffic controllers in the stack, and the speed at which we can adapt to changing conditions. We envisage that the performance of the managed versus unmanaged stack, particularly at the lowest layers, will be a crucial consideration. One way to tackle is this is to switch on lower layer packet classification only when the network management system senses congestion; the performance degradation of the protocol stack will help reduce network congestion and the extra CPU overhead will be offset by improved network conditions.
The fundamental principle behind the research is that a resource allocation system must manage the sources of traffic in a network. The source of traffic is ultimately not the protocol stack, but applications. We envisage that, in future, applications will be integrated into network management systems; applications will be able to ascertain what network resources are available and network management systems will be able to provide much better reporting as to what applications are causing network congestion. However, it is unlikely that all applications will be made manageable for some time and by managing the stack, one can control all traffic flowing from all applications, regardless of whether they are manageable or not.
Tasks
There are three basic tasks:
Traffic Analysis
We will analyse the traffic conditions of the various networks used in the project both before and after the RAMS is deployed. Tadlon with assistance from UCL will analyse traffic conditions and write a report of findings which will be presented at seminars attended by Texaco's IT staff. The report will both give an overview of traffic conditions on Texaco's LAN, compare the different types of traffic that one finds in the networks of a large corporate, an academic institution and a small company's LAN and describe what impact deployment of the RAMS has on applications.
Design and implementation of RAMS
UCL will build this prototype. Management hooks will be incorporated into TCP stack on a Sun Workstation. The RAMS prototype will consist of an agent running on each workstation which manages the protocol stack. A central manager will both gather information about network conditions from each agent running on a workstation and allow human managers to configure how resources should be allocated.
Applications development
Tadlon will examine different ways in which applications that are not managed (i.e. not Z/ftp) can be re-configured. Emphasis will be placed on analysing client server applications, particularly where server computers are controlled by RAMS but the client computers are unmanaged. Tadlon will investigate the impact of only being able to control part of the network sources.
Communic8 will enhance Z/ftp to provide SNMP management and RAMS awareness. The first task will be to define a Management Information Base (MIB) that describes the various components of Z/ftp. Management hooks then need to be added to Z/ftp. An SNMP agent will then be developed that implements the MIB and interfaces with the management hooks. For the second stage, the SNMP agent will be modified so that it can be controlled by RAMS.
The project will require two full time UCL researchers for a duration of twelve months for the analysis, development and trials of the resource management system, as well as project management and dissemination of results. UCL will also require one months effort from a hardware technician and one months effort form a systems programmer, these being essential during the initial analysis and final trials where workstations and software will have to be installed at different sites.
Communic8 will require 29 months effort for their part in the resource allocation management system analysis and trials as well as for the development of Z/ftp enhancement for integration with the rest of the system. This effort also covers technical support and project management.
Tadlon will require 4 months effort to carry out a detailed analysis of the applications and traffic on the various networks and to examine what impact RAMS has on applications.
Texaco will provide their Local Area Network (LAN) for traffic monitoring.
The overall objective of the project is to develop a saleable resource allocation management system that is applicable to the needs of commercial users, and can be tailored to the requirements of their networks and distributed applications.
The specific objectives to be attained during the project are:
To develop an understanding of the use of distributed applications in a commercial environment and the related network load imposed by these applications.
To use this understanding in developing a resource allocation management system that will improve the utilisation of network resources by distributed applications.
To develop a demonstrable prototype of the system that can be used with existing commercial applications.
To demonstrate the effectiveness of the system by conducting an analysis of its performance and comparing this to results of the initial analysis.
In order to ensure that these objectives are achieved the project will be split into a series of discrete activities. These activities are shown below:
project management,
analysis of user site traffic and application usage,
prototype specification and development,
prototype/ Z/ftp integration,
prototype trials and performance analysis
prototype assessment and dissemination of results.
Each activity is associated with a deliverable. These deliverables will be the means by which the project coordination committee will assess the success of the various stages of the project and will be the primary means of demonstrating the progress of the project to the monitoring officer, i.e. they will act as milestones.
The relationship between the projects activities and the deliverable is shown in the chart in section 5.2. Descriptions of the deliverables are given in section 5.4.
The philosophy behind this plan is to start with a firm understanding of what the commercial requirements are before launching into the design and implementation of the resource allocation management system. Using Tadlon's expertise in commercial distributed systems and Texaco's real user requirements at an early stage will ensure that the innovative input from UCL and Communic8 is directed accurately towards real needs, thus maximising the prospect of producing an exploitable system.
The following chart shows how the effort of different partners will be distributed over the 12 months of the project on the different activities. The management activity will be spread throughout the project and is omitted from the chart for clarity.
The contributions by each partner are given below. All costs are for the single year of the project.
UCL Budget:
Category Breakdown Cost
Staff 2MM, Initial network analysis, 55961
8 MM, prototype development,
5 MM, prototype integration,
6 MM, prototype trials,
3 MM, final analysis and
dissemination,
2MM, coordination and management
Travel: to user sites and management meetings 255
Consumables: office supplies, 5820
access to shared departmental
resources,
equipment maintenance.
Total 61991
Communic8 Software Europe Budget:
Category Breakdown Cost
Staff 2MM, Initial network analysis, 125934
10 MM, prototype development,
4 MM, prototype integration,
8 MM, prototype trials,
2 MM, final analysis and
dissemination,
3MM, coordination and management
Materials 5000
Fees for 53000
Trials and
Testing
Licensing of 95000
new
technologies
Travel to development, integration and 5000
management meetings
Total
283934
Tadlon Budget:
Category Breakdown Cost
Staff 4MM, application analysis and 38000
subsequent review
Travel to user sites and management meetings 2000
Total 40000
Texaco Budget:
Category Breakdown Cost Staff network support and management 2000 Travel To management meetings 500 Infrastructur use of network for analysis and trial, 7500 e Total 10000
The following table summarises the breakdown of partner effort into the different activities:
Activity UCL Communic8 Tadlon Texaco Total Analysis of user site traffic 2 2 2 0.2 6.2 and application usage Prototype specification and 8 10 0 0 18 development Prototype / Z/ftp integration 5 4 0 0 9 Prototype trials 6 8 0 0 14 Prototype assessment and 3 2 1.5 0 6.5 dissemination of results Project coordination 2 3 0.5 0.1 5.6 Total (man months) 26 29 4 0.3 59.3
5.4. Deliverables
The following deliverables will be generated:
User Site Analysis Report, due end month two: This report will summarise the work performed by Tadlon and UCL in analysing the relationship between network and distributed application usage over sections of Texaco's corporate network with possible additional analysis of Tadlon and Communic8 client sites. Conclusions on the impact of the results of this analysis on RAMS requirements will be given.
Demonstration of Prototype Stack Management, due end of month seven: This demonstration will show the operation of the RAMS management application and TCP/IP stack management agent.
Demonstration of Prototype Application Management, due end of month nine: this Demonstration will show the operation of the Z/ftp management agent.
Prototype Assessment Report, due month twelve: This report will provide an assessment of the changes to network and distributed application observed during prototype trials. A plan outlining the steps required to turn the prototype to a commercial product will be developed.
Seminar and Demonstration of Integrated Prototype: A seminar will be given describing the results of the initial user site analysis, the RAMS design and how it was influenced by the analysis, the effects of using RAMS on network performance and plans for the future. This will be accompanied by a demonstration of integrated stack and application management using the RAMS prototype.
UCL will be the coordinating partner. Dr. Jon Crowcroft will be the project leader assisted by David Lewis. Both have extensive experience in managing large projects in collaboration with industry, both through involvement in the RACE program and through bilateral projects with BT, Xerox, Sun and Bellcore.
A Project Coordination Committee will be established and will meet approximately every month. This committee will consist of at least one person from each partner. It will be responsible for monitoring the project and reviewing deliverable status and contents. In the event of problems in the execution of the project plan the committee will be responsible for initiating remedial action including a rescheduling of the plan if necessary.
During the development and testing phases and site analysis and trial phases further technical coordination meetings will be held between the relevant partners only. At least one attendee of every technical coordination meeting should attend the next project coordination meeting to submit a report.
The major technical risk facing the project is that the Resource Allocation Management System proposed will not provide the intended benefits to the commercial applications being targeted. The project consortium and project plan have been structured in such a way as to minimise this risk. This has been done by supplementing the key technical skills of the academic partner in this field with the commercial needs, skills and resources of the industrial partners. These consist of: the requirements of a user organisation (Texaco Ltd) for an analysis of its distributed applications, the skills of a consultancy (Tadlon) specialising in distributed applications software development, which are essential in analysing the requirements of the user organisation, the skills of a distributed application vendor (Communic8), required in analysing the relevance of RAMS, based on knowledge of their customer's requirements. Communic8 will also apply essential skill in incorporating their own products into the resource allocation management system.
The following resources will be deployed by each partner:
UCL:
Effort: 26 MM
To be used in assisting with the initial application analysis, developing the RAMS prototype and protocol stack enhancements, and testing and evaluating these both at UCL and at Communic8.
Equipment: UCL will use existing equipment (SPARC workstations) for software development and for use in the initial and final analysis phases.
Infrastructure: UCL operates a complex local area network supporting 100+ research related workstations and a similar number of teaching machines. This will provide an ideal testbed for the initial testing of the RAMS prototype. As well as the Communic8 software a wide variety of other applications, including multimedia conferencing and information servers, are in operational use across this network, providing a rich environment for additional traffic analysis and prototype evaluation.
Communic8 Software Europe:
Effort: 29MM
To be used analysing, designing and developing an SNMP management agent and RAMS awareness within Z/ftp. Communic8 will also provide direct technical administration and project management time.
Equipment: Communic8 Software Europe will provide their entire LAN resources for the duration of the project. This comprises some 40 x PC's, 30 x X-Stations, 6 x HP 9000 RISC Systems, 4 x IBM RISC System/6000, 1 x Integraph, 4 x Windows NT Servers, 1 x ICL DRS6000, 2 x DEC VAX, 1 x IBM AS/400, 1 x IBM System/36, 5 x SUN SPARCstations, 1 x SCO, 1 x Novell Netware Server. 4 Segment Ethernet LAN, 64K Internet connection. Communic8 Software Europe will also have to purchase some industry standard software including HP Openview and a Network General Sniffer.
Infrastructure: Licensed copies of the Z/ftp product will be provided to UCL as part of the integration and testing process.
Tadlon:
Effort: 4MM
To be used in conducting the initial analysis of how distributed applications are used over and affect commercial LANs. The conclusions of this analysis and its input to the design of the RAMS will be presented to other partners.
Equipment and infrastructure: equipment at network sites will be used for the analysis and evaluation work
Texaco Ltd:
Effort: 0.3MM
This will consist of technical effort required to support the traffic analysis work at the Texaco's site, support for the installation and operation of the prototype and attendance at seminars held to present findings.
Equipment: Texaco will supply a SPARC workstation specifically for the use in the analysis and evaluation work.
Infrastructure: Texaco will allow use of their operational network for analysis work
The primary route to exploitation from this project will be through the integration of the RAMS technology with Communic8's Z/ftp product set.
Background
As a commercial organisation, Communic8 are seeking to define and deliver a marketable world class product or products at the end of the project. Indeed Communic8 can only afford to be involved in the RAMS project if the quality and focus of the research is oriented towards a perceptible customer need which can be supplied as a product at a marketable price.
As recorded before, Communic8 is already building world class network products (Z/ftp) which fill a direct customer need and are way ahead of competitors in the US, Europe and Pacific Rim. This trend must be successfully continued to ensure the commercial mandate for Communic8 can be maintained. Communic8 believe that the RAMS project will provide world leading technology which can be packaged and exploited globally.
The RAMS project dovetails neatly with work Communic8 is already doing in the field of network optimisation. Within the existing framework of international standards, Communic8 will be providing products which meet forthcoming international customer directions. To ensure that Communic8 stay in line and on target with both international standards and emerging technology Communic8 are producing dynamic 2, 5 and 10 year plans for RAMS based products. These plans will be constantly monitored and adapted as external standards and technology change.
Productisation
Initially Communic8 will be able to kick off existing products such as Z/ftp for the first RAMS products. Using this technique Communic8 will be able to market test a demonstrator to ensure Communic8 have got the substance and packaging right. Once Communic8 have established the basic substance, Communic8 will look to provide multi-platform products through packaging. Additionally to multi-platforms, Communic8 will be able to package RAMS based products to fit particular market segments. For instance, the Microsoft Windows user will not have the same requirement of functionality as the administrator of an IBM Mainframe. The IBM mainframe administrator will expect to pay substantially more for the product than the Microsoft Windows user. Further, the products will be modularised to enable a user to start cheap and build as required, enabling extendibility to be integral to the product marketing plan.
Pricing
Pricing for products is very market sensitive with an overall downtrend in retail price and upward trend in functionality, and will need to be determined close to point of launch. It is anticipated that the RAMS based products will be able to contain a high profit margin percentage due to the multiple packaging options available to Communic8
Global Marketing
Using existing European distribution links, Communic8 will be able to establish RAMS based products with speed to any EU based corporate. As soon as possible Communic8 will seek to promote sales in the United States, Japan and Pacific Rim. In common with all Communic8 products, RAMS based products will have national language support appropriate to the country of use.
Advertising & Marketing
Market awareness is a key part of Communic8's sales strategy. In common with existing product ranges, Communic8 anticipate investing significant sums in advertising and promotion. Typically this will involve running an advertising campaign in key periodicals which will establish RAMS based products with the target audience. Communic8 back-up advertising campaigns with comprehensive customer marketing to ensure that key organisations are aware of the RAMS based products at all levels. It is important that leading industrial organisations are seen to be adopting RAMS based products, which in turn will re-inforce the market position with middle tier organisations.
Future Planning and Products
Using long-term planning Communic8 will be able to bring a series of continuously enhanced RAMS products to market. The key strategy is to stay within defined International standards to enable global corporations to justify purchase. The RAMS based products are likely to define a new international standard for resource management but this must still remain within prevailing international standards to ensure marketability.
Whilst the RAMS research is likely to lead to highly complex and technical solutions, the first products will have to very focussed and simple in approach to avoid overwhelming the client. By setting the initial objective in easy reach, we can slowly but more surely migrate customers onto the more complex and sophisticated deliverables from RAMS. Such a policy is important, as it prevents the customer having to be involved in the long evaluation procedures associated with large or complex products.
Other marketable products and services will come from this proposal:
The RAMS prototype, when developed further, would be sold in a number of ways. It would be sold in a similar way to NFS; just as an NFS client allows access to files on distributed file servers, so a RAMS client would enable controlled access to network resources. The exploitation route for this type of product would be to sell it to operating system vendors and workstation manufacturers, such as Sun, on whose machines the initial demonstrator will be developed. RAMS could also be sold as an add-on to other network management platforms which use the industry standard SNMP protocol, such as HP OpenView. Network Management systems currently concentrate on managing network devices, rather than distributed applications; RAMS would solve this shortcoming. Communic8, in collaboration with UCL, will during the project assess RAMS' market potential.
RAMS would also be sold as part of a consultancy package. Debugging networks is as complicated as debugging software and with RAMS, it will be possible for Communic8 and Tadlon to approach network consultancy from an applications oriented angle. This approach is likely to be much more attractive to IT departments who are interested in consultancy services that identify problems with their distributed applications, rather than just their network equipment.
Partners will benefit from the results of this project in the following ways:
Texaco will benefit from an in-depth analysis of the requirements of their networked applications. If used, the resource allocation management system will also improve the efficiency of the organisation's internal network, essential in an area where data mobility is key, and reduce network management costs.
Tadlon will benefit from widening their knowledge of the operation of a distributed applications of primary importance to their client base.
Communic8 will benefit from having access to a unique and jointly developed system, well suited for inclusion in their existing product base and which is commercially viable.
UCL will benefit from being able to demonstrate the applicability of its research work to a commercial environment. UCL also expect to play an active part in any future commercial exploitation of the project results.
All partners expect this relatively modest project will lead to further, fruitful collaborations between the academic and industrial partners.
UCL expects that the RAMS research will lead to new research projects in the management of distributed environments. A logical extension to the research proposed here would be to design a new type of protocol stack. From the outset, this would be designed so that it could be managed by a network management system. The stack would be dynamically configurable so that an application could construct a suitable protocol object to fit its network requirements (whether it was reliable, synchronous, asynchronous, what performance characteristics it required etc.). Object oriented techniques could be used to deal with the traditional problems of protocol layering; inheritance may provide a more elegant and efficient way of implementing layering. This kind of research has already been started (e.g. CEC Hipparch project), but there are many problems. Efficient object oriented languages (e.g. C++) are statically bound; negotiating context between dynamic end protocol stacks is problematic; dynamic configuration makes possible the construction of invalid protocol objects (e.g. they deadlock).
Each partner will hold the IPR of their own work. UCL normally retains IPR in its research to ensure that it can use its software in future research projects without prejudice.
UCL will license its software to commercial partners and UCL 's research staff will actively participate in the exploitation of the software developed during the project. Professor Kirstein, Head of UCL CS, has given his approval. Dr Jeff Skinner of UCL Ventures will advise on all aspects of the commercialisation and ensure that it is done according to college rules. UCLi, UCL's contracts division, will draw up appropriate exploitation agreements. A formal collaboration agreement will follow by 1st October 1994.
The work has clear industrial relevance but is sufficiently far removed from the market that public money is critical to the viability of the project. Successful completion of the programme will help the UK to maintain and improve its position in the Internetworking field particularly in its rapidly expanding commercial application. As the Internet is rapidly being accepted by the business community, there is an enormous opportunity for the UK to exploit its position. This project is only possible due to the combination of technical expertise in Internet protocols and network management supplied by the academic partner and the expertise and knowledge of commercial requirements and exploitation routes provided by the industrial partners. It is however only with the help of public funding and support for academic cooperation that the industrial partners are able to participate.
This work clearly falls within the HPIP programme. The project will develop a resource allocation management system that works by introducing information flows between the network and the protocol stacks and distributed applications running in end user systems. The results of the project will improve the way in which network resources are utilised.