GN1 (GÉANT) Deliverable D13.2

Transcription

1 Contract Number: IST Project Title: GN1 (GÉANT) Deliverable 13.2 Technology Roadmap for Year 3 Contractual Date: 31 July 2002 Actual Date: 16 August 2002 Work Package: WP8 Nature of Deliverable: R - Report Dissemination Level: PU - Public Author: Roberto Sabatino, DANTE David Sutherland, DANTE ABSTRACT This document updates the Technology Roadmap given in section of the GN1 (GÉANT) contract. The developments carried out during Year 2 of the contract are reviewed briefly and a revised list of topics for investigation in Years 3 and 4 are given. KEYWORDS IPv6, QoS monitoring, SSM, optical networks GEA v August 2002

2 CONTENTS 1 EXECUTIVE SUMMARY Introduction GN1 Service developments Multicast Premium IP Creation and utilisation of a distributed service test bed Support for special projects Monitoring and Network management Deployment of multidomain monitoring infrastructure Performance Response Team Technology testing and development IPv Optical Networking developments Router testing Table 1: Technology Roadmap...7 GEA v August 2002

3 1 EXECUTIVE SUMMARY This document reviews the activities relating to the investigation of new technologies and services during the second year of GN1 (GÉANT) and describes an updated Technology Roadmap for such activities in the third and fourth years of the GÉANT service. The broad conclusions of work in other areas is reported. In each of the technological areas of interest, recommendations for further work are made and form the basis of the updated Technological Roadmap. GEA v August 2002

4 2 Introduction This document defines the technical road-map for the year 3 development activities for GN1. The activities are defined in three groups namely, service development, monitoring and network management and technology testing. For each group of activities the work planned for year 3 is set out below. To give additional context to the work program a summary of the expected status of work at the end of year 2 is provided. 3 GN1 Service developments One of the main purposes of GEANT is to extend the basic best efforts IP service with new advanced services. In that respect we have spent considerable efforts for the development of services such as multicast, Premium IP and Scavenger Services. Multicast is deployed on all GEANT nodes and offered to any NREN that wishes to make use of it (in fact most NRENs do). For Premium IP we have devised an architecture and implementation model which has been piloted with a number of user groups and is available in all GEANT nodes, although more work is needed for its operational management. Scavenger services are based on a simpler subset of the techniques used to offer Premium IP services, and as a result is simpler to implement and requires much less operational management. We expect to have completed the work on Scavenger services before the end of year-2. In year-3 we will focus on enhancements to multicast and Quality of Service and will work on the addition of a distributed testbed service and support for special projects. 3.1 Multicast A basic multicast service has already been implemented at all the GÉANT nodes and is used by most connected NRENs. This service is being complemented by the provision of beacons which provide enhanced monitoring of multicast traffic. The beacons are able to measure network metrics for multicast such as loss, delay, jitter, order and number of duplicate packets. In year 3 we plan to carry out the following additional developments: History feature for beacon: currently beacon can only provide live data, whilst it would be useful to have more historical data available, in order to be able to monitor network behaviour changes. This work will be completed in Q1 of year-3; Per group monitoring: multicast sessions are organised as network group communications. We will develop and deploy tools which can perform networking monitoring on a per-group basis, showing which users and how much traffic are associated with each group. Group information is essential for resolving user-perceived performance problems for multicast, as users will always refer to a particular multicast group communication problem. This work will be complete by Q3 of year-3; Single Source Multicast (SSM): SSM is a recent development of multicast and its design is a reflection of the fact that most of today s multicast sessions follow the one-tomany model rather than the many-to-many. SSM sessions can be combined to form many-to-many multicast communications. SSM requires a simpler network configuration than the many-to-many multicast and thus offers improved reliability. An initial study of the implementation issues for SSM will be followed by test cases and eventually deployment on GEANT, if the test cases are satisfactory. The test results will be available in Q2; Experimentation with high capacity streams: to date multicast has been used by applications with low bandwidth demands, typically less than 5Mbps. Given the availability of Gbps on GEANT and the ongoing development of applications capable of generating and processing high bandwidths, it is expected to run applications with bandwidth requirements of hundreds of Mbps. These applications can vary from GEA v August 2002

5 transmission of high quality video for remote teaching of surgery to transmission of high quality audio and video of concerts. The high bandwidth demands for multicast will put under severe test the reliability of the implementation of multicast across GEANT and NRENs and will therefore undergo a planned set of experimentation; This work will ebb carried out throughout year-3 and conclusions available in Q4. Multicast over IPv6: the issues related to deploying multicast over IPv6 will be analysed. This work will remain a study, as it is not expected to run multicast over IPv6 during year Premium IP Based on the work carried out in TF-NGN we have successfully piloted a Premium IP service across a combination of GÉANT and connected NREN networks. Whilst from a technology point of view Premium IP services have proved to be viable, nevertheless there are significant challenges in managing their provision and monitoring their subsequent performance. Work in year 3 will concentrate on remedying these weaknesses by developing management tools that will automate provisioning of service. In addition we will implement a QoS Monitoring infrastructure to monitor the performance delivered to users. Specific work items are: Web-based User interface for collection of requests: currently a set of forms need to be filled in manually which provide all the technical and administrative information required for the setting up of a Premium IP between a pair of end users. Typically this information contains the expected performance values requested, (bandwidth, delay), the technology used in each network domain, technical contact person etc. All this information can be collected using web interfaces which can store the information in a format accessible by the other management tools that we will develop; Resource allocation: a set of tools to check the current resource allocation and determine whether a new request can be accommodated. The tools will deal also with scheduled and regular requests as well as one-time requests for Premium IP service; Router configuration update: once a Premium IP request is accepted, the GEANT routers need to be configured to perform the necessary functions. The configuration needed is conceptually simple but in practice translates to a complex set of instructions on the routers and is prone to human error. We will develop tools that will automatically provide updated configuration for the routers. The network management tools outlined above will be available by Q3 of year Creation and utilisation of a distributed service test bed One of the goals of GEANT is to make available additional advanced services such as multicast, QoS, IPv6 and others. These are very complex services which require complex configurations on the routers. In addition, often the services interfere with one another. Given these complexities and challenges it is not possible to trial in the first instances the new services on the production routers. Instead, the services must be proven first in a controlled laboratory environment which can simulate as far as possible the real operational environment. Once the set of laboratory tests is complete with satisfactory results, testing can move to the operational network. Given the number and complexity of the services that we must develop on GEANT and on NREN networks it is essential to have access to testbed facilities. Testbeds are essential for carrying out tests that range from testing the advanced services mentioned earlier to simple tests for the operational network such as an Operating System upgrade. GEANT does not currently have a dedicated testbed, which has a negative effect on the operation and development of GEANT. We have had access to test facilities of router vendors and to GEA v August 2002

6 national testbeds, but access to these is very limited, and a more permanent and accessible solution is required. The creation of a permanent distributed testbed on GEANT, to be used by DANTE and NREN engineers for the development of GEANT and NREN networks would fulfill the requirement. The testbed must comprise a representative mixture of routers that are encountered in GEANT and NREN networks, such as Juniper, Cisco 12k, Cisco 7k. The routers will be distributed across GEANT and interconnected to one another using MPLS/CCC functionality available on the production GEANT routers in order to create a VPN for the testbed. The testbed must also contain traffic generation and analysis equipment such as SmartBits. DANTE has already obtained on behalf of GEANT an open ended loan of four SmartBits which can be used in the testbed. The testbed will be particularly useful for the phased introduction of IPv6, enhancements to multicast and QoS, and will also be essential in the router testing activities that will be carried out in year-3. The activities in year-3 will include the detailed definition of the testbed, the procurement, installation and management of the hardware. This result will be achieved in Q1 of year-3 and will be followed by its continuous use in support of operation and development of GEANT and NREN networks. 3.4 Support for special projects Special projects are those whose connectivity requirements are not by default fulfilled by the standard best efforts IP service accessed via the NREN access ports to GEANT. These can be IST projects, GRID activities or other. The exact requirements can differ substantially between each project. These will be analysed on a case-by-case basis and a technical solution will be proposed for each one of them. In some cases nothing additional must be done (i.e. access can be via the normal NREN access using standard best Efforts IP), in other cases Premium IP or VPN functionality of some form may be required. The GEANT working groups have tested, during year-1 and year-2, a variety of techniques to fulfill a variety of requirements, but it has not been possible to define consistent services because of the lack of consistent requirements. In year-3 the requirements of at least five special projects will be analysed and technical solutions proposed for each one of them. This work will continue throughout year-3 and results available at the end of Q4. 4 Monitoring and Network management GÉANT only makes use of IP technology, in contrast to TEN-155 where a mixture of ATM and IP technology was employed. The IP technology deployed is capable of performing at the line speeds used in GÉANT, however, in terms of manageability and performance measurement the technology is immature and somewhat unreliable. This is reflected, for example, on the need to do measurement of traffic flows based on sampling which lead to an inherent inaccuracy given the variable packet sizes and the variable life time of streams in the Internet. In year-2 we have continued to study a number of traffic measurement tools, with particular emphasis on tools for monitoring QoS and verifying SLAs. This activity is complementary to the work on developing Premium IP services for GEANT and NRENs. In year-3 we will capitalise on the studies conducted during year-2 and implement a multidomain monitoring infrastructure for the QoS of the network and the services it offers (Best Effort, Premium IP, Scavenger). In addition, a new activity on the organisation of a Performance Response Team will take place. GEA v August 2002

7 4.1 Deployment of multidomain monitoring infrastructure During year-2 a comprehensive study of tools and architectures for monitoring QoS and verifying SLAs has been conducted. In particular the SEQUIN project has proposed an architecture, whilst TF-NGN has conducted tests of several SW packages and mechanisms to monitor network performance parameters. In particular tools and methodologies for monitoring one-way-delay have been conducted, which have highlighted the pros and cons of systems based on GPS synchronisation as opposed to those based on Network Time Protocol (NTP) synchronisation. During year-3 we intend to draw conclusions from these experiences and implement a multidomain monitoring infrastructure for QoS. The architecture required has been defined during year-2, whilst the choice is still open for tools and methodologies (GPS vs. NTP) to use. In year-3 a cost vs. benefit analysis will be performed which will lead to the formulation of a formal proposal for the infrastructure. It will then be procured, installed and managed and will become part of the basic GEANT infrastructure and future networks. This result will be achieved by Q3 of year Performance Response Team Understanding end to end performance has always been a complex task. In the event of poor performance, it is extremely difficult to identify whether performance problems are due to the applications or to the network. To date this has been conducted in an ad-hoc manner with mixed results. For a more structured approach to resolving end to end performance problems we suggest the creation of a Performance Response Team (PRT). This is made up of a combination of individuals with a good understanding of network behaviour and individuals that understand specific application behaviour. The role of the PRT is to assist in resolving end to end performance problems observed by end users. The activity will focus on defining the organisation of a PRT and will provide a sample implementation. A limited number of user cases (up to five) will be used to verify the concept and operation of a PRT. We will provide recommendations for the organisation of a PRT that is able to operate at a pan-european scale by Q4 of year-3. 5 Technology testing and development Work in year two has focussed on evaluation of new routing equipment and IPv6 and developments of optical networking. During year two evaluation of routers and developments of optical networking have been conducted mainly as studies with no hands on experience, mainly because of immaturity or unavailability of systems to test. This is set to change during year-3 where a number of router vendors (Alcatel, Avici) are developing routers with interesting sets of functionality, whilst for optical networking possibilities for obtaining hands-on experience are starting to materialise with suppliers of services and equipment. IPv6 activities in year-2 have focused on the establishment of the GTP-V6 overlay network based on a Juniper M5 router. This served as the platform to study conformance and interoperability issues. The most significant IPv6 activity during year-2 was the implementation of the infrastructure for 6NET. IPv6 remains a high priority activity for year-3 and will focus on implementation steps for GEANT. 5.1 IPv6 There will be two main activities for IPv6 during year-3. First, the GTP-V6 overlay network established during year-2 will continue to operate and will be extended with a router from Hitachi. This will result in an overlay IPv6 network in a dual star topology which will continue to be used to perform conformance and interoperability tests between different router suppliers. GEA v August 2002

8 Second, and most importantly, IPv6 will be introduced on the production GEANT routers in dual stack fashion. In preparation for this it will be necessary to perform activities such as performance analysis of dual stack routers, co-existence of IPv4 and IPv6, management, monitoring and MIB development. Some clear targets are set out for the introduction of IPv6 on GEANT: In autumn 2002, GEANT should support dual stack functionality, with static routing. This is the simplest form of support for IPv6 but represents a key milestone. Two NRENs (REDIRIS and ARNES) and Juniper are currently co-operating with DANTE in order to achieve this result; By summer 2003 dynamic IPv6 routing should be supported on the GEANT routers, complemented by network management and monitoring tools; NRENs will be encouraged to connect to the GEANT pilot IPv6 service to carry production-like IPv6 traffic. The transition from a pilot IPv6 service to a full production service is targeted for year Optical Networking developments This activity is essential for the understanding of the components or network elements for future GEANT and NREN networks. During year-2 it has focused on information gathering and dissemination whilst in year-3 we expect to obtain more direct hands-on experience. The work will focus on: 40Gbps: transmission systems capable of delivering 40Gbps clear channels are already available, although no supplier offers them in their portfolio yet. This is mainly because of lack of demand combined with additional capex needed to offer the service. On the other hand, GEANT and NRENs foresee the need of 40Gbps clear communication channels emerge within the next two years. By establishing co-operation with suppliers of equipment and services and other IST or FP6 projects it is expected to gain direct experience with these transmission systems for a better appreciation of the issues which surround their deployment and operation, including the operational costs. 10Gigabit Ethernet: similarly to 40Gbps, long-haul 10GE is a technology which is currently proven but not available commercially, for similar reasons to those of 40Gbps. Bandwidth on demand: GEANT has lost an important feature that was available on TEN-155, i.e. the ability to provide circuits with a high granularity of bandwidth to separate users. This has been partly compensated by differentiated IP packet treatment (Premium IP). The work on this topic will focus on studying and testing the techniques to provide circuit switching of wavelengths and lower capacity channels; Integrated IP and optical network management: In the future it is expected to have more direct access to the underlying optical infrastructure whilst at the same time operating an IP network on top. Having one common network management system for the IP and optical layers has the advantage of being able to manage all resources coherently, as opposed to independently which can lead to mis-use or sub optimal use of resources available. In year three systems for integrated network management will be studied and where available they will be tested. This is a complex set of tasks, and it will be a challenge to complete all these optical networking development activities in year-3. Interim results will be available during year-3 and more work will be required during year Router testing GEANT serves as a platform for testing new router technologies. On TEN-155 a similar activity took place which enabled a thorough evaluation of Juniper routers which eventually lead to their operational deployment. During year-2 evaluation of new routers has taken place in the form of meetings with suppliers, which in each case have outlined the immaturity of GEA v August 2002

9 the products in respect of the future requirements for GEANT. A series of meetings with Alcatel, Avici, Pluris Marconi have all outlined deficiencies (support for 40Gbps, multicast, diff-serv, IPv6) which are currently being addressed. We will perform performance and feature testing of Alcatel and Avici routers during year-3. This will take the form of initial laboratory testing followed by introduction on GEANT for a limited test period. This work will be complete by Q4 of year Table 1: Technology Roadmap Topic Details Plans Available Service Development Service Development Service Development Service Development Monitoring and Network management Multicast Premium IP Creation and utilisation of a distributed service test bed History feature for beacon Per group monitoring Single Source Multicast High capacity streams Multicast over IPv6 Web-based User Interface Resource allocation Router configuration update Definition of permanent testbed Yr 3 Q1 Yr 3 Q3 Yr 3 Q2 Yr 3 Q4 Feasibility study Yr 3 Q 3 Yr 3 Q 3 Yr 3 Q 3 Yr 3 Q1 Support for special projects Up to 5 special projects Results in Yr 3 Q4 Deployment of multidomain monitoring infrastructure Cost benefit analysis Infrastructure proposal Yr 3 Q3 Monitoring and Network management Performance Response Team Define organisation Verify concept Yr3 Q4 Technology testing and development IPv6 Operate GTP-V6 overlay Plan implementation steps for GÉANT Extend in Yr3 Q1 Yr3 Q3 Technology testing and development Optical Networking development 40Gbps 10Gigabit Ethernet Bandwidth on demand Integrated IP and optical network management Interim results by Yr3 Q4. Work will continue in Yr4 Technology testing and development Router testing Complete testing of Alcatel and Avici routers Yr3 Q4 End of Document GÉANT Technology Roadmap, Year 3 GEA v August 2002