 |
||||
|
Tutorial 1: Multi-Service Interworking over MPLS For many enterprises, Frame relay/ATM services are a major part of their
current solution for data traffic. As an end-to-end offering, such Layer 2
services and their associated service level agreements provide an enterprise
with a guaranteed level of performance, reliability and security their
business can rely on. Yet the price performance of emerging Ethernet
services is very attractive for enterprises to address changing business
needs at select sites where additional bandwidth is required. For carriers,
there are benefits from convergence of existing and services on a common
MPLS network. This half day tutorial focuses on the emerging industry
standards/agreements to support multiservice interworking over a MPLS
network. These enable the offering of a Layer 2 service such as Ethernet
using dissimilar access circuits (i.e. different link layer technologies:
Ethernet, Frame relay, ATM , etc) over a MPLS network. This interworking
specification enables carriers and service providers to introduce new
Ethernet services, while preserving existing infrastructure like ATM and
Frame Relay. Key existing and emerging industry standards/agreements are
referenced. Several enterprise service migration scenarios and benefits with
multiservice interworking over MPLS are highlighted.
Tutorial 2: Pseudowires and VPNs This tutorial will review standards based pseudowire technology and how this technology can be used to transport of a variety of traffic including frame relay, ATM, Ethernet, PPP/HDLC and emulate the QoS and OAM behavior of the transported protocol. As pseudowire technology has become widely accepted, carriers have begun detailed planning for rolling out the technology in support of L2 and L3 VPN services, and are now looking closely at operational aspects such as QoS and OAM. This tutorial will discuss possible scenarios in which pseudowires are adding value to service provider networks, and consider the operational aspects of making pseudowire deployments ubiquitous. In addition to reviewing the technology itself this tutorial will discuss important network applications and the corresponding benefits to service providers of pseudowire technology. These applications include the transporting of multiple protocols over a converged MPLS network as well as emerging applications which leverage pseudowire technology as a multiplexing layer for the efficient carriage of packet protocols over diverse physical media. Tutorial 3: MPLS Architecture, Services and OAM
This tutorial will first introduce the basic concepts of MPLS -- label switching, data and control plane separation, as well as label distribution methods. The presentation will then highlight how the key applications of MPLS operate using these functions. Finally, we will present some key OAM concepts and features and show how they can be used to provide a complete, carrier-class network solution for service providers. Tutorial 4: Triple-Play Services Richard Loveland , Ron Haberman and Arnold Jansen, Alcatel Triple play networks require careful planning of the foundation network, services and network management. Building the right network foundation and management infrastructure is a critical success factor to gain and sustain market momentum and hit the right cost targets in a rapidly evolving competitive landscape. To become a leading triple play service provider it is essential to design a network topology and service deployment strategy that is able to meet a variety of service requirements and operational needs, now and in future. To do so requires a clear understanding of the enabling technologies and implementation choices at hand. This tutorial is structured in three parts: - Access: Transforming the broadband access infrastructure in order to extend the service reach for triple play broadband service bundles in a cost and performance optimized way. - Aggregation and Service creation: Examines the challenges and enabling technologies associated with network topology choices, subscriber management for different modes of operation and service creation. Deployment scenarios and the operational tools required will be covered throughout the discussion. - Subscriber and Policy control: Operational support aspects related to end-to-end service control and assurance covering service creation, subscriber management, policy orchestration and support for various modes of operation. BT's 21st Century Network Building the World's Biggest Next Generation Network This presentation will provide a technical case study on BT's next generation network (21CN), one of the world's largest and most advanced MPLS/IP/Ethernet networks. In June 2004, BT publicly announced it was going to be the world's first national telecoms operator to take the bold step of replacing its entire existing legacy network infrastructure with a next generation network built using MPLS, IP and Ethernet. The plan announced was that by 2010 BT's 16 legacy networks (PSTN, ATM, PDH, Frame Relay, TDM, etc) would no long exist, they would be replaced by a single, national IP/MPLS,/Ethernet network. 21CN must support not just next generation services such as triple play and video on demand, it must also support traditional legacy services such as emergency services private circuits and the UK's national public switched telephony network. It must be a truly national network, covering all locations from remote rural to major cities. It must support the requirements of over 30 million customers, from the residential subscriber to largest corporate. It must support all services, from those requiring simple best effort, to those requiring over 99.9999% availability, delivery, and performance guarantees, and it must enable any communications services provider to connect into it seamlessly. To meet all of these criteria, BT is pushing the envelope of MPLS networking to the very limits of its capabilities, and in some cases beyond what can be delivered today. From day one, 21CN is supporting Graceful Restart (RFC3623, 3478, 3473) for all routing and MPLS protocols, pseudo wires (RFC3916), IP VPNs (RFC 4364), MPLS Fast ReRoute (RFC4090), and 8 levels of QoS. Because 21CN is the UK's next generation network national infrastructure it is carrier grade and industrialized, with the scalability, flexibility, resilience and robustness to support any application, at any grade of service, to any location across the UK. If 21CN fails to meet these very stringent requirements, the consequences could be catastrophic for residential, private and public sector businesses. IP/MPLS Network Design Challenges and Technical Solutions For NGN Intensive discussion is currently under way in many Service Providers (SPs) about how to deploy IP/MPLS based NGN, because SPs want a scalable and strong NGN architecture which can provide triple-play service (VoIP/Video broadcast/Internet access) for a large number of customers. (For example, NTT have a plan to deploy NGN for more than several tens of millions of customers.) It is also notable that several SPs are seeking a way to integrate their enterprise VPN services and triple-play service networks on the same NGN infrastructure. Even though MPLS is a strong technical tool, a lots of missing pieces are still missing. This presentation will introduce IP/MPLS based NGN design challenges and their technical solutions from the perspective of the SP's NGN deployment plans. It will expose the scaling and functional limitations of current MPLS- TE technology, and discuss the necessary technical enhancements for the NGN. The discussion will mainly cover, but not be limited to, the following technical items. - P2P MPLS-TE network designs and tools for realizing high-grade VoIP services - P2MP MPLS-TE network designs and tools for realizing high-grade Digital Video Broadcasting services - PW and MPLS-TE network designs and tools for realizing high-grade L2/L3 VPN services - MPLS network designs and tools for IPv4 and IPv6 dual stack IP routing and forwarding - MPLS-TE scaling issues and enhancements for Inter-AS/Area TE environments - MPLS-TE network designs and tools for triple-play and VPN service converged network - MPLS management architectures and techniques for triple-play and VPN service converged network - MPLS OAM challenges and enhancements for NGN The discussion of network design will be illustrated with descriptions of realistic deployment scenarios based on NTT's considerable experience of supporting triple-play and IP/MPLS service networks, and will draw on NTT's research and planning activities as well as information gathered from other Service Providers that plan to deploy IP/MPLS based NGN. Network Design for Large Scale Triple Play Services in MPLS supporting IP-Multicast Nicolai Leymann and Thomas Beckhaus, Deutsche Telekom Deutsche Telekom is currently building one of the largest triple play enabled networks. This network will cover 50 cities in Germany with speeds up to 50 MBit/s deploying a VDSL based access network and an MPLS enabled backbone. IPTV as a key building block offers 150 different Live TV channels including several HDTV channels. The platform is fully IP Multicast enabled to support high scalability. The presentation will give an extensive overview about the underlying network design and the implementation based on the requirements of the Triple Play service. It also covers problems being encountered during the design phase and corresponding solutions. The following topics and questions will be covered in detail: . Requirements of Triple Play Services: Description of requirements for Triple Play Services. How can a network scaled to support 100.000s of Triple Play customers, what kind of reliability is needed. What are the QoS parameters to be supported and which applications a most critical to packet loss, delay, etc. . Network Design: Detailed description of the end to end network design being deployed by Deutsche Telekom AG to provide Triple Play services to end customers. How to fulfill the requirements being identified and what kind of solution is needed. Description of the mechanisms to optimize rerouting times for Multicast traffic and to minimize packet loss. . IP-Multicast Support: Challenges for implementing IP-Multicast in an MPLS based core network to support IPTV. Where are the pros and cons of the different solutions? What kind of design was chosen to support a scalable, reliable and secure IP-Multicast transport? . Scalability: Description of the mechanisms to provide a highly scalable network; what are the key parameters regarding scalability and where are trade offs necessary (e.g. number of end users; number of parallel TV channels; routing protocol limitations). . Open issues and future activities: Triple Play Services in general are still in an early phase. Several issues where identified during the design phase which need to be supported in the future by standardization bodies and equipment vendors. The presentation will give a detailed description of future requirements and potential solutions from an ISPs point of view. This includes support for P2MP LSPs and TE, additional management and monitoring capabilities and scalability.
Operational Experiences of Telenors Next Generation Multiservice Network
Telenor's multiservice network based on MPLS technology has been in service
for about a year now. During this year, we have migrated the exsisting IP,
L3VPN and L2VPN customer base to the new network and also in process
migrating 2-3 new ASes of inqired companies into the network. We would like
to share the operational experiences related to technologies such as:
Kireeti Kompella, Juniper Networks MPLS was initially envisioned as a SP technology primarily for use in WANs. Here, the complexity of configuring and managing MPLS is less of a concern. However, as MPLS is used in wider contexts, such as metro networks and enterprises, simplifying configuration and management of MPLS devices and networks will play a key role in the adoption of MPLS in these contexts. This talk describes several techniques that can be applied to simplify MPLS provisioning. Many of these techniques are available today, but by and large, their use is piece-meal. The goal here is to bring these together in a coherent approach to reduce OpEx and advance the applicability of MPLS. Combining MPLS and 802.1x on the Access Node for Dynamic Service Provisioning Geraldine Calvignac, France Telecom IP/MPLS technology is already spreading from core to backhaul networks. The introduction of MPLS technology down to the access network would bring some benefits to support multi-services using MPLS VPNs, for retail, business and wholesale. The support of the MPLS control plane right from the Access Node (AN) could bring dynamicity and automaticity. Indeed, adding a new customer could be accomplished by setting up the correct service and interface parameters and then triggering MPLS signaling across access and backhaul or even core networks. In addition to the MPLS control plane, advantage could be taken of the possibility that 802.1x offers to load on the AN, the profile of a network service at the connection phase of the client. The AN being an 802.1x authenticator would be able to download the information necessary to trigger the connection of the client to a specific MPLS VPN that supports the service, wherever the client is located at the time of connection. The client\\\'s authentication would trigger, for instance, the setup of a pseudo-wire from the client Attachment Circuit, or the connection of a client logical interface to an emulated LAN using VPLS, or even the connection to a Virtual FIB i.e. VRF of a L3VPN service. Thus a dynamic provisioning would be enabled: once authenticated, a service would be triggered where the client is located at that time, without pre-provisioning of the network. This would also be of particular interest for nomadism. One Step Toward "Auto-Programmable" Networks Using MPLS Traffic Engineering A plethora of Traffic Engineering techniques have been designed over the last two decades to optimize network resources and meet traffic SLAs using a variety of protocols such as ATM, SONET/SDH, IP, MPLS and so on. Furthermore, a number of off-line and on-line path computation techniques and algorithms have been proposed that differ in term of efficiency, scalability, ability to quickly react to networks changes, ... Despite the variety of such available tools, over the years, Service Providers and large enterprises have been facing several challenges to efficiently engineer their traffic: (1) Traffic matrix determination especially in moving environments where topology is subject to changes due to failures and added network elements, apparition of new traffic types with different traffic characteristics and profiles (e.g. voice, video, peer to peer), mix of traffic types, ... (2) Ease of provisioning, (3) Ability to quickly (re)optimize the network and adequately reroute flows upon network topology changes and traffic load or traffic matrix changes And the list above is by no mean exhaustive. The aim of this presentation is to show a set of MPLS Traffic Engineering techniques that can be used so as to meet these requirements whereby: * The traffic matrix does not have to be known a priori and each router dynamically resizes its Traffic Engineering Label Switch Path (TE-LSP) according to the measured traffic load, * A set of one or more mesh(es) of TE LSPs can be automatically set up thus easing the provisioning task, * TE LSPs are dynamically routed in the network according to the traffic demand and available network resources. The presentation will not only cover the protocols and standardization aspects but will also include several simulation results highlighting the benefit of such an approach.
Policy-based Management for Multi-domain Networks Dimitri Papadimitriou, Alcatel Bell Resource reSerVation Protocol-Traffic Engineering (RSVP-TE) is the tool of choice for multi-domain Label Switched Path (LSP) provisioning in particular for single carrier environments. Techniques developed for multi-domain TE LSP provisioning includes, for instance, mechanisms such as LSP stitching and shuffling. TE reachability, availability, (some level of) traffic-oriented performance and blocking probability are the basic criteria for path computation, selection and establishment, in addition to the cost associated in traversing a given set of domains. In this context, policy could provide an important role in the decision process. For example (see Fig. 1), transit domain A may be more expensive but provide lower delay or loss than transit domain B while being less expensive but provide higher delay with the same loss than transit domain C. Likewise, a transit domain may give precedence to incoming requests from its directly connected customers compared to requests from other domains. In both cases path computation and selection will result in multiple transit domains where policies could be used to govern which incoming requests get the better service. However, since so far, policy has received very little attention while playing a fundamental role in facilitating multi-domain TE operations. Policy-based Management (PBM) enables network administrators to operate their network through rule-based policies that plays a key role during the service provisioning phases. The latter are then translated automatically into individual device configuration directives, aiming at controlling a network as a whole. This presentation will first motivate introduction of policing for traffic engineering operations in multi-domain networks. It will then describe the role of PBM as an enabling technology for controlling TE operations in large-scale distributed systems, and for enforcing operational coherency through policy rules. Policy mechanisms including implementation techniques, enforcement criteria, and their integration into multi-domain network environments will then be analyzed in depth. Use cases for policing combined resource and traffic oriented traffic engineering performance objectives during service provisioning will illustrate applicability of the PBM mechanisms. Optimization of Mobile Backhauling Architecture with MPLS Pseudowires Frederic Jounay, France Telecom As a fix/mobile service provider, France Telecom aims at using the same network infrastructure to carry all types of service (residential customer, business user and mobile operator). This strategy underlines the willing to reduce network costs (CAPEX, OPEX) by sharing a common network to support current services and easily integrating new services. To reach this crucial objective, France Telecom is evolving its network architecture to Ethernet/IP Access Nodes and IP/MPLS Metro networks. Unfortunately, 2G/3G mobile traffics keep using TDM/ATM from the time being and it is quite important to be able to carry such legacy flows over Packet-Switched networks without noticeable degradation of quality or impact on the mobile services SLA. The proposed solution relies on the use of an end-to-end Pseudo-Wire (PW) to let the recently deployed packet-oriented network agnostic to the legacy user protocol (TDM or ATM). Indeed PW is considered as a MPLS key technology allowing encapsulation of different non-IP traffics (TDM, ATM, Ethernet or Frame Relay). The first phase consists in setting up a SS-PW between each Mobile Base Station (BTS or Node B) and its corresponding Controller (BSC or RNC) which aggregates the encapsulated traffic. At near term, the access network is so far based on Ethernet/IP and Metro routers are MPLS-aware (namely LSRs). Hence from the access side the dedicated modem connected to the Base Station or directly implemented inside the BTS/Node B is a PE. The introduction of numerous PEs leads to a significant increase of signalling adjacencies. Moreover this situation is going to be more and more complex due to the on-going HSDPA and future HSUPA deployments. Indeed the more the bandwidth is increased, the more the number of Base Stations is increased. Hence when this massive deployment occurs, it will increase the number of signalling sessions to setup and maintain PSN tunnels and PWs. A solution to solve this scalability issue consists in using MS-PW. In this scenario the Metro Edge Node or the Access Node evolves to an S-PE role. The presentation first describes the evolution from SS-PW to MS-PW architecture, and then focuses on routing issue and on the three steps for MS-PW setup: - Provisioning at each PE of the Attachment Identifier which defines the binding of an Attachment Circuit with a PW - Discovery of the remote PE which terminates the PW. So far BGP is often proposed to advertise AI to PEs. In access network a trade off must be found between simplicity (static default gateway) and hard complexity (dynamicity with BGP or other protocols) - Signalling to setup the PW by distributing the associated PW label All these features have to be adapted to the access networks specificities where cost and hardware/software complexity have to be reduced as far as possible. A particular attention is also dedicated to protection scenarios based either on protection at PSN tunnel level (FRR with RSVP-TE) or at PW level (PW redundancy) to guarantee fast restoration in case of link or node failure. Mobile Backhaul aka "Wireless over Pseudowires" Multiple carriers have deployed backhaul networks for 2G and 3G mobile traffic using pseudowire emulation over IP/MPLS. This presentation describes: . The business drivers for using MPLS for 3G backhaul . The network architectures and technologies used . Project outcomes . Future possibilities
MPLS Based Services over PBT, from VPN to Pseudowires This session discusses the use of Ethertype multiplexing to combine PBT connectivity, channel associated signalling, and various MPLS and other adaptations to deliver services directly on Ethernet infrastructure. The session will cover pseudo wires, MACinMAC, IPVPN and broadband aggregation and illustrate the relative simplicity of combining MPLS and other services with emerging Ethernet capabilities. This greatly expands the options for packtizing the metro and gives operators greater flexibility in how they distribute functionality in the network as increased choices come to play in how the PSN is assembled.
Triple Play Services: Re-Inventing the "BRAS" This presentation will describe the set of functions that need to be optimized for the efficient and cost-optimized delivery of voice, video and managed data services over a unified service infrastructure. It will specifically discuss the new set of requirements induced by the introduction of BTV and VOD applications, and propose ways of optimally implementing all the relevant per subscriber and per service policies and functions where they can scale across the service delivery architecture.
MPLS and Carrier Ethernet Competing or Complementing Technologies Luis Aguirre-Torres, Corrigent Systems This presentation focuses on those network deployments where Carrier Ethernet has been chosed as the service enabling technology (e.g. Triple Play), and shows how in some cases MPLS control and forwarding plane can be used to enhance availability and manageability of the network. It also weights on the trade-offs of pairing both technologies in terms of complexity and expense. We use a live triple play packet transport network as example of an implementation where Carrier Ethernet offers the underlying service infrastructure and MPLS is used to provide the convergence layer for more than triple play services (i.e. TDM services). It explores the usefulness of MPLS OAM and MPLS FRR in next generation packet transport networks.
A Scalable MPLS Layer-2 Transport Solution Pasula Reddy, Redback Networks MPLS layer-2 point to point as well as point to multipoint networks faces scalability challenges in multiple dimensions: in the provisioning plane, in the LDP based signaling plane and in the data forwarding plane. This presentation will go into details of these challanges and look at possible solutions such as circuit creation on-demand, spoke connectivity as defined by hierarchical VPLS and MAC hiding, to solve some of the issues in large layer-2 MPLS networks..
Managing a Large MAC Address Space with VPLS Marc Lasserre, Lucent Technologies The VPLS forwarding paradigm is based on having a MAC address learning as a standard IEEE 802.1D bridge does. Since Ethernet MAC addresses are not hierarchical, it is not possible to perform address summarization, as is done with IP addresses. As a large number of customers use VPLS services - in the process, exposing potentially a large number of end-user MAC addresses in service provider networks - it is necessary to devise schemes to handle such Ethernet MAC address needs. In this presentation, the speaker will discuss various methods used, such as the use of routers as CPEs and MAC address limiting, to reduce the total number of MAC addresses that VPLS nodes need to manage. He will present a general approach to solving this problem, citing specific details on how HVPLS is well suited for such enhancements. GMPLS for Domain-Wide Ethernet Label Switching
This session is about
proposed extensions to GMPLS for controlling Ethernet and the specific label
paradigm termed "domain wide" label that is currently proposed as
part of the GMPLS Ethernet Label Switching (GELS) area of the IETF. It focuses
on the technology implications of controlling the standard Ethernet data plane
so that both legacy connectionless services as well as new connection-oriented,
"circuit" services also called PBT can be provided in a powerful layer 2
infrastructure. GMPLS is presented as the ideal technology to automate the path
placement and path management of these circuits.
Point to Multipoint LSPs represent the lastest advances in MPLS. P2MP-TE in particular is targeted at very critical and highly visible applications as Broadcast Video feeds and real-time financial data. Solid, scalable OAM is required to ensure these services. The presentation will discuss the technical issues and scalability challenges of P2MP OAM. It will then go on to discuss the emerging solutions that can meet these challenges. Challenges and Solutions for OAM in Point-to-Multipoint MPLS Adrian Farrel, Old Dog Consulting Ltd. and Zafar Ali, Cisco Over the last decade Service Providers have been deploying unicast MPLS Traffic Engineering to deliver VPN services, Fast Re-Route (FRR), bandwidth optimization, bandwidth guarantees, and explicit routing. Service Providers are rolling out, or have major plans to roll out, multicast services such as triple play and broadcast video over their existing and next generation MPLS/GMPLS backbone networks. These applications have driven the development of MPLS/ GMPLS extensions to create and maintain point-to-multipoint (P2MP) Label Switched Paths (LSPs). P2MP MPLS LSPs are at least as vulnerable to data plane faults or to discrepancies between the control and data planes as their P2P counterparts. For the Service Providers deploying services based on P2MP MPLS LSPs, the detection and specification of how to handle those defects is important because such defects may impact service level specification commitments for customers of their network. These service level requirements for P2MP applications are quite stringent. This presentation outlines the operational and maintenance (OAM) requirements of P2MP LSPs, and outlines solutions to meet the OAM requirements related to P2MP LSPs, which include LSP ping and traceroute functionality for both P2MP MPLS-TE and multicast LDP as well as new MIB modules. The presentation stresses the reuse of existing mechanisms used for P2P LSPs, and shows how we can apply them to P2MP MPLS TE LSPs in order to simplify implementation and network operation. The challenges and tradeoffs associated with approach are examined especially with regard to tracing data paths, testing packet replication, and managing P2MP LSPs with many destinations. The presentation explains the options to regulate the use of LSP Ping to reduce the number of responses and to prevent the ingress LSR from being swamped. Alternative mechanisms are described and put in the context of the current standardization activity within the IETF. LSP traceroute requirements are also equally applicable to P2MP LSPs, and we should be able to trace the route to a single destination or to all destinations of the P2MP tree. The presentation again stresses on the reuse of existing LSP traceroute mechanisms used for P2P LSPs, and shows how we can apply them to P2MP MPLS TE LSPs. Enhancements to P2P LSP traceroute are described that allow the recognition and correlation of branch points so that the ingress can construct the correct P2MP LSP tree. A third OAM requirement is for manageability through SNMP MIB modules. The presentation will briefly introduce new work within the IETF to extend the MPLS TE MIB for use in P2MP environments, and will describe how the existing LSR and LDP MIBs can be used in support of P2MP MPLS LSPs MPLS P2MP OAM.
High Availability in MPLS Networks Zubair Ahmad, Orange Business Services Achieving 5 9's availability figures means a downtime of ~5 min in a year. Enterprise applications have become more and more demanding in terms of network and service availability. This is mainly due to the ongoing Data & Voice convergence within the enterprise environment. There are no off-peak periods due to Network globalization. MPLS VPN Service Providers may face difficulties and challenges in scheduling maintenance windows with their enterprise customers. The concept of maintenance windows may be diminishing in the future. This presentation provides an overview of various High Availability building blocks in MPLS VPN networks. It provides a high level comparison among the key architecture choices i.e. Graceful Restart approach based on IETF specifications versus the vendor proprietary Non-Stop Routing alternative. It discusses about the optimization of Planned Maintenance operation relying on In-Service Software Upgrade (ISSU) and BGP Graceful Shutdown mechanisms. The presentation also describes the potential interactions among the high availability functions and fast failure detection techniques, such as BFD and suggests few approaches to allow for seamless co-existence of these functions in the MPLS networks. Adapting MPLS FRR to Create Resilient Rings Ahmed Abdelhalim, Foundry Networks Fiber rings play a key role in many service provider networks today. Over the years, several technologies evolved to address ring resiliency like SONET ring APS, some vendor proprietary resiliency protocols, and recently IEEE 802.17 (RPR). Yet, MPLS FRR can be a viable and attractive approach in this area. A ring topology presents an interesting challenge for MPLS FRR implementations. Yet, with some adaptation, MPLS FRR can be leveraged for implementing resilient packet rings. Such approach has the merit of requiring only MPLS capable hardware, i.e., no need for special resilient ring interfaces. In addition, it is a unified resiliency mechanism that obviates having resiliency at layer 1 or 2 to augment resiliency at the MPLS layer, and avoids having these multiple layers of resiliency running as ships in the night. The presentation will discuss how MPLS FRR can be adapted for ring topologies, as well as the pros and cons of such approach. Fast Protection in MPLS L3VPN Networks This paper covers fast reroute protection in MPLS VPN networks with a CE-to-CE scope. Firstly we provide some failure statistics taken from an operational MPLS VPN networks. We then list mechanisms that can be used today for fast recovery upon link and node failures in MPLS VPN networks and compare them in terms of recovery scope, performances, and scalability. We highlight current limitations for PE node protection. Finally we discuss new mechanisms that could be used so as to ensure fast protection upon PE failures and analyze their pros and cons. This presentation will discuss the technologies of implementing IPv6 and IPv6 BGP/MPLS VPN (RFC 4364) in MPLS network. First, we will give a brief over view on the drivers of IPv6 and IPv6 VPN deployment. We then discuss the technical implementation of IPv6, especially focus on IPv6 BGP/MPLS VPN over IPv4 MPLS backbone, based on IETF draft-ietf-ppvpn-bgp-ipv6-vpn: new BGP extensions for IPv6 VPN, routing information exchange, use of route reflectors, and Inter-AS/Inter-Providers IPv6 VPN for connecting customers IPv6 VPN sites across multiple ASes/Providers. The general service provider requirements will be addressed, as well as the challenges in design and deployment of IPv6/IPv6 MPLS VPN in IPv4 MPLS network, including security consideration, Inter-AS, and Multicast VPNv6 support.
Delivering MPLS Services Over L3VPN Kireeti Kompella, Juniper Networks and Kenji Kumaki, KDDI This document describes procedures that can be used to deliver MPLS services over Layer 3 Virtual Private Networks (L3VPN). Using these procedures, a VPN customer can signal MPLS Label Switched Paths (LSP) from Customer Edge (CE) router to CE router. Although a Service Provider (SP) network carries the Customer LSP (C-LSP) from one CE router to another, the C-LSP does not interact with the SP network\\\'s routing or signaling infrastructure.
GMPLS-based Multi-Layer Network (GMPLS-MLN) Concepts, Architecture and Services GMPLS-based multi-layer networks (GMPLS-MLN) is a promising technology to provide integrated control of packet and optical layers in service providers\\\' networks [MLN-REQ, MLN-EVAL]. This presentation reviews concepts, architecture, and services of the GMPLS-MLN. In this presentation, we start by reviewing the concepts and basic architecture of GMPLS-MLN. In the GMPLS-MLN, client networks, e.g. routers, are supported by a server network, e.g. OXCs and ROADMs, using a multi-layer GMPLS control plane. We demonstrate the flexibility of the control architecture by defining three interworking control models between client and server networks: overlay, peer, and augmented models. We then describe how these models address the architecture requirements of GMPLS-MLN. We discuss various factors influencing the architecture choice, including the following: - Trust model between the packet and optical networks. - Scalability requirements - Resource utilization, traffic engineering scope (single layer TE, multi-layer TE) - Deployment aspects - Fault recovery and stability We discuss how the proposed models can be realized using existing GMPLS protocol mechanisms and potential extensions to the GMPLS protocols, including key concepts such as ISC, FA-LSP, Virtual TE-link, and VNT. We conclude by describing how network services can be provided for client networks using a GMPLS-MLN. An IP/MPLS network is discussed as an example of a client network. The vertical and horizontal relationships for the data plane, control plane, and service plane are described. [MLN-REQ] Requirements for GMPLS-based multi-region and multi-layer networks (MRN/MLN) <draft-ietf-ccamp-gmpls-mln-reqs-00.txt > (work in progress), January 2006. [MLN-EVAL] Evaluation of existing GMPLS Protocols against Multi Layer and Multi Region Networks (MLN/MRN), <draft-ietf-ccamp-gmpls-mln-eval-00.txt>, January 2006 Multi-Layer Models: IETF vs. ASON/OIF
Work on multi-layer networks is still in progress for ASON. This talk
discusses some recent activities within the ASON standardization efforts to
support control of multiple layer networks. In particular, issues related to
multi-layer discovery and routing are addressed.
GMPLS Network Design Considering it's Control/Data Plane Resiliency Tomohiro Otani & Kenichi Ogaki, KDDI R&D Lab This presentation describes the guideline of designing the IP-based data-communication network (DCN) for the GMPLS out-of-band control plane from the point of actual network deployment considering the evaluated results of the control plane resiliency and scalability The IP-based DCN for the GMPLS control plane requires high availability without affecting the control plane operation by using, for example, fast convergence of routing and the bandwidth enough to convey control packets. However, as the various protocols on the GMPLS control plane and the DCN are complexly interacted with each other, the evaluation assuming an actual environment is indispensable in addition to the theoretical analysis. We utilized the testbed consisting of actual elements such as GMPLS capable routers, PXCs and WDM equipment with hundreds of kilometers transmission line in our laboratory environment. Although the disjointedness of SRLG must be considered between the control plane and the data plane, such disjointedness may not be completely ensured under the actual network environment. Therefore, as the worst case, both control and data planes between two sites are configured through the same transmission line. Firstly, the influence with resiliency in the IP-DCN to the GMPLS control plane was evaluated by the fast IGP failure detection mechanism of BFD. By speeding the convergence up to several hundred milliseconds by BFD, the LMP session as well as GMPLS neighbors could be maintained, and other GMPLS operations were not consequently affected except for the extra-restoration time due to the DCN routing convergence. Furthermore, the scalability of the GMPLS control plane over the IP-based DCN was evaluated by adding emulated nodes from a network tester, assuming that the GMPLS network scale grows. The traffic flow over a physical link of the DCN accommodating control channels between two sites were measured at the moment of the OSPF initial database synchronization by enlarging the network scale up to a few hundreds nodes. Traffic increase was much smaller than the GbE line bandwidth used in the DCN and it is verified that the bandwidth of even Fast Ethernet will be acceptable. The more detailed results will be presented in the conference, including data plane and control plane interaction effect, and those series of evaluation is quite useful for the deployment of GMPLS commercial networks.
Issues on GMPLS Inter-carrier E-NNI and a Prototype Node based on Linux The Generalized Multi-Protocol Label Switching (GMPLS) [1] is a set of network control protocols and an important controlling technology for next generation's photonic core networks. It is strongly required to support the end-to-end LSP under multi-carrier environment. However, GMPLS has some issues for interoperability between carrier domains to solve. The Interoperability working group (IWG) of Kei-han-na info-communication open laboratory studied such issues and we developed a prototype node of GMPLS E-NNI. In this paper, we introduce the prototype node as the E-NNI border node. We also verified that LSPs is established between carrier domains using this prototype.
Scaling MPLS in Multi-Area Networks Bruno Decraene, France Telecom Next generation networks use and provide VPN services and therefore need MPLS. The size of these networks is big and is getting bigger as IP/MPLS is spreading in aggregation networks, so several IGP areas are common. However extending LSPs across areas is quite a challenging task. This presentation covers the set up of MPLS LSPs between nodes belonging to different IGP areas, when Traffic Engineering is not required. We first present a case based on current service provider requirements. Then, we describe existing mechanisms to address this case and discuss their limitations . We next present a new solution using an LDP extension to ease the set-up of inter area LSPs and better address current networking issues. Finally we provide a high level comparison of the existing and new mechanisms in terms of performances, impact on IP routing, survivability, and scalability. Verizon Business MPLS Enabled Core Network Design Studies: using WANDL IP/MPLS View In today's economic environment, it is uncommon for a carrier to build a core network from scratch. Instead, the establishment of any core network is often the result of the natural evolution of an existing backbone network. Although this method often result in an extended network build out, and makes the network design more challenging, the benefits are readily apparent. Networks can realize immediate relife from bottlenecks. Also network's return on investment (ROI) will accelerate as new services are introduced due to the ability to use the latest technologies. This joint presentation from Verizon Business and WANDL, Inc. will address several network design challenges that Verizon Business faced when designing and building its MPLS-enabled Core Network. Each challenge will be discusses with examples, and the solutions will be presented. Verizon Business Design Challenge #1: Designing an MPLS core network that can sustain dual network failure - As the trend of network convergence onto a single MPLS core backbone continues to grown, and the transmission technology moves from the conventional ring based SONET architecture into linear Ultra Long Haul technology, designing an MPLS core network that has the ability to restore 100% of traffic during dual network element failure becomes critical. Verizon Business Design Challenge #2: Designing an MPLS network with backbone trunks of various sizes - Carrier backbone networks have traditionally used a tiered structure with one backbone size for the core backbone, and another bandwith size for edge-to-edge, edge-to-core trunks. In MPLS based networks, backbone trunks could range from OC12 all the way up to OC768. This presents a new set of MPLS traffic engineering challenges, especially during the network failures. Verizon Business Design Challenge #3: Designing an MPLS network with ever more sophisticated customer routing requirements - Verizon Business is experiencing ever more sophisiticated customer routing requirements for its traffic. MPLS-based network makes it possible for carriers to satisfy their customer's unique requirements; it also creates network design challenges. The design of the MPLS core network is an evolving process. While MPLS-based networks introduce many advantages, their inherent flexibility also introduces many challenges which need to be addressed during design process. MPLS-based Multicast - A Service Provider Perspective This presentation will discuss the current state of the MPLS-based multicast standards that are emerging from the IETF and other standards bodies. It will cover service provider requirements for MPLS-based multicast and how MPLS-based point-to-multipoint transport (P2MP LSPs) addresses these requirements in order to support current multicast based services. Other services (e.g. IPTV, broadband etc.) which could be supported by MPLS multicast in the future will also be covered.
Is your MPLS Network Ready for IPTV? Martin Lai, Agilent Technologies The introduction of Triple Play services such as IPTV represents a new and exciting time for consumers. The expected financial rewards from IPTV and other multicast services are triggering a remarkable evolution in Telco networks this decade. Network equipment manufacturers and service providers are now faced with the challenge of ensuring their devices and networks can meet scalability, performance and reliability expectations for multicast services including IPTV. There are several options being proposed within the IETF to ensure efficient multicast service delivery over MPLS networks. In this presentation we will discuss several of the mechanisms being proposed including Point-to-Multipoint (P2MP) extensions, multicast VPNs (mVPNs) and multicast VPLS. We will describe the test methodologies involved in validating the interoperability, scalability and performance of devices and networks implementing these approaches. You will also be presented with key test scenarios to evaluate and compare the different mechanisms available for efficient multicast delivery and measure the impact they have on the end user viewing experience. Enabled Operational Modes Using PCE-Based Architecture The emerging PCE protocols open the door for valuable new operational modes. The PCE is for example an efficient enabler for vertical integration (between layers) and horizontal integration between domains. This presentation explore some innovative modes such as Interdomain TE scenarios, enhanced overlay models using PCE-to-PCE enriched communications, provisioning and recovery performances improvments. Beyond the protocols mechanics, the presentation aims at demonstrating the potential key role of the PCE-based architecture which may foster undeployed technologies existing but sometimes considered too complex. The combination with a Policy Based Management architecture is thus falling into the scope of this introduction to PCE-based architecture applicability.
PCE-Based Inter-Layer Traffic Engineering in MPLS/GMPLS Networks This presentation describes a framework for the PCE-based path computation architecture to inter-layer MPLS and GMPLS traffic engineering. It provides suggestions for the deployment of PCE in support of multi-layer networks. This document also describes network models where PCE performs inter-layer traffic engineering, and the relationship between PCE and a functional component called the Virtual Network Topology Manager (VNTM). We also present our PCE prototype system that demonstrates inter-layer TE in MPLS/GMPLS networks and its experimental results.
Peter Busschbach, Lucent Technologies In this presentation, Lucent CTO Office Director Peter Busschbach will discuss work currently underway in three different standards bodies and other forums regarding network management and resource control. The three activities are: · The IETF's work on Path Computation Element (PCE), which intends to develop solutions for situations in which the computation of an end-to-end LSP is difficult to achieve using existing mechanisms. This is the case, for example, with LSPs that cross operator boundaries. · The ITU-T's work on RACF (Resource Admission Control Functions), focused on providing resource management for sessions that are established via IP Multimedia Subsystem (IMS) or other service control functions. · IPsphere's work on a service management infrastructure to be used between service providers and between customers and service providers. The presentation will describe areas of difference and overlap among the three and conclude with a proposal on how the three can be aligned to create a single powerful solution.
Increasingly Complex Demands for Path Computation in Converged Networks The trend for convergence of triple play services over IP/MPLS networks places increased demand for complex service placement algorithms within the network. It is no longer sufficient to throw bandwidth at the problem as Service Providers must extract better value from their deployed resources and must distinguish the traffic for different services they offer. High specification SLAs force network operators to think carefully about the bandwidth and quality guarantees that they are making, and require sophisticated network planning to ensure that a variety of conflicting constraints and demands are met for each customer service. This presentation examines how the requirements of voice, video and data traffic map to different path computation requirements for minimized cost and delay, maximized resilience, and guaranteed bandwidth. The different objectives of Fast Reroute and end-to-end protection will be discussed and the consequences for network planning and path computation will be highlighted. The speaker will identify how different computation techniques can achieve suitable paths only under specific conditions, and will go on to examine the benefits and risks of various network reoptimisaiton techniques in traffic engineered MPLS networks. Dr. Jay Perrett is Chief Science Officer & Head of R&D with Aria Networks Limited where he works on the development of advanced machine learning solutions to complex path computation and network reoptimization problems. Before joining Aria Networks, Dr. Perrett was CTO of Applied Insilico and applied machine learning techniques to the drug discovery and biological research industries. With a long history in Academia and the telecommunications industry, spanning UK and US Universities and Research institutes and companies including Lucent Technologies, Redback and Movaz Networks, Dr. Perrett is ideally placed to bring together the advanced problem-solving abilities of next-generation machine learning applications with the high demands of path computation in complex and growing converged Next Generation Networks.
Open Issues in Hierarchical Recovery Payam Torab, Lambda Optical Systems Generalized Multiprotocol Label Switching (GMPLS) is recognized as a universal control plane with wide applicability to various switching technologies, ranging from packet switching in IP networks to photonic wavelength and waveband switching in optical WDM networks. Today, this wide applicability is being demonstrated by the emergence of multi-region networks, where a single instance of GMPLS control plane controls multiple switching technologies (regions) at the same time. In this presentation, we examine a few areas around fault management and recovery in multi-region networks. In particular, we study 1) GMPLS applicability to end-to-end recovery in these networks, and potential areas for protocol extension, and 2) the implications of optical transparency on fault management, in particular fault localization, in case one or more network regions employ all-optical switching. Regarding GMPLS applicability to recovery, we look at the standard GMPLS end-to-end recovery techniques available today, and how they can be used to achieve end-to-end recovery in multi-region networks. We will discuss the following points and more: . Since end-to-end recovery options in one region correspond to link recovery options in another region, there has to be a one-to-one correspondence between end-to-end and link recovery options in GMPLS, which is not the case at this time. In particular, the notion of link restoration is currently missing from the GMPLS standard. . Assuming the GMPLS end-to-end and link recovery options are unified, we discuss the potential need to signal or distribute additional information to coordinate recovery at different regions. In particular, recovering an H-LSP can affect the diversity of paths that use the H-LSP as a link, a fact that has not been studied with sufficient detail so far. A related problem to recovery in multi-region networks is the implications of optical transparency on fault management. Multi-region networks with an optical core are moving towards transparent optical switching for scalability and economic reasons, but this also makes SONET-style performance monitoring and fault management at each node more difficult. Network performance monitoring can still be done at each node, but often with limitations. Of particular interest is the complexity of localizing soft failures such as BER degradation, and how one can benefit from the vertical hierarchy of multi-region networks to make fault localization faster and more efficient. We share lessons from our experience with a multi-region network that does not have one, but two layers of optical transparency: A DWDM network with multi-granular optical crossconnects (MG-OXCs) that support transparent switching at both waveband and wavelength levels. Here, we show that the GMPLS fault localization performance (in terms of speed and usage of monitoring resources) for soft failures can be significantly improved if the vertical hierarchy of multi-region networks is considered in the design. We conclude the presentation with final thoughts on applicability of GMPLS as a fast, reliable and robust control plane technology for multi-region networks.
Invited Presentation: Future Directions in MPLS QoS Since the early days of the MPLS there have been great expectations that it could improve the QoS capabilities of IP networks. In reality much of the MPLS QoS architecture simply mimics the capabilities that have been developed for IP, such as Diffserv. However, MPLS has had a significant impact on QoS for a variety of reasons. MPLS VPNs have turned out to be the catalyst for a great deal of Diffserv deployment - indeed the majority of service provider Diffserv deployments are in the context of MPLS VPNs. And the decision to use RSVP as the signaling protocol for MPLS-TE has provided a key building block for scalable admission control - an extension to the IP architecture that seems likely to be increasingly important with the advent of large scale video over IP. This talk will explore the future direction of MPLS QoS, including the development of more scalable admission control techniques based on congestion marking, and the challenges of delivering end-to-end SLAs over the networks of multiple independent providers.
|
|
|||
|
|
||||
