This tutorial focuses on the emerging industry standards/agreements supporting multicast in BGP/MPLS VPN and VPLS networks. Key options, benefits and tradeoffs are explored for solutions supporting several prevalent service provider and enterprise applications. Key existing and emerging industry standards/agreements are referenced.

back to tutorials ^

Widely deployed broadband is coming to mobile networks in the form of High Speed Packet Access (HSPA) Evolution Data-Optimized (EV-DO) and WiMAX networks enable new mobile applications and services. Some of the biggest challenges that will result from this data and bandwidth explosion will be in the backhaul and aggregation functions that take place in the RAN. Historically RAN backhaul networks have been voice centric, low capacity and a source of major operational expense for mobile operators. New infrastructure technologies need to be able to support legacy access types, leverage cost points of newer transport types such as Carrier (Metro) Ethernet, and leverage the high availability, resiliency, and Quality of Service capabilities of IP/MPLS technology.

This tutorial will examine the principal drivers for a new IP/MPLS backhaul transport infrastructure that accommodates the scaling needs of the evolving mobile networks. Key challenges, options, benefits and tradeoffs are explored for solutions supporting several prevalent applications. Key existing and emerging industry standards/agreements are referenced.

back to tutorials ^

The Broadband world is currently experiencing a major transition in mindset, as Service Providers are looking for ways to migrate from the legacy technologies that have their roots in dial-up access to next generation methods of providing access. PPP was a fantastic technology in its day for identifying individual subscribers and applying individual policy control, but it is inefficient and doesn’t scale well to today’s Quad-Play technologies. Today’s deployments must decrease OpEx costs, provide new opportunities for increasing ARPU, and shorten deployment cycles.

This tutorial will discuss Next Generation concepts around applying subscriber polices directly to native IP sessions and how Ethernet and MPLS architectures are becoming increasingly more important to these models. VoIP, IPTV, VoD, and wireless access must be seamlessly integrated, while connecting and disconnecting from the appropriate networks and VPNs. We will explore the requirements of next generation, access agnostic Service Provider networks, with special attention towards QoS, Admission Control, Multicast, and MPLS aggregation.

back to tutorials ^

MPLS became a pivotal technology for service providers in the late 1990s primarily due to the success of MPLS-based layer 3 VPNs. This technology succeeded largely by replacing layer 2 technologies such as Frame Relay due to the superior scalability of the L3VPN approach. At the same time, demand for L2 services did not evaporate, and the realization that L2 services could be tunneled over MPLS led to a vision of converged networks based on MPLS. That vision is now very much a reality with MPLS being deployed in the majority of service provider networks. The interesting question now is whether MPLS will continue to be the technology of choice for service providers or whether some new technology can knock it from its perch. In this talk we will consider the recent crop of challenger technologies (such as PBB-TE) and whether they are likely to compete with or complement MPLS. We will also examine some of the architectural features that have made MPLS very effective at evolving to accommodate new services, and suggest reasons why MPLS will continue to flourish for at least another decade.

back to program ^

Available Soon

back to program ^

Our objective in this presentation is to present a case study on carrying broadcast video in high definition format using RSVP P2MP in the core and PIM-SSM at the distribution layer in Cox’s network. Our solution will meet the video stringent QoS SLA like packet loss rate, peak-peak jitter, and to minimize packet loss rate during link or node failure, we will be using facility-based protection, and to protect against application failure, we will be using spatially diverse streams such that the primary and secondary paths are maximally disjoint. Our solution will keep vital statistics like packet drop, and jitter per single program transport stream.

back to program ^

The proliferation of VoIP, IP video and other collaboration services across enterprises puts increased pressure on the ability to understand, evaluate, anticipate and respond to the demands of the real-time applications across a customer's network. This situation is intensified through the convergence of wide-area networks (WAN) and the deployment of IP MPLS technology. As MPLS and its primary benefits become more of a commodity, how can service providers differentiate themselves and avoid the price erosion? Presentation Proposal: Increasing customer loyalty and stemming churn is crucial. The solution is to create market disruptive solutions that will displace well-seated players. The MASERGY vision is the implementation and extension of network-embedded applications that simplify network management for the enterprise and provide profitable growth opportunities for service providers. Simplified services embedded in the VPN eliminate the need for software, hardware, integration, validation and maintenance on behalf of the enterprise. Businesses get real-time access to network traffic statistics and view application performance and data flows across a company's MPLS-based WAN. Using a web-based portal, businesses can set threshold alarms, verify service level agreements (SLAs), view granular detail, analyse network traffic and generate management reports.

back to program ^

After the establishment of MPLS in most networks, did the promise of reduced operational cost materialize ? This talk will explore some of the perceived MPLS complexity, and some of the proposed "simple and cheaper" solutions. MPLS, in particular is a technology that can be very simple, or extremely complex. Some of the more simple technologies recently proposed to replace MPLS are already evolving toward by slowly adding a duplication of the existing MPLS technology. Deployment of the more complex MPLS technology needs to be carefully evaluated against the operational cost. MPLS is still the best and most cost effective technology for a simple network.

back to program ^

Most MPLS applications require end to end LSPs between all communicating PEs. As MPLS networks grow, particularly with more intelligence pushed to the network edge, the number of PEs in networks has risen substantially. Service Providers need a means to allow both their IGPs and LDP state to scale.

A number of proposals have been made for limiting the IGP and/or LDP state within the network core. This presentation will compare and contrast those options and their impact on
scaling MPLS applications like L3VPNs.

back to program ^

The problem of scaling a network running a link-state IGP has been adequately addressed using the notion of multiple areas (or levels). However, while there has been some discussion of how LDP can be run over a multi-area topology, there are still open questions on how LDP itself can be made to scale well.

This talk presents the concept of Hierarchical LDP (H-LDP) in some detail, and shows how H-LDP scales well in multi-area topologies. It also describes how H-LDP can be used to solve other problems, such as running LDP in multi-AS environments (one example being inter-AS option C for IP VPNs and VPLS.)

back to program ^

The somewhat alarmist title of this talk indicates a problem being encountered by some of the larger network operators as they consider rolling out MPLS Traffic Engineering to support full edge-to-edge connectivity at different service levels across their core IP/MPLS networks.

Simple math shows that a core network with 1,000 PEs might require as many at 999,000 LSPs to achieve a full edge-to-edge mesh. The use of separate LSPs for different functions or qualities of service provides a linear multiplier.

There is nothing intrinsic in MPLS-TE routing or signaling that prevents support of such large numbers of LSPs, but the impact on core routers can be considerable. Problems associated with RSVP-TE "soft state" have largely been discounted through the refresh reduction mechanisms of RFC 2961, but not enough consideration has been given to the sheer volume of information that might have to be held at a transit LSR, not to the quantity of LSPs that might need to be managed by a management application responsible for a router at the core of the network.

This talk will outline the service types that might motivate fully-meshed edge-to-edge connectivity. It will examine the scaling properties of two key network topologies – the snow-flake network and the ladder network. And it will deduce the number of LSPs that might be seen at a core LSR in these networks and explain why such large numbers of LSPs may cause problems.

But MPLS-TE scaling can be improved considerably by the use of hierarchical LSPs as described in RFC 4206. The speaker will quantify the benefits of hierarchical LSPs in the context of the network models already described, and will explain some of the additional management overheads created by this solution.

As an alternative solution, the presentation will introduce multipoint-to-point (MP2P) TE-LSPs. Such LSPs can be shown to make a significant impact on the scaling properties MPLS-TE in particular in ladder topologies. MP2P could be achieved through relatively simple extensions to the existing RSVP-TE signaling protocol.

back to program ^

MPLS VPN management has historically been challenged by significant technical challenges to gaining full edge-to-edge and network-wide visibility into VPN routing and traffic dynamics, including insight into the core network’s ability to deliver on CoS and understanding of per VPN, application-specific traffic. For example, the need to collect traffic information at the PE interfaces has made it has been practically impossible to scale the collection and processing of sufficient traffic and routing information throughout a large service provider MPLS VPN core and edge network to provide visibility into per VPN traffic dynamics. Furthermore, the virtualized routing nature of MPLS VPNs makes it challenging to manage the routing integrity of each VPN.

This session will describe a new approach to overcoming the historical scalability barrier in MPLS VPN network management, so that service providers can gain an unprecedented level of service and core network traffic visibility. The session will address:

The State of MPLS VPN and Core IP Traffic Analysis

• The scalability barrier: the technical and practical limits on traffic visibility in MPLS VPN networks
• Additional service challenge of analyzing service traffic in the virtualized MPLS VPN routing infrastructure
• Service assurance challenges stemming from lack of core network and per-VPN traffic visibility

Breaking the MPLS VPN Traffic Analysis Scalability Barrier

• An approach to baseline and monitor the routing integrity of all VPNs in the network
• A method for achieving full visibility into per-VPN and the entire core network’s service traffic from the P to PE perimeter
• How the new approach differs fundamentally from previous approaches in terms of technology and visibility
  • Control plane as foundation for data plane analysis
  • Beyond columns and rows to topology-based correlation
  • A time-moveable model to understand service traffic’s past, present and future

back to program ^

The trend toward convergence of multiple network services onto a common network service infrastructure underscores the need for ultra high reliability to ensure 7/24/365 access to applications and resources for businesses. MPLS based services including Carrier Ethernet VPNs based on VPLS, IP VPNs and pseudowire services leverage built-in MPLS resiliency to ensure high availability. Complementing these with new access protection techniques, non-stop service/Non-stop routing capabilities of the network elements and service–aware network management combine to deliver the protection, restoration and reliability requirements critical to businesses.

The presentation will discuss current resiliency capabilities for MPLS services, their applicability to today’s needs, management considerations and the state of standards.

back to program ^

Now that VPLS has become prevalent for business services, it is critical to provide SLAs that comparable or better than IP VPN services.
Management of redundant customer access connection is key to provide this kind of service.

This presentation will compare the pros and cons of the features available at the PE-PE and PE-CE level, including :

• dual homed spoked MTU in H-VPLS (LDP vs BFD, Mac Withdrawal message ...)
• VPLS PE running STP
• Dedicated VPLS instance for STP (management PW)
• loop avoidance using Mac Move detection We will conclude with future developments in the area.

back to program ^

As BGP/MPLS VPNs grow in popularity, more critical applications are provided over this infrastructure. Although such deployments are designed with appropriate redundancy, not all types of failures can be addressed this way. For example, if the label switched path from the ingress PE to the egress PE is broken in the data plane, and if this failure is not reflected in the control plane, the PE router would continue to attract traffic from the CE router, only to drop it due to the faulty LSP. MPLS OAM mechanism can provide fast detection of such failures, however, it is equally important to have fast failure recovery.

In this talk we will look at different options for automatic fast reaction to a failure, and see what type of deployments and failure scenarios would benefit from the fast failure detection coupled with automatic recovery mechanisms.

back to program ^

Over the last few years service providers express interest in evolving their transport infrastructure towards packet-based technologies. In this presentation we examine two possible technologies to accomplish this goal: Provide Backbone Transport (PBT), and MPLS Over Ethernet (MPLSoE).

back to program ^

Public Ethernet services are exploding in popularity due to the ubiquity and low cost of Ethernet interfaces on end-customer equipment, the relative simplicity and almost universal use of Ethernet-based networks within enterprises, and the emergence of carrier-class equipment and standards capable of supporting both metro and wide area carrier Ethernet networks. However, both the standards and vendor communities have produced a wide range of choices for how a service provider could deploy Ethernet services, including, but not limited to, native Ethernet-based services using Q-in-Q encapsulation with 802.1ad Provider Bridge Networks (PBN); native Ethernet-based services using Mac-in-Mac (801.1ah) encapsulation with Provider Backbone Bridges (PBB); native Ethernet-based services with PBB-TE (801.Qay); MPLS-based Virtual Private LAN Services (VPLS); Ethernet transport over next-gen SONET/SDH; and Ethernet transport on DWDM optical transport platforms. The purpose of this talk is to compare and contrast these various alternatives to see where each might make the most sense for a service provider to deploy.

back to program ^

Unwanted traffic is all the traffic that we don't really want to transport over the Internet. There is an ongoing change of the nature of this traffic, from being just kids that played nasty tricks to severs and routers, we have seen develop first to something that threatens business critical traffic, and beyond that to an entire business of its own.

back to program ^

This talk covers several recent developments in Ethernet technology, namely Provider Backbone Bridges (PBB, or IEEE 802.1ah) and Provider Backbone Transport (PBT, aka PBB-Traffic Engineering or PBB-TE IEEE
802.1Qay).

Ethernet has been evolving to allow scalable provider backbones. PBB introduces the concept of recursion and VLAN partitioning to support many virtual customer networks with great scalability and improved security. This reduces the need to learn customer MAC addresses and virtually eliminates the possibility of VLAN exhaust. PBT capitalizes on certain aspects of PBB, providing a simple, traffic engineered point-to-point and point-to-multipoint capability that delivers on the resiliency benchmark levels that providers require.

Because PBB and PBT technologies are backwards compatible, this evolution of Ethernet, toward greater scale and reliability with traffic engineering capabilities, is ready today. A final aspect of this discussion will include examples of PBB and PBB-TE network applications.

back to program ^

This presentation will discuss the solutions of Carrier/Metro Ethernet services. The focus is on network design considerations and practical deployment scenarios. The assumption is that Service Providers require scalability, reliability, QoS, P2MP/MP2MP, Multicast, and OAM.

The discussion will include the following:

• Short overview of the building blocks - technologies for Carrier/Metro Ethernet implementation, H-VPLS, 802.1ad, and 802.1ah (PBB).
• Architecture design options and considerations: Single core vs. multiple cores; single service vs. multi-services; small scale vs. large scale; static vs. dynamic; more layers vs. fewer layers.
• The use case studies, real world deployment scenarios.

back to program ^

Transport networks are on the verge of significant change. Residential bandwidth is exploding, driven from high speed internet access and IPTV, pushing fiber capacity far into the access networks. Enterprises continue to ask for more, with 10/100, GbE and 10GbE services on the rise. Additionally, greater demands for Ethernet services create the need for wholesale private line packet access networks, ultimately leading to ubiquitous Ethernet connectivity in the WAN.

The question is sometimes asked – why not just connect routers together via “dumb” WDM transport? In other words, why have a value-added transport network in addition to the service switching network? Although transport networks exist separate from service networks for several reasons, the fundamental reason is cost - both in capital and operational expenses. In general, per port costs of transport equipment are significantly lower than the equivalent costs of service switches. Additionally, transport is typically less complex to operate and inherently more reliable keeping operational expenses low. Given that transport equipment is located in many times the locations of service switches, these costs differences are significant.

Applying these arguments to packet transport, it is generally known that Pseudowires over MPLS (PW) are the packet transport mechanism of choice for carriers today. PWs offer the scale and reliability features that allow it to be used in traditional transport applications. However, to date PWs have been deployed mostly on router platforms. So the relatively simple application of long lived, point to point transport is being burdened by the costs of high value-add service switching features such as layer 3 VPNs and general layer 3 switching. This paper proposes that PWs, as the next generation of transport technology, will drive the deployment of packet transport equipment that is lower cost, more reliable and easier to operate than current alternatives.

back to program ^

An increasing number of mission critical applications are carried over IP/MPLS networks. A cost effective choice for interconnecting the IP/MPLS ports is switched ethernet. Fault management for ethernet is being standardized by the IEEE 802.1ag CFM working group. This technology is pivotal to deploy carrier class ethernet networks. When a carrier chooses to combine IP/MPLS and carrier class ethernet in its network, it is important to leverage the synergy between the two technologies to get the maximum benefit. In this presentation we explore specific areas of synergy such as 1. CFM triggering FRR and a comparison with BFD 2. Interworking of MPLS OAM such as ping, traceroute and VCCV with CFM-802.1ag to localize the faults in VPLS networks.

back to program ^

Deutsche Telekom runs one of the largest Triple Play Network on top of an IP-Multicast enabled MPLS based infrastructure. Recently new Ethernet based transport mechanisms like PBB-TE for backbones gained a lot of interest and are heavily discussed by vendors and ISPs.

The presentation provides an in-depth view from a service provider’s perspective regarding recent developments in the area of Ethernet based backbone transport. It compares mature and widely deployed technologies like MPLS with the new emerging approaches.

back to program ^

Many SP are providing L3VPN service. Their customers also want the transport of IP multicast. In this circumstance, standardization body and vendors develop the solution to provide multicast L3VPN services. Softbank Telecom has Multicast L3VPN service in early phase of that circumstance. BGP P2MP solution is key of Softbank Telecom's Multicast L3VPN service. This presentation will illustrate the experience of Multicast L3VPN service and its advantage of solution.

back to program ^

This presentation describes the MPLS-based network architecture for a high quality media transmission service to broadcasters. It is a challenge to employ such services over packet centric networks with guaranteed QoS, low latency and jitter, since a well-established ATM or TDM-based networks is currently utilized to transport live video and recoded video from a local station to a central station for broadcasting and editing. The introduction of MPLS protocol suites is one of candidates to envision such services over packet centric network while converging with other application services.

In this presentation, we describe our vision on future media transport network architecture using MPLS technology. Firstly, we overview our plan of the media transmission service with guaranteed QoS, bandwidth on-demand, converged with data transfer and voice. Considering broadcasters requirements, we also indicate our designed MPLS network architecture introducing Diffserv aware Traffic Engineering (TE) and policy based routing using BGP to prioritize multiple classes of services as well as a low latency IP-based codec with 1+1 protection based on route diversity. In order to provide reliable services, we developed the network management system not only to activate media transmission services by provisioning Diffserv-TE LSPs and IP-based application flows on top of those LSPs but also to facilitate operation, administration and maintenance (OAM) such as network topology discovery, hierarchical inventory management between LSPs and application flows, and alarm management. MPLS-based OAM functionalities are under consideration to mitigate with the NMS.

We indicate the evaluated results of the media transport network testbed to verify the end-to-end media transmission services are compatible with those over ATM or TDM-based network. Additionally, recent nation-wide trial results considering actual network deployment are also covered during conference.

back to program ^

Multiple alternative mechanisms are studied by the IETF l3VPN working group for multicast in BGP/MPLS VPNs. The aim of this paper is to present the different alternatives proposed for the different building-blocks of the mVPN architecture, and examine them in the light of operator and VPN customer’s requirements, focusing on their ability to be a good standard. A use case is presented showing how using the BGP-based approach for the different building blocks can address the needs for consistency and adaptability of an operator deployment.

back to program ^

Service providers offering multi-play residential services must quickly take on the challenge of deploying video services over broadband connections. While it is not cost effective for a service provider to engineer a congestion free residential metro or an aggregation network that provides Broadcast TV and VoD content simultaneously, network congestion from oversubscription or network failure can dramatically degrade the quality of experience for all video service subscribers.
A cost-optimized and standard based solution not only prioritizes quality demanding traffic it also rejects new bandwidth consuming sessions when the network approaches congestion thresholds and intelligently re-uses or re-captures bandwidth resources as sessions end - dynamically allocating and limiting network resources as necessary. To implement this intelligence, the network requires per flow admission control mechanism that can guarantee tight SLA for end-user quality of experience.
This presentation reviews options for just such a cost-optimized architecture. It highlights the pros-and-cons of RSVP over Diffserv and the recently standardized mechanisms for "RSVP Aggregation over MPLS TE/DS-TE", which allows deployment of the RSVP VoD CAC approach in MPLS networks, L3VPN and L2VPN. Particular focus is given to the benefits of inherited MPLS TE capabilities (aggregate bandwidth reservation, fast reroute, constraint based routing) in the solution.

back to program ^

There has been an increasing trend among enterprises to rely on MPLS WANs to deliver business critical applications across their dispersed sites. Accordingly, enterprises are expecting their service providers to offer SLAs based on network performance. Yet setting up such SLAs can be challenging and several factors should be considered.

This presentation will show attendees how to:

• Set up SLA performance metrics that are meaningful for the enterprise based on quality of experience and end user productivity.
• Focus SLAs on the portion of the network that service providers control
• Implement SLAs that are easily verifiable by enterprises through per packet measurement of network flow metrics
• Ensure that MPLS infrastructure is proactively controlling service levels

Most importantly, attendees will learn about a new breed of MPLS SLAs called Application-centric SLAs. Leading telcos and MSPs have embraced these application-centric SLAs, which are based on quality-of-experience indicators and technology that provides an exhaustive, per packet measurement of network metrics.

back to program ^

The additional traffic counters made available by deploying MPLS are helpful but not sufficient for calculating traffic matrices (the amount of traffic between source and destination pairs) in a network:

• RSVP LSPs would provide end-to-end counters only if there were a full mesh of LSPs. Most RSVP implementations, however, are based on a partial mesh (for example, a mesh in the core).
• LDP LSPs are often setup end-to-end in a network but the leading router vendors' implementations do not provide sufficient data for measuring end-to-end traffic.
• The statistics gathered from LSP measurements are often not consistent with interface measurements. This happens when the two sets of measurements take place on different time scales (for example, it might take 15 minutes to do a round of collection on 10,000 LSPs versus 2 minutes for 100 interfaces) and/or because the individual LSPs exhibit higher traffic variance than the aggregate interface measurements.

In this talk, we outline a unified method for obtaining a coherent traffic matrix from inconsistent and incomplete statistics. The method has a firm theoretical foundation in information theory and has been verified and deployed in many operational networks as part of a commercial product.

back to program ^

Reconfigurable Optical Add / Drop Multiplexers (ROADMs) enable service providers to maximize the transport capacity of their fiber plant and provide a low-cost distributed metro aggregation and transport infrastructure. ROADM-based optical systems, which can be deployed in single ring or multi-degree interconnected rings, are used in metro networks to multiplex and add / drop wavelengths and aggregate sub-wavelength traffic. Current ROADM deployments typically support at least forty 10 Gbps wavelengths. 40 Gbps wavelengths are also possible, and 100 Gbps wavelengths are expected in the future. ROADMs are used to transport circuit-switched SONET, digital video, and packet-based services. Packet based services generally require ROADM transport of 1 Gbps and 10 Gbps Ethernet ports. For packet-based services, including triple play, an important consideration is how to multiplex many lower rate lightly filled 1 Gbps Ethernet ports into 10 Gbps wavelengths and onto highly filled 10 Gb/s Ethernet handoff ports. Similarly there will be a future requirement to multiplex lightly filled 1 Gbps and 10 Gbps Ethernet ports into 40 Gbps and in the future 100 Gbps wavelengths and highly filled Ethernet handoff ports. There are two possible approaches, circuit switched multiplexing and packet switched multiplexing. This presentation will detail the advantages of using integrated MPLS based packet switched multiplexing of Ethernet ports in ROADM-based optical networks. Circuit-switched Ethernet port multiplexing is facilitated by using a G.709 Optical Transport Network (OTN) digital wrapper with G.7041 GFP encapsulation. This provides a standard solution, but it is bandwidth inefficient for 1 Gb/s Ethernet transport as the smallest G.709 container is a 2.488 Gbps OPU1. Efficiency can be improved if the ROADM also includes a SONET STS-1 fabric. For this case the 1 Gbps Ethernet port can be GFP-F mapped to a virtual concatenation of 21 STS-1s, which is multiplexed into an OC-192 and then into a 9.995 Gbps G.709 OPU2. However the circuit-switched approach does not support statistical multiplexing of the packet traffic itself, and therefore does not provide aggregation from lightly filled lower rate ports onto highly filled handoff ports. Without packet multiplexing, underutilized 1 Gbps Ethernet ports that enter the optical network are transported and handed off to the router or BRAS as 1 Gbps Ethernet ports, resulting in inefficient use of router or BRAS ports.

back to program ^

GMPLS protocols for control of optical networks are now mature enough to claim deployment. However, one of the major obstacles for GMPLS deployment faced by the service provider is “how to operate and maintain a GMPLS based IP + Optical network”. This problem is more acute when the network is based on all optical nodes, and the fact that data and control channels are separate. This presentation details these challenges faced by GMPLS technology, along with solutions to address them. The presentation starts by outlining applicability of MPLS OAM techniques to GMPLS networks and highlights the associated gaps. E.g., for a hierarchical LSP with transiting through a non-PSC network, LSP Ping can be used to localize failures to a GMPLS tunnel, but it cannot be used to localize the broken leg of the GMPLS LSP. Similarly, MPLS traceroute display an entire tunnel as a single MPLS hop and cannot be used to trace a GMPLS LSP. The presentation highlights these gaps and discusses possible solutions for GMPLS OAM. In a network consisting of optical cross connects (OXC) and DWDM gear, there are requirements to perform commissioning of the links, e.g., monitoring and power-checking, before GMPLS LSP can carry user’s traffic. This requirement presents some issues that hinder GMPLS deployment. E.g., alarm reporting needs also be disabled during the commissioning phase. If the commissioning is done for a specific route that GMPLS LSP takes and when the LSP is rerouted due to a failure or reoptimized, the LSP needs to be commissioned before it can carry user traffic. These aspects need to be considered in implementing protection and restoration, or in realizing make-before-break procedure for reoptimization. In addition to LSP commissioning requirement mentioned above, there are additional operation requirements on GMPLS technology. E.g., on one hand, GMPLS needs to communicate alarms along the path of the LSP within the optical network; on the other hand it needs to be able to provide alarm suppression, aggregation and correlation to avoid undesirable actions at multiple layers of the network. Similarly, due to traditionally heavy use of management plan in control of optical network, GMPLS needs to be able to provide more management functions that its packet counterpart. It is also an operational requirement that IP applications running on top of GMPLS LSP should not need to boot strap in the event of a failure in the optical network or during reoptimization. Trace route and LSP ping are also important OAM requirements. GMPLS protocol extensions provide mechanisms that can be used to address the above-mentioned requirements. Nonetheless, there are implementation gaps. This presentation outlines how GMPLS is able to address the OAM requirements, what are implementation gaps and key interop issues. It summarizes the major obstacles for large scale deployment of GMPLS based IP+Optical networks.

back to program ^

This presentation first describes three major potential issues in GMPLS based network operation from the service provider’s point of view.

• The GMPLS based network service provisioning requires highly flexible and granular policy, which may be difficult by using today’s average policy description technique for GMPLS. This is because GMPLS service users are sometimes large service providers who have diverse preferences and sensitivities in their service level experiences, which may be decided on negotiation basis.
• The GMPLS based protection and restoration will be missing a way to effectively cope with large and unexpected incidences. It may not be easy to “find correlated and remotely provisioned LSPs” such as of the same user, and to “identify a plan to restore the LSPs in a way which is the most efficient” to each user and incidence, because of the protection and restoration plan’s difficulty to predictably set and to determine in a unified way on the incidence.
• The expected GMPLS based new services (e.g., Layer 1 VPN), will not be easily managed with only today’s GMPLS technology. Many new services and most of traditional service orders are in nature reservation based Semi Permanent Connections. When accepting requests, the optimal route should be calculated based on the future network , demands, and other reservations. In addition, today’s GMPLS based technology misses capability to treat scheduling (reservation) concept and/or support from reservation managing entity (ies).

The presentation then proposes a framework of a GMPLS based network operation. It utilizes decomposed Path Computation Element (PCE) functionality as key mediating components for the network operation.

• The service instances (LSPs) are provided by the trigger of operators when they accept the simulated route assignment based on network status and requests. Fundamental functions are similar to PCE, including topology/resource update, LSP management, and path calculation.
• In order for the selected groups of service instances to be restored in a unified way for large scale failures (e.g., at the same time and in the same manner), the automated restoring actions of network nodes are interrupted and helped by operators through the centralized managing instance which can use the same PCE functions.
• New reservation based service instances will be handled in the centralized managing instance which has access to database on future reservation and network planning (e.g., network construction and expansion). The accepted reservation must be assigned a route which can be calculated by the combination of the PCE functions with expansion in time/schedule conditions and the planned future network topology and resources information. This also enables to assign a route for backup path at the service provisioning time as well as at the failure recovery time (protection and restoration).

The presentation concludes with analysis of today’s GMPLS and PCE standard techniques in order to implement the proposed PCE based GMPLS network operation framework.

back to program ^

The Interoperability working group (IWG) of Kei-han-na info-communication open laboratory studied GMPLS E-NNI interoperability under multi carrier environment. We developed a prototype node of E-NNI for establishing interoperability between carriers. In this paper, our prototype node of E-NNI is extended with some functions and is able to treats L2SC with D-plane equipment.

back to program ^

This talk covers the R&D project on IP and Optical Networking in NTT Network Service Systems Laboratories.

The IP and Optical Networking research project is conducted to make backbone network architecture flexible and resilient [1]. As IP based technology becomes common in carrier's network, unexpected new services emerge (Gyao, You-tube, Joost, etc.). As ICT-based social life prevails, outage of the communication network has enormous impact on continuation of our social life. Therefore the communication network has to be flexible to unexpected change of traffic and resilient to unexpected network failure.

In the IP and Optical Networking where IP routers and Optical transport nodes are managed in an integrated manner, multi-layer network operational tasks are conducted using Generalized MPLS-based control plane with support of PCE (Path Computation Element). GMPLS-based control plane offloads various complicated tasks from the network management system (NMS) and improves interoperability between network node systems. PCE offloads complicated traffic engineering computation from network node systems and NMS. We can concentrate on the core tasks in designing the NMS. We will present the detailed analysis of those benefits.

In the IP and Optical Networking, traffic engineering is done in both IP/MPLS and Optical layers to improve flexibility and resiliency in the IP and Optical Networking. Virtual network topology is constructed using a set of optical label-switched paths. IP/MPLS traffic engineering is conducted over the virtual network topology. In conducting traffic engineering in both IP/MPLS and Optical layers, we integrate traffic engineering database so that we can perceive the routing topology and diversity structure across layers. We can therefore optimize resource utilization and improve survivability of the networks. We discuss the applicability of multi-layer traffic engineering against network operation tasks: network design, network monitor, optimization, and failure recovery.

Finally we review other traffic engineering techniques including CSPF, automatic bandwidth adjustment, and global optimization in IP/MPLS layer. After giving taxonomy in two aspects: (1) optimization (local and global), and (2) control epoch (static, quasi-dynamic, and dynamic), we examine scalability, efficiency, resiliency, flexibility, and operational overhead. We discuss the applicability of those techniques against roadmap of the network architecture evolution.

back to program ^

Current Radio Access Networks in GSM/UMTS/CDMA networks leverage ATM to aggregate and transport voice/data/video flows from base stations to the core network. This model scales well, due in part, to the multi-user (thousands of users) aggregation capabilities of the base stations and due to the fine grain nature of ATM’s QoS and Traffic Management capabilities.

The transition to packet networks and the emerging introduction of micro base stations (so-called femtocells) is about to challenge this equilibrium. Femtocells power these home mobile access networks driven by the need to reduce macro-cellular costs associated with backhaul (OPEX and CAPEX). While these home mobile networks attempt to decrease macro-cellular costs, they increase the number of base stations homed to a core network by a few orders of magnitude introducing large numbers of (hitherto unknown) micro flows. What is the best method of transporting these micro flows ? IP tunneling ? MPLS LSPs ? What about security as the home mobile network traverses the public Internet ? What are the requirements of the “Mobile Access Internet” in terms of delay/latency/jitter etc ? How do packet networks scale/change to meet these requirements?

The challenge of new packet based networks is to offer QoS/CoS in support of IP-based multimedia (voice, data, video) services while scaling to support the proliferation of micro-flows generated in support of these services. This presentation investigates the QoS mechanisms of packet networks used to guarantee performance, especially with regards to support of wireless clients (base stations, micro base stations and femtocells). It will examine security considerations in traditional IP networks as well as the availability of MPLS at the network edge and its impact on the RAN's ability to leverage MPLS transport services. The presentation will also address how the IP/MPLS network can scale to support 100K mini base stations found in highly distributed RANs and will cover key management issues including IP address and MPLS come-and-go label management.

back to program ^

Orange is currently migrating the legacy PDH/SDH-based infrastructure so far used for the 2G, 3G mobile backhauling to a truly IP/MPLS network by using Pseudowire. The presentation first will provide an overview of the Orange architecture selected and the rationales for using MS-PW and FEC129 for the PW setup. Then it will discuss some technical issues and scalability challenges: (1) optimized and consistent AII addressing plane configuration, AII auto-discovery methods to allow PE/S-PE to be informed of AII configured on end devices and to maintain dynamically PW routing tables based on AII information, (2) PW segment protection to face T-PE failure in case of dual-homed RNC to two different T-PE, (3) PW interface parameters update without service disruption, typically upgrading the number of timeslots or the number of ATM cells to be encapsulated in a TDM/ATM pseudowire … The paper will present the related use cases experienced by a SP like Orange and will give indications on some mechanisms that can be used so as to satisfy these requirements.

back to program ^

Backhaul requirements are evolving rapidly, requiring more bandwidth, higher reliability, support for advanced QoS, and lower cost. Drops in revenue per bit in service providers’ business plan and scalability limits of current TDM backhaul are challenges faced by wireless service providers. MPLS can provide an ultimate solution to improve overall service reliability, scalability, and QoS. In this presentation, we present the network architecture for backhaul network, and address the challenges and solutions in designing such a network. Traffic requirements for Sprint next generation WiMAX network, and backhaul and QoS deployment roadmap are provided.

back to program ^

As the demand for services over MPLS networks grows, network operators are increasingly bundling multiple links together to increase the capacity of the paths in the network. At the same time, various Equal Cost Multi-Path techniques are used at various protocol layers such as BGP, OSPF, ISIS, RSVP-TE and LDP to provide multiple paths. This leads to numerous combinations of load balanced paths at different layers (IP, MPLS, Ethernet). With services being offered for native IP, L2VPNs, VPLS, L3VPNs, IPv6 and others, a one size fits all approach for load balancing is limited. In this presentation, we will look at the reasons for load sharing, the various layers where load sharing is used, highlight the criteria used for load sharing at each layer and recommend the best schemes to be used for each traffic type to boost network capacity utilization. We will further look at some interesting techniques of speculation in MPLS packets and neutralizing polarization effects.

back to program ^

In this session, we will review existing capabilities to perform VPLS OAM functions and discuss further enhancements. We will look at how BFD, VCCV, LSP ping functions can be complemented with MAC level OAM functions. Ethernet OAM functions as defined by IEEE 802.1ag will also be reviewed. This will lead to comparing the IETF and IEEE OAM toolsets and discussing the differences and synergies between them.

back to program ^

Demand for Ethernet-based services is on the rise, and the key driving force behind this is continuous growth of data traffic in the metro/access networks. Many service providers and enterprise network operators are already moving towards the evaluation and deployment of next-generation Ethernet services, which promise not only a new source of revenue, but also improved efficiency in the delivery of data services.

The challenge is that Ethernet is traditionally 'best-effort'. Now, industry bodies and equipment manufacturers are making considerable efforts to ensure Ethernet services are 'carrier class' and can be transported over IP/MPLS core and aggregation networks. Key areas of concerns such as quality of service (QoS), resiliency and scalability required to deliver carrier-class services are being addressed. Carrier network devices are adopting evolving technologies such as Bidirectional Forwarding Detection (BFD), Link Aggregation Control Protocol (LACP), Ethernet Connectivity Fault Management (CFM), Multiple Spanning Tree Protocol (MSTP), Stacked VLANs and Virtual Private LAN Services (VPLS). Therefore, testing is an essential first step in ensuring carrier-class Ethernet services are delivered to subscribers.

This paper presents an overview of Ethernet services, infrastructure and associated testing, and an appreciation of the challenges involved in evaluating carrier-class services and infrastructure being developed and deployed.

back to program ^

Service providers are looking to scale their Carrier Ethernet service offerings to support thousands VPNs with multiple service levels over 10G cores. To meet these needs, network equipment manufacturers (NEMs) have unleashed the next wave of "purpose built" products with new protocol features. This is moving the industry closer to the goal of an end-to end carrier-grade Ethernet network, where service providers can deploy multiple services over a single Ethernet backbone.

But how close are we to actually meeting these goals? The conformance standards for basic services of MEF 9 are no longer sufficient. This requires leaping into the new world of MEF 14.

Ixia and Isocore will provide the road map to help service providers and NEMS navigate an MEF 14 world, where they will need to conduct performance tests (data plane forwarding and control-plane scale) to show scalability with QoS across multiple service levels. They will provide real-world examples as to how to develop a testing plan to ensure that products and networks meet the challenges of MEF 14 for services and the underlying MPLS infrastructure.

back to program ^

This talk covers several recent developments in the space of Operations, Maintenance and Administration (OAM), which involves interactions between devices for the purposes of fault management and performance monitoring, for Ethernet and similar capabilities in MPLS.

Specific focus will be on the development of IEEE 802.1ag and ITU-T
Y.1731 functionality based on requirements and expectations in Carrier Transport networks also leveraging upon the work on Ethernet Services by MEF. IETF has also recently defined the Layer 2 OAM framework which allows utilizing the MPLS LSP and PW OAM functionality and complementing it with functionality available across other technologies, especially when these services are being emulated across IP/MPLS networks.

The comparison allows to take a look at what have been the main areas where the OAM functionality has focused on thus far and the synergies across these functions is also explored providing the service provider and the user community end-to-end operational ease and OPEX savings.

back to program ^

The Generalized Multi-Protocol Label Switching (GMPLS) suite is expected to achieve network resiliency against not only failures but also unexpected traffic surges. Specifically, the extension of signaling protocols such as the Resource reSerVation Protocol for Traffic Engineering (RSVP-TE) to support Virtual Concatenation (VCAT) along with the Link Capacity Adjustment Scheme (LCAS) [1] is now gaining much attention from service providers and can provide sophisticated bandwidth control mechanisms for Layer 1 networks to adapt packet traffic flow within Time Division Multiplexing based Label Switched Paths (TDM-LSPs).

First, this presentation clarifies the services that utilize the bandwidth-on-demand (BoD) capability of the GMPLS based VCAT/LCAS control mechanisms. Although there are several discussions concerning the on-demand leased line services, the objective of this study includes the creation of a new reliability class of Ethernet transport services. Namely, the combinational usage of the Generic Framing Procedure (GFP) and GMPLS based VCAT/LCAS control mechanisms to allow service providers to control adaptively the bandwidth of backup LSPs without completely disrupting the connection of Ethernet Traffic flows according to network conditions. To achieve such a Bandwidth Variable Restoration (BVR) strategy, we fully utilize the concept of “member LSP sharing” described in the VCAT&LCAS draft [1] discussed in the IETF CCAMP-WG. Namely, the GMPLS suite dynamically forms VCAT groups (VCGs) to which shared member TDM-LSPs belong. Then, the GFP functionality of TDM-XCs accommodates client Ethernet LSPs to the assigned VCGs. Through the employment of the BVR strategy based on the GMPLS based VCAT/LCAS control mechanisms, the service providers achieve efficient network resource usage and appropriate priority control to allocate bandwidth to client Ethernet LSPs.

Then, the OA&M architecture for networks employing the concept of “member LSP sharing” is discussed from the view point of management scalability. Also, the extension of “Call” mechanisms to manage and control VCGs is addressed. Through the assessment of real operational scenarios and OA&M issues, the requirements are summarized for not only the protocol extensions but also the implementation issues of the GMPLS based VCAT/LCAS control functionality in the network elements.

In conclusion, we believe that the GMPLS based VCAT/LCAS control mechanisms provide a novel solution to enhance the flexibility, resiliency, and efficiency of network operations. The establishment of this technology will pave the way to deploy new services such as scheduled services and Layer 1 VPN services. Finally, we should note that this technology is applicable even to future ITU-T G.709 based OTN transport networks.

back to program ^