Hundreds of Columns Sorted by Topic
Optimizing Content Delivery in the Core with P2MP LSPs
by Muralidhar Devarasetty and Mazen Khaddam,
2/17/2009
Optimized content delivery has become a critical requirement due to the increased level of media-rich traffic on networks. These bandwidth intensive and quality-sensitive offerings raise scaling and operational challenges across multiple dimensions. They also require a mechanism complementary to VPNs to deliver content securely, with resiliency and with maximum cost-efficiency. Point-to-multipoint (P2MP) label switched paths (LSPs) are now being deployed to resolve these issues and meet the needs of these emerging applications. In the last 9 to 12 months, the cost-effectiveness and numerous benefits of P2MP LSPs have been driving many major service providers, MSOs and content delivery networks to upgrade their networks to use P2MP LSPs, which is now an open-industry standard.
The growing number and type of media-rich services are posing new technological challenges in terms of optimized content insertion, transformation and distribution. Many providers are expanding their HD offerings to thousands of channels, while offering increasingly sophisticated collaboration tools, such as video conferencing and enterprise triple-play bundles. These offerings combine advanced managed enterprise services, including RFC 2547bis--based VPNs, Layer 2 VPNs, pseudowires, and virtual private LAN service (VPLS) for diverse geographic regions: metro, national and worldwide. As such, uncompromised performance, optimized cost structures, carrier-class reliability with low latency, and superior QoS are becoming critical infrastructure requirements.
Converge! One Minute Videos
Requirements for Optimized Content Delivery: Resiliency, Performance and Operational Simplicity
Superior convergence with protection against failures is essential for mission-critical applications. As well, a high level of resiliency is needed to support real-time applications that typically require convergence times less than 50 milliseconds. For example, loss of an I-frame in an MPEG stream can cause visible degradation of video quality. Loss of a frame on a SIP stream can also adversely affect voice quality. Traditionally, broadcast networks met the high-availability need with dual diverse connections, but at the cost of doubling the resources. MPLS, on the other hand, does not share common network paths or resources, thereby providing a zero-loss environment.
With the recent shift to HD video and increase in HD broadcast channels, demand is growing for high throughput in the infrastructure. Video delivery over packet networks mandates that underlying network elements support high performance for packet replication and provide flawless system characteristics for delay, jitter and packet loss.
The zero-loss requirement for video broadcast demands superior control plane and forwarding performance. Typical IGPs and multicast protocols take tens of seconds for network convergence in the event of a link or node failure. Alternatively, MPLS offers inherent protection against failure with its Fast Reroute capabilities. MPLS resource reservation and traffic-engineered paths have tremendous advantages for over-the-top, on-net and other premium services.
Service delivery requirements vary considerably and force service providers to have multiple networks, thus providing the potential to appreciably increase delivery costs. Significant savings can be achieved by leveraging common MPLS and P2MP LSP mechanisms. Multicast-based video and content distribution leverage native multicast forwarding using a PIM-like control plane. Point-to-point Layer 2 services require an MPLS and LDP infrastructure for control plane and data plane forwarding. VPLS E-LAN services leverage several point-to-point Layer 2 circuits to emulate a broadcast network. Layer 2 and Layer 3 VPNs for managed communication services require BGP, MPLS and LDP technologies. Conversely, MVPNs require MPLS, GRE and PIM-based technologies for service delivery. VoD service requires only a call-admission mechanism for guaranteed resources for the complete network path and video source.
Challenges with Content Delivery
Traditionally, multicast technology options for content delivery were limited, and implementations were based on PIM protocols in the infrastructure core. PIM-based multicast deployments pose a series of challenges:
- IP multicast does not allow end-to-end traffic engineering (TE) paths and guaranteed QoS. It is difficult to determine the actual path IP multicast traffic takes and its affects on link and node failures. In contrast, a multicast VPN (MVPN) with P2MP LSP technology enables service providers to set up statically and dynamically routed redundant paths to multiple branch LSPs, thus making it easier to troubleshoot failures. P2MP LSPs also simplify operations by removing the need for PIM in the core.
- Most current networks use only a small percent of the available bandwidth and are allowed a maximum of 50 percent of the available utilization. For broadcast services, this low link utilization is not cost effective. A real class of service with the capability to control network traffic by service and by customer becomes a critical need when broadcast, multicast and unicast services are offered on a convergent network.
- Financial services, news services and stock exchanges have high-touch content that requires low latency, nonstop operation and high security. Although IP multicast can achieve the broadcast requirement, it lacks the security offered by combining MVPN services and a P2MP LSP delivery mechanism.
Advantages of P2MP LSPs
P2MP has many advantages over traditional PIM infrastructures, which falls short in meeting the requirements of today's broadcast and multicast services. PIM networks cannot promise subsecond convergence times due to their dependence on integrated routing protocols. As well, PIM provides limited control over the traffic path, as many services depending on multicast routing tend to follow a similar path. This approach often results in overutilization of network segments, suboptimal load distribution, underutilization of links and service degradation.
Lack of Control over Multicast and Unicast Traffic
Many providers still depend on point-to-point transport technologies to offer broadcast and multicast services. For example, emerging metro Ethernet services based on VPLS, being Layer 2 in nature, cannot use the PIM core, thus network elements replicate broadcast and multicast packets to every egress PE leveraging point-to-point MPLS LSPs. Such services can easily create a choke point on the ingress PE (Figure 2). All multicast traffic in the network between the same pair of nodes will follow the same path and degrade network resource utilization. This approach also creates challenges for bandwidth and high-availability modeling to protect against failures.
Optimizing Multicast Rollouts with High-Performance P2MP LSPs
A P2MP LSP is an LSP with multiple destinations and it has the ability to replicate packets at the MPLS layer. This approach optimizes bandwidth by eliminating multiple unicast streams via replicating MPLS packets at every relevant or designated branching router and MPLS switch.
P2MP LSPs leverage all of the rich and proven MPLS capabilities, such as TE, (superior) control over service traffic (unicast and multicast), enhanced security, faster convergence and simpler operation. For example, proven Fast Reroute mechanisms with convergence times of less than 50 milliseconds are leveraged by P2MP LSPs and offer outstanding failure recovery times for video and other real-time applications (Figure 3). P2MP LSPs also reduce complexity by eliminating the need for less stable multicast routing protocols. Moreover, service providers can combine MVPN directly with P2MP LSPs, thereby avoiding the compounded complications of having to concurrently manage MPLS, IGP, VPN IGP, PIM and, possibly, GRE. In the case of VPLS, P2MP LSP eliminates ingress replication choke points by transporting traffic over P2MP LSPs.
Conclusion
P2MP LSPs offer several advantages over PIM-based network rollouts. A P2MP LSPs-based solution also simplifies overall complexity of the solution for services including Multicast services, IP VPN, VPLS, or E-LAN services. In conclusion, P2MP LSPs offer the following advantages:
- Traffic engineering and path control
- Resiliency and high availability with Fast Reroute convergence
- Streamlined, integrated operations with a consistent MPLS model for delivery of point-to-point and multipoint services
About the Authors
Muralidhar Devarasetty is a Sr. Technical Marketing Engineer at Juniper Networks, Inc. Muralidhar holds a BS degree in electronics engineering, with several networking certifications, including JNCIE and CCIE. He has more than 13 years experience in networking and has worked in various capacities, including Sr. Systems Engineer, Principle Network Consultant and Networking Engineer.
Mazen Khaddam is a Principal Networking Architect at Cox Communications, Inc. Mazen has had the foresight to design highly scalable and resilient networks for the rapidly growing Cox IP network infrastructure. With over 15 years of experience in the communication fields, Mazen has a bachelor and a masters degree in electrical engineering. Mazen also has numerous certifications from Juniper Networks, Cisco Systems, Sun Microsystems, and HP. About Juniper Networks
About Cox Communications
Send us your response to this article.
Learn How to Get Your Column Published on this Site
