Improving Service Availability of Ip Networks PDF
Content
Improving Service Availability of IP Networks
Author: Suman Choudhary [email protected] Overview A typical network is composed of access, aggregation, IP Edge and backbone layers, with a mix of optical and IP infrastructure, interconnected in a hybrid way of p2p, ring, partial or full mesh topologies. Reliability mechanisms in optical transport are mainly associated with fiber topology, independent of service. However, devices at Ethernet/IP layer rely on a diversity of control plane protocols to implement services with widely varying characteristics. Thus, reliability mechanisms at Ethernet/IP layer are specific to a given service and can differ much between session based services, L2/L3 VPN, Multicast, etc Reliability for IP Devices at Aggregation and Backbone Devices at the aggregation and backbone layer mostly perform packet forwarding for unicast and multicast services based on pre-calculated or configured paths/tunnels. At these layers, the prerequisite for high network reliability is to ensure availability of redundant paths, faster fault detection and notification, and fast convergence. In many networks, the aggregation and backbone layer are composed of multiple hops. The greater the number of network hops, the greater the number of failure points and the reliability mechanisms are increasingly more complex. Flattening the network reduces number of network hops and failure points. In the IP network, the link layer of an Ethernet or SDH network can perceive a port fault within 10 ms through the failure of photoelectric signals. IGP protocols such as OSPF, ISIS, and BGP use hello packets to monitor and detect failures which could take several seconds. Technologies such as BFD (Bidirectional Forward Detection), Ethernet & MPLS OAM mechanisms support failure detection in the order of few milliseconds. BFD provides low-overhead detection of faults even on physical media that don't support failure detection of any kind, such as Ethernet, virtual circuits, tunnels and MPLS Label Switched Paths. BFD supports interactions with most control plane protocols in IP/MPLS including OSPF, IS-IS, BGP, RSVP, LDP, VRRP, PIM, pseudo-wire, and TE tunnels. In addition, BFD is flexible to be used under different IP networking technologies and topologies. Reliability at IP Edge Layer The devices in the IP edge layer mainly comprising of BRAS, PE and SR work in conjunction with AAA, DHCP servers and DNS, enabling voice, data, unicast, multicast and different flavors of Layer 2 & Layer 3 services. Service intelligence is highly concentrated in the IP Edge layer. Ensuring minimal device downtime and fast switchover to standby infrastructure is critical to ensure higher service availability at IP Edge layer. One approach is to segregate delivery of residential and enterprise services on different equipment. Further, segregated delivery of voice, video and data services for residential customers on separate devices can further enhance reliability at the cost of increasing operational complexity and cost. Different services can have different maintenance windows and operational issues in IP Edge equipment for new feature support, software upgrades, etc, yet will not impact working of other services. Any device failure localizes the impact to only one of the services. As operators emphasize more on CAPEX and OPEX reduction, device convergence at IP Edge layer is becoming a trend. In this case, a single carrier class device is used to deliver all services for residential and enterprise markets alike. With such deployments, devices in IP Edge layer can adopt hot standby or cold standby models to deal with link and hardware related failures. In the Hot standby model the user and session information is synchronized between the master and slave devices real-time. Should a link/node failure occur, failover mechanisms will detect failure and trigger fast switchover within few 100’s of milliseconds. The user sessio n is not interrupted. In cold standby architecture spare infrastructure is installed to handle the customer sessions and traffic. However, in this case, the user experiences interruption of services.
1
HUAWEI TECHNOLOGIES CO., LTD.
Certain services like CGN, DPI, in-built caching, etc implement features that have application layer awareness and maintain lots of state information, thus making it more challenging to implement redundancy solutions. In such scenarios, it is recommended to segment the IP edge architecture into two layers; devices implementing basic subscriber management for voice, video and data can be distributed across many PoPs; devices implementing CGN, DPI, IPSec, video caching, etc. can be deployed at fewer centralized PoPs with cold standby model. This approach ensures that value added services (VAS) do not adversely impact the availability of critical services. Certain services such as multicast, L3 and L2VPN involve service awareness in both aggregation and backbone networks. The reliability mechanisms for these services require special consideration. Summary At the aggregation and backbone layer, reducing number of network hops reduces the number of failure points and failure scenarios. Centralizing IP service intelligence whenever possible at IP Edge and distributed deployment of IP Edge improves the reliability of network and reducing the network down time. For services such as VPN and Multicast which involve service awareness across multiple network layers, reliability mechanisms must be considered separately.
2
HUAWEI TECHNOLOGIES CO., LTD.
Author: Suman Choudhary [email protected] Overview A typical network is composed of access, aggregation, IP Edge and backbone layers, with a mix of optical and IP infrastructure, interconnected in a hybrid way of p2p, ring, partial or full mesh topologies. Reliability mechanisms in optical transport are mainly associated with fiber topology, independent of service. However, devices at Ethernet/IP layer rely on a diversity of control plane protocols to implement services with widely varying characteristics. Thus, reliability mechanisms at Ethernet/IP layer are specific to a given service and can differ much between session based services, L2/L3 VPN, Multicast, etc Reliability for IP Devices at Aggregation and Backbone Devices at the aggregation and backbone layer mostly perform packet forwarding for unicast and multicast services based on pre-calculated or configured paths/tunnels. At these layers, the prerequisite for high network reliability is to ensure availability of redundant paths, faster fault detection and notification, and fast convergence. In many networks, the aggregation and backbone layer are composed of multiple hops. The greater the number of network hops, the greater the number of failure points and the reliability mechanisms are increasingly more complex. Flattening the network reduces number of network hops and failure points. In the IP network, the link layer of an Ethernet or SDH network can perceive a port fault within 10 ms through the failure of photoelectric signals. IGP protocols such as OSPF, ISIS, and BGP use hello packets to monitor and detect failures which could take several seconds. Technologies such as BFD (Bidirectional Forward Detection), Ethernet & MPLS OAM mechanisms support failure detection in the order of few milliseconds. BFD provides low-overhead detection of faults even on physical media that don't support failure detection of any kind, such as Ethernet, virtual circuits, tunnels and MPLS Label Switched Paths. BFD supports interactions with most control plane protocols in IP/MPLS including OSPF, IS-IS, BGP, RSVP, LDP, VRRP, PIM, pseudo-wire, and TE tunnels. In addition, BFD is flexible to be used under different IP networking technologies and topologies. Reliability at IP Edge Layer The devices in the IP edge layer mainly comprising of BRAS, PE and SR work in conjunction with AAA, DHCP servers and DNS, enabling voice, data, unicast, multicast and different flavors of Layer 2 & Layer 3 services. Service intelligence is highly concentrated in the IP Edge layer. Ensuring minimal device downtime and fast switchover to standby infrastructure is critical to ensure higher service availability at IP Edge layer. One approach is to segregate delivery of residential and enterprise services on different equipment. Further, segregated delivery of voice, video and data services for residential customers on separate devices can further enhance reliability at the cost of increasing operational complexity and cost. Different services can have different maintenance windows and operational issues in IP Edge equipment for new feature support, software upgrades, etc, yet will not impact working of other services. Any device failure localizes the impact to only one of the services. As operators emphasize more on CAPEX and OPEX reduction, device convergence at IP Edge layer is becoming a trend. In this case, a single carrier class device is used to deliver all services for residential and enterprise markets alike. With such deployments, devices in IP Edge layer can adopt hot standby or cold standby models to deal with link and hardware related failures. In the Hot standby model the user and session information is synchronized between the master and slave devices real-time. Should a link/node failure occur, failover mechanisms will detect failure and trigger fast switchover within few 100’s of milliseconds. The user sessio n is not interrupted. In cold standby architecture spare infrastructure is installed to handle the customer sessions and traffic. However, in this case, the user experiences interruption of services.
1
HUAWEI TECHNOLOGIES CO., LTD.
Certain services like CGN, DPI, in-built caching, etc implement features that have application layer awareness and maintain lots of state information, thus making it more challenging to implement redundancy solutions. In such scenarios, it is recommended to segment the IP edge architecture into two layers; devices implementing basic subscriber management for voice, video and data can be distributed across many PoPs; devices implementing CGN, DPI, IPSec, video caching, etc. can be deployed at fewer centralized PoPs with cold standby model. This approach ensures that value added services (VAS) do not adversely impact the availability of critical services. Certain services such as multicast, L3 and L2VPN involve service awareness in both aggregation and backbone networks. The reliability mechanisms for these services require special consideration. Summary At the aggregation and backbone layer, reducing number of network hops reduces the number of failure points and failure scenarios. Centralizing IP service intelligence whenever possible at IP Edge and distributed deployment of IP Edge improves the reliability of network and reducing the network down time. For services such as VPN and Multicast which involve service awareness across multiple network layers, reliability mechanisms must be considered separately.
2
HUAWEI TECHNOLOGIES CO., LTD.
