A New Time Critical Architecture for MPLS Provisioning

Susan Hares
NextHop Technologies

In 2004, Carriers are deploying VPNs with hundreds of VPNs per multi-service router/switch. Each of these MPLS VPNs may have 2000 - 10,000 routes per VPN, and hundreds of lines of configurations. A critical component in maintaining 5 9's of uptime for each VPN is the ability to perform time critical changes to each multi-service device.

A multi-service IP/MPLS router/switch supporting 500 VPNs may have 100 to 2,000 lines of configurations per VPN. The higher numbers for configuration lines are used for networks that do extensive prefix filter for routes or security filters. Often 5% to 10% of these routes may require configuration changes based on some network event. This results in a potential of 10,000 to 20,000 configuration changes to the router --- perhaps having to be done per second -- in order to handle a set of VPN changes.

A full download of an initial or rebooted configuration for a box with 500 VPNs may require 100,000 lines of configuration to be re-scanned by the router. This load on a rebooted multi-service router/switch may cause delay in rebooting a switch/router upon hardware failure.

Within a few years, the multi-service IP/MPLS routers may need to do thousands of configuration transactions per second. Carriers will not accept multi-services IP/MPLS routers with configuration rates that do not support the customer's need for the current level of response for time critical events.

This talk describes a new time critical remote management architecture for the interaction of OSS devices to configuration process on multi-service IP/MPLS devices. This architecture has been deployed in the Advanced Management Interface (AMI) is available as part of the NextHop GateD NGC portable IP/MPLS stacks. The architecture utilizes hierarchical configuration structures and transaction oriented processing to provide time critical configuration and network management. Time critical network management is defined as queries or events that have critical impact on operational events.

Current SNMP and XML queries do not always differentiate between time critical queries and large data gather. Critical SNMP data queries may be queued behind a large bulk transfer of a routing table. Most logging information is not time-critical and may also impact a systems ability to respond to time critical events.

This new Time Critical Network Management architecture allows:

  • Support for thousands of virtual routers within a single box running IP/MPLS switches,
  • Support for al IP/MPLS signaling and forwarding protocols,
  • Ability for multiple management process or multiple CLI to interact with the routing processes support the IP/MPLS switches,
  • Locking of a portion of the configuration hierarchy to a particular management process for a particular Time critical function (configuration, events, queries),
  • Synchronization of configuration or Time critical functions on flexible configuration boundaries (features within a protocol, protocol, groups of protocols or full node configuration), and
  • Roll-back of configurations based on flexible configuration boundaries,
  • Ability to hot-swap the configuration between failure, and
  • Support for "journaling" of the configuration

    This Time-Critical architecture has been deployed in products for enterprise and carrier class IP/MPLS devices in a variety of MPLS services in a variety of management systems. This talk will conclude with lesson about scaling of the Time-Critical Management architecture for large numbers of routing process, routes, and MPL Services.

    Bio:
    Sue Hares is recognized as one of the world's foremost experts in routing technology. She leads the technology qualification, development, and strategic planning functions at NextHop. Prior to launching NextHop Technologies, Ms. Hares spent 13 years at Merit Network Inc. where she most recently directed the Merit GateD Consortium. Ms. Hares was also a senior engineer at both Allen-Bradley Corp. and ADP Inc.

    An active participant in the design, specification and implementation of routing protocols, Ms. Hares co-chairs the Internet Engineering Task Force’s (IETF's) inter-domain routing group that is standardizing border gateway protocol. She is also a member of the NANOG (North American Network Operators' Group) steering committee.

    Ms. Hares earned a bachelor of science degree in computer engineering from the University of Michigan.