Bringing redundancy to Layer 3 switching
 
|
|||
 | |||
|
Resiliency and redundancy are critical qualities in today's LANs. From the resiliency of load-balancing server network interface cards to the redundant fabrics and power supplies of backbone switches, LAN technologies continue to make high-availability networks more commonplace.
Now, thanks to the IETF's Virtual Router Redundancy Protocol (VRRP), the same qualities are coming to Layer 3 switches and routers that serve as statically configured default gateways.
Determining the default gateway is the first step in routing data from one network domain to another. Some end systems use dynamic route discovery protocols, such as Open Shortest Path First (OSPF)or Routing Information Protocol (RIP), to determine their correct default gateways. OSPF and RIP are able to circumvent any failed outages to come up with the optimum gateway.
But many organizations prefer to use statically configured default gateways to reduce processing overhead in the end system. The risk of this process is that a router serving as a default gateway becomes a single point of failure, and failure can spell catastrophe for end systems that rely on the default gateway for WAN connectivity or access to other LAN domains.
VRRP is designed to guard against such failures, and to do so with a relatively simple, efficient architecture. And although dynamic route discovery protocols are able to locate a new default gateway, VRRP provides for faster and more efficient failover in the event of an outage.
Furthermore, the protocol functions in load-sharing applications, as well. For example, VRRP lets a router act as a master for one IP subnet and a backup for another that has its own master. Two routers configured this way can load-share, with each router acting as a redundant backup for the other.
The master
When VRRP is turned on, one of its first duties is to provide the master router with a virtual media access control (MAC) address, one that can be taken over by the next-in-line backup. (VRRP allows for more than two routers, but two is the most likely scenario.)
MAC resides at Open Systems Interconnection Layer 2 and represents a device's physical identification; switches know where to forward IP packets by relating the IP address to the MAC address of the next-hop switch or router. By employing a virtual MAC address, VRRP ensures that end systems won't need to change their default-gateway address in the event of failover. In fact, end systems won't notice that a backup has taken over for the master.
Key to VRRP's operation is a process in which the master router sends a special VRRP packet, known as an advertisement, to the backups. Advertisement intervals can be set by the user; the VRRP default is 1 second.
If the advertisements suddenly stop, the backups set interval timers, typically for three times the advertisement frequency.
If no further advertisements appear, the backups assume the master is down and the failover routine is activated. From that point, the election of the next-in-line master typically takes less than a second.
Meanwhile, each router must report its status regularly to the network management console via SNMP. Network management may also be alerted if the VRRP packets show anomalies in any other areas - for instance, if a normal handshaking error occurs or authentication fails.
Providing protection
Although it is a relatively simple protocol, VRRP is effective protection for networks with statically configured gateways. And while some end systems employ discovery tools for locating new gateways in case of failure, VRRP represents a far more efficient solution to the problem - VRRP applies intelligence to the network rather than to the end system. Compared with the alternative - setting and managing individual discovery protocols in each of the network's end systems - VRRP is an efficient, simple technology.
On a macro level, VRRP represents another step organizations can take toward high-availability computing.
Years ago, organizations concentrated their resiliency/ redundancy efforts largely on their data center systems and servers.
Now organizations are realizing that high-availability computing requires high-availability network.
Related Links
Kraus is product marketing manager for Layer 3 switching at 3Com. He can be reached at Charlie_Kraus@ 3Com.com.
