Summary

To reach a high level of availability, you have to focus on these areas: network device reliability, device redundancy, link redundancy, fast convergence, a correct network design, and a well-documented network. There are two general types of redundancy: component and chassis. Component redundancy provides protection against a component failing inside a chassis. Chassis redundancy protects against failure of a device.

In hardware redundancy, if you're going to have dual SEs, they must be in slots 1 and 2. The secondary SE is in a standby state, with the exception of its uplink interfaces, which can be used. RPR provides hardware redundancy for the MSFC and PFC cards. The primary SE uses its own MSFC card, whereas the secondary is in a standby state. When a switchover occurs, it can take from 2 4 minutes because the MSFC card must be initialized. A switchover occurs if the MSFC and PFC fail on the primary, there is a clock synchronization problem, or a manual switchover is initiated. RPR+ has the secondary MSFC fully operational, which means the switchover takes only between 30 60 seconds. To configure RPR+, use the redundancy and mode rpr-plus commands.

Proxy ARP allows for a basic level of redundancy for default gateways if a device ARPs for a destination that is not on the same segment, a Cisco router can respond back with its own personal MAC address. Proxy ARP has problems with dealing of failures of the default gateway it is typically used to dynamically discover it, though. IRDP uses ICMP to dynamically discover default gateways. RPs announce themselves and then end stations use them. The problem with IRDP is that when an active RP fails, it can take up to 30 minutes for the end station to start using another RP.

HSRP is a Cisco-proprietary protocol that provides default gateway redundancy and is invisible to the end stations in the VLAN. A single virtual IP and MAC address is used per group. An active RP, elected by the RP with the highest priority (or IP address, if a tie occurs), forwards traffic. A standby RP monitors the active RP. There are six stages an HSRP might go through: initial, learning, listening, speaking, standby, and active. An RP goes into a speaking state when an election occurs, or if it is the active or standby RP. The active RP can tell the rest of the RPs about the virtual addresses. To enable HSRP, use the standby ip command on an RP's interface. HSRP supports both preemption and interface tracking.

SRM provides Layer 3 redundancy between dual MSFC cards. One is the designated RP and the other is nondesignated. The nondesignated RP is operational, but all of its interfaces are disabled. SRM's advantages include using only one set of IP addresses, fewer RP peers, and configuration needs to be done only on the designated RP.

VRRP is an open standard for default gateway redundancy and works on Ethernet, VLANs, and MPLS VPN media types. VRRP has a master and backup RPs. Either a virtual IP address or a real IP address (of the master) is used.

GLBP is an enhanced version of HSRP. It allows for up to four RPs to forward traffic from the group. RPs are grouped together and each group is assigned one or more virtual addresses. The AVG is responsible for address management, whereas the AVFs forward traffic the AVG can also be an AVF. GLBP supports three types of load balancing: round-robin (default), weighted, and host-dependent.

SLB allows an IOS RP to load-balance traffic across a group of real servers. An end station sends traffic to a virtual IP address and the RP forwards this traffic to one real server in the SLB group. In directed mode, SLB performs NAT on the end-station's packet; in dispatched mode, SLB puts a real server's MAC address in the packet, but leaves the virtual IP address as is (it is assumed that the real server understands about the virtual address).