- Physical Network Topology and Availability
- Layer 2 Availability: Trunking —802.3ad—Link Aggregation
- Layer 2 Trunking Availability Strategies using SMLT and DMLT
- Layer 2 Availability: Spanning Tree Protocol
- Layer 3—VRRP Router Redundancy
- Layer 3—IPMP—Host Network Interface Redundancy
- Layer 3—Integrated VRRP and IPMP
- Layer 3—OSPF Network Redundancy— Rapid Convergence
- Layer 3—RIP Network Redundancy
- Conclusion
- About the Authors
Layer 3IPMPHost Network Interface Redundancy
The purpose of the server redundant network interface capability is to increase overall system availability. If one server NIC fails, the backup will take over within two seconds. This is IP Multipathing (IPMP) on the Solaris™ operating environment.
IPMP is a feature bundled with the Solaris operating environment that is crucial in creating highly available network designs. IPMP has a daemon that constantly pings the default router, which is intelligently pulled from the kernel routing tables. If that router is not reachable, another standby interface, in the same IPMP group, then assumes ownership of the floating IP address. The switch then does a re-ARP for the new MAC address and is able to contact the server again.
A typical highly available configuration includes a Sun server that has dual NIC cards, which increases the availability of these components by several orders of magnitude. For example, the GigabitEthernet card, part number 595-5414-01, by itself has an MTBF of 199156 hours, and assuming approximately 2 hours mean time to recovery (MTTR), has an availability of 0.999989958. With two cards, the MTBF becomes 9 9's at .9999999996 availability. This small incremental cost has a big impact on the overall availability computation.
FIGURE 13 shows the Sun server redundant NIC model using IPMP. The server has two NICs, ge0 and ge1, with a fixed IP address of a.b.c.d and e.f.g.h. The virtual IP address of w.x.y.z is the IP address of the service. Client requests use this IP address as the destination. This IP address floats between the two interfaces: ge0 or ge1. Only one interface can be associated with the virtual IP address at any one instant. If the ge0 interface owns the virtual IP address, then data traffic will follow the P1 path. If the ge0 interface fails, then the ge1 interface will takeover and associate the virtual IP address, and then data traffic will follow the P2 path. Failures can be detected within two seconds, depending on the configuration.
.FIGURE 13 High-Availability Network Interface Cards on Sun Servers