- Physical Network Topology and Availability
- Layer 2 Availability: Trunking —802.3ad—Link Aggregation
- 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
Layer 3Integrated VRRP and IPMP
By combining the availability technologies of routers and server NICs, we can create a reusable cell that can be reused in any deployment where servers are connected to routers. This reusable cell is highly available and scalable. FIGURE 7 shows how this is implemented. Lines 1 and 2 show the VRRP protocol used by the routers to monitor each other. If one router detects that the other has failed, the surviving router assumes the role of master and inherits the IP address and MAC address of the master.
Lines 3 and 5 in FIGURE 7 show how a switch can verify that a particular connection is up and running, which can be por-based, link-based, or Layers 3-, 4-, and 7-based. The router can make synthetic requests to the server and verify that a particular service is up and running. If it detects that the service has failed, then the VRRP can be configured, on some switches, to take this into consideration to impact the election algorithm and tie this failure to the priority of the VRRP router. Simultaneously, the server is also monitoring links. Currently, IPMP consists of a daemon, in.mpathd, that constantly pings the default router. As long as the default router can be pinged the master interface (ge0) assumes ownership of the IP address. If the in.mpathd daemon detects that the default router is not reachable, automatic failover will occur, which brings down the link and floats over the IP address of the server to the surviving interface (ge1).
In the lab, we can tune IPMP and Extreme Standby Routing Protocol (ESRP) to achieve failure detection and recovery within one second. The trade-off, is that because the control packets are on the same network as the production network, and because ESRP is a CPU intensive task, if the switches, networks, or servers become overloaded, false failures are possible because the device can take longer than the strict timeout to respond to the peer's heartbeat.
FIGURE 7 Design PatternIPMP and VRRP Integrated Availability Solution