HP Integrity rx2800 HP Integrity Network Adapter Teaming Whitepaper - Page 48

• Link loss detection - "Do I have physical link?" • Transmit path validation - " Can I successfully transmit a frame onto the network?" • Receive path validation - "Can I successfully receive any kind of frame from the network?" These basic redundancy mechanisms do not provide for redundancy beyond the network adapter port in the team. In other words, these basic redundancy mechanisms have no intelligence to detect network conditions that would prevent a particular team member port from being used by the team. If one teamed port is isolated from the rest of the team on another network, the isolated teamed port would most likely pass all three of the basic redundancy mechanisms tests: a) it would have link, b) it could successfully transmit heartbeat frame (it doesn't care if another team member receives it or not), and c) it could receive any kind of broadcast frame that may be transmitted on the isolated segment (from another network device). As a result, the basic redundancy mechanisms don't provide for intelligently choosing which teamed ports to use or not to use. In addition to the basic redundancy mechanisms not being intelligent enough, use of Active Path alone doesn't always allow for optimal use of teamed resources. Active Path lacks the granularity to choose the BEST teamed port. Active Path determines if teamed ports are good or bad (in other words, binary result). Active Path can't determine good versus best. Fast Path, on the other hand, provides the granularity that Active Path lacks. Fast Path can make the same decisions that Active Path can make (in other words, good versus bad), however, Fast Path goes one step further and determines good versus best. For example, refer to the diagram of an NFT team in Figure 4-7. The HP Integrity server's Primary team member port is connected to Switch A. Switch A's uplink to the core switch has been unplugged/disconnected. This results in the HP Integrity server being effectively disconnected from the entire network except for Switch A. Only the 50 users on Switch A can access the server, whereas the 50 users on Switch B, the 250 users connected via the Core Switch, and the router connection to any external network can no longer communicate with the server. Even though a non-Primary port is connected to this isolated segment, non-Primary ports in an NFT team are not used for any transmit or receive communication. If this was a TLB team instead of an NFT team, the server would still only be able to communicate with the network represented by Switch A since only Primary ports can receive traffic from the network. When this type of situation occurs on the network, it would be better for the server to proactively failover and choose a new primary from one of the team ports that still has connectivity with the core segment. In addition, the team should discontinue the use of any teamed ports that don't have connectivity with the core segment to prevent transmitting on a teamed port that can't reach the intended target (for example, client, router, etc.). The use of heartbeat frames between teamed ports will not provide enough information for the teaming driver to make the best failover decision. In other words, if the Primary port can't receive heartbeat frames from the non-Primary and vice versa, the only thing the teaming driver knows is that the teamed ports are no longer attached to the same network segment. The teaming driver still doesn't know which network segment is the best segment to choose. 48 The Mechanics of Teaming for the Advanced User