A Higher Level of Redundancy
The good news is that trunking is about to go standard. But as usually happens in the industry, select vendors are pushing ahead with proprietary features that extend the new IEEE 802.3ad standard.
First, a bit of history for those who have not followed this technology: Trunking is a technique that allows a networking device to bond together multiple physical links into a group that works like one logical link. In a LAN environment, trunking can be used on any type of network capable device -- a Layer 2 switch, a Layer 3 switch, or a computing device such as a server or desktop computer. The networking device is, of course, equipped with multiple physical links. These physical links are, in turn, considered to be part of a larger entity known as the trunk port. To the higher-level protocol functions, such as Spanning Tree or SNMP, the trunk port is viewed as one big logical link into the network.
The subtle fact behind the trunk port is that physical links are associated with a set of MAC-addressed ports -- in an Ethernet environment. When frames or packets are sent into the network, it first becomes necessary to pick one of the MAC-addressed ports, within the larger logical trunk port, over which to send the data. This selection is done by something called the MAC selection algorithm. The MAC selection algorithm's job, however, doesn't stop at this point. In addition, it plays another vital role -- the enforcer that makes sure the frames being transmitted into the network arrive in order at the destination network device. To do this, the algorithm creates a set of associated frames known as a session. The common characteristic of a session is that all frames in it have the same source and destination addresses. By controlling the session, the MAC selection algorithm makes sure all frames going between the source and a particular destination travel on the same physical link within the trunk group, thereby arriving at the destination in correct order.
As mentioned, all of this complexity is covered by the IEEE standard 802.3ad. To get the standard completed as quickly as possible, however, the working group restricted the scope of its efforts to point-to-point links only. This means that this sophisticated trunking capability -- now called link aggregation -- is standards compliant only if all physical links on one device are connected to the same networking device on the other side of the point-to-point arrangement. The standard does not allow some of the physical links emanating from one device to be connected to more than one networking device on the other side. In other words, physical links within the trunk group cannot be spread across more than one switch or router.
The standard's restriction on one-to-many connections affects the issue of reliability and redundancy. The 802.3ad standard does an excellent job of ensuring that packets will be rerouted to the available link should a link failure occur between point-to-point device. It does not, however, cover cases when the networking device fails. When this occurs, all physical links between the devices go down and all transmission capability is lost.
One-to-many connections solves this problem, albeit in a non-standard manner. Since the physical links are connected to different networking devices, a device failure affects only those links connected to it. Transmission is still possible over the physical links within the trunk group that are connected to different networking devices.
The rules, therefore, are simple: 802.3ad link aggregation provides standards-compliant, link-level redundancy; and proprietary extensions to the standard deliver both link- and device-level redundancy. Even as suppliers are building products that offer the 802.3ad standard, some are also extending the standard to offer the next higher order of redundancy and availability. --Sam Alunni is vice president of networking at Sterling Research (Sterling, Mass.). Contact him at firstname.lastname@example.org.