Trunking Goes Standard
Trunking is a technique that allows you to take several ports on a networking device and group them together to work as one big link. Four fast Ethernet ports, for example, can be trunked to work like one big 400 Mbps pipe. For those cramped by the speed limitations of their current networking technology, trunking provides a cost-effective way to pour bandwidth into critical network hotspots.
There is nothing new in this concept. Over the years trunking has appeared in many incarnations in many different market segments. For example, I remember reading in the mid-1980s about the coming of the new 3745s, which were going to stripe SNA data over multiple T1 links -- a clear example of trunking -- in the wide-area networking space. More recently, trunking has quietly penetrated some LAN environments where brave networking designers were willing to risk using a proprietary technology in exchange for low-cost bandwidth expansion. Overall, however, trunking has remained out of the mainstream and out of the consciousness of most network planners.
But trunking's low profile existence is about to come to an end. With the 802.3ad specification now entering the final phase of its ratification process, trunking is about to become an IEEE standard, shedding its dreaded proprietary label. If anything, trunking now stands a good chance of becoming a stock player in many LAN environments.
So what are some of the virtues of the 802.3ad trunking specification? Scalability comes first. The specification, now called link aggregation, covers all three Ethernet speeds -- 10, 100, and 1,000 Mbps -- giving link aggregation maximum market appeal. Regardless of the size and speed of their enterprise network, planners can expect to find link aggregation shipped as a standard feature on networking equipment from their preferred supplier. For those with larger enterprise networks, scalability translates into a breadth of solutions for different parts of the network: link-aggregated backbones running at multiples of 1,000 Mbps, all the way down to link-aggregated wiring closets running at speed multiples of 10 or 100 Mbps.
Luckily, the link aggregation standard did not have to sacrifice link redundancy. This extremely desirable feature remains part of the new standard. For those not familiar with this concept, link redundancy amounts to an automatic failover capability where the physical links act as redundant backups for each other. Should a link fail, packet flows are redirected to another physical link within the logical link aggregation group, and most importantly, this redirection is done with a subsecond response time. Obviously, such redundancy will be highly prized when used for switch-to-server connections -- the second major application area in which link aggregation is expected to be important.
Although the benefit of standards has been proven, vendors are not standing still at the soon-to-be-approved IEEE revision. Some are already pushing into new proprietary territory. Multipoint link aggregation is one area in which promising technical extensions are being made. Link aggregation, as it stands now, only allows for point-to-point connections between two connected switches or a switch and a server. The multipoint extension will allow multiple switches -- not just their physical links -- to be combined into one, large logical connection. In such a multipoint link aggregated environment, data will flow through the network across different physical links connected to different switches. Ultimately, however, all of the logically associated data will be reassembled into its correct order by the multipoint link aggregation service at the destination point.
Most of the technical work on the link aggregation specification will be completed this year. Formal ratification is expected by 2000. --Sam Alunni is vice president of networking at Sterling Research (Sterling, Mass.). Contact him at alunni@sterlingresearch.com.