11-29-2010 08:16 AM
I've had a couple questions about how VCS Technology support for automatic inter-switch link (ISL) formation and automatic Trunk formation differ from LAG and Etherchannl and how the VCS Logical Chassis is different than switch stacking. I think the short answer is the degree of simplicity VCS provides compared to these techniques. Let me explain a bit more.
LACP, as defined in 802.1ad, offers a means by which multiple ports on different switches can form a single logical link between the switches. This is helpful when you need to bypass spanning tree to increase bandwidth as now two (or more) physical links can be used between two switches. Spanning tree is aware of the logical link aggregation group (LAG), not the underlying physical links.
In an Ethernet fabric using VCS technology, multiple ports on two VDX switches can be connected together. Each connection automatically comes up as an inter-switch link with NO configuration requirement for the ports on either switch. It's entirely automatic. LACP requires configuration settings on each port that will participate in a LAG. Hence, VCS ISL formation between two VDX switches is built-in and automatic by default.
Stacking solutions have been around for some time. Typically, special ports are required to form the stack; a single topology, a loop, is required; a single master switch is elected for the stack; and there are limitations to scale-up due to the limitations on the number of special stacking ports available. It's fairly easy to create an initial stack to meet a specific over-subscription ratio within the stack, but often very hard (if not impossible) to maintain that ratio as more switches are added. In some cases, the minimum over-subscription ratio is too large to meet the demands of east-west traffic flows often found in virtual server configurations.
In a VCS Ethernet fabric, no special ports are needed to form an ISL connection. If two ports are connected between two VDX switches, an ISL automatically forms. Further, if more than one ISL connections is added between two switches a trunk automatically forms (no configuration required on any of the ports involved) provided the ports on each switch are within the same "port group". In the current generation of VDX switches, a port group has a maximum of 8 ports, so a single trunk can flow up to 80 Gbps. A VCS Trunk is similar to an Etherchannel, but is self-forming without special configuration settings.
The VCS trunk provides very high utilization rates across all the links in the trunk as the frame forwarding across a trunk is done in hardware, frame by frame. This avoids problems in Etherchannel caused by hashing frame headers for load balancing that can create hot spots. The impact of that is more tinkering required by the network administrator, as discussed in this blog post by Ethan Banks, "The Scaling Limitations of Etherchannel -Or- Why 1+1 Does Not Equal 2"). Now multiple trunks can be configured between VDX switches if need be. This provides dead simple configuration, adjustable over subscription ratios all the way down to 1:1 if so desired, built in link resiliency within the trunk and support for any topology you care to use, not just a loop.
Another point to make is when two VDX switches are connected with an ISL, they form a VCS Ethernet fabric with Distributed Intelligence. That means they share a common distributed control plane. With stacking, there is often only a single control plane created in the master switch, but this is not the case with the VCS technology.
Finally, with VCS technology, all switches in an Ethernet fabric are managed as if they reside in a single logical chassis. This is very similar to the ability of stacking to support management of all switches in the stack as a single management entity. But, in stacking, it is typical to have to log into the master switch to configure the stack. With VCS technology, you can log into any switch in the Ethernet fabric for management as there is no master switch.
I hope this helps clarify the degree of innovation available in VCS technology. On the surface it may sound like "what has come before", but when you dig into the details, I think you will see there are some valuable differences.