Now that virtual technology has permeated the server infrastructure and is starting to branch into other areas of the data center, many enterprises are learning a hard lesson: virtualization giveth and virtualization taketh away.
Nowhere is this more obvious than in the network. It seems that all the money being saved in hardware consolidation will have to go toward upgrading network architectures to accommodate the much more fluid data environments that virtualization fosters.
A key consideration is what to do about old-style STP (Spanning Tree Protocol) architectures. As Storage Switzerland's George Crump describes it, STP's fatal flaw is its habit of finding the most optimal data path between two points and then shutting down alternate routes. This was fine in earlier times when the goal was to avoid complex data loops, but in the free-wheeling world of virtualization, constant re-working of network paths not only overloads network controllers and traffic managers, it leaves a lot of network capacity sitting idle.
Enter TRILL, the Transparent Interconnection of Lots of Links, the latest creation of the IETF (Internet Engineering Task Force). It's a more intelligent protocol for Layer 2 networks that not only accommodates multi-path networking, but incorporates a high degree of intelligence and monitoring to constantly shift and re-shift traffic flow to make the most efficient use of available resources. Best of all, it stands a good chance of combining Layer 2 edge and Layer 3 aggregation infrastructure with core routing functions to finally do away with the complex multi-tier networking designs that STP has forced on the industry over the past several decades.
Vendors, at least, are already showing enthusiasm for TRILL. Cisco has incorporated it into the new FabricPath platform that is being incorporated into the NX-OS operating system. The company sees an initial market in clustered application environments in HPC settings and has already lined up key clients like NASA's Ames Research Center and the Lawrence Livermore National Laboratory.
Meanwhile, Brocade is matching Cisco step-for-step with its new Virtual Cluster Switching portfolio, which combines TRILL with Data Center Bridging (DCB) technology to flatten out Layer 2 networks and unite Ethernet and storage networking under a single fabric. The company is eyeing a seamless transition to converged fabrics, with one of the primary goals being the elimination of STP architectures.
TRILL is also emerging as a key component in advanced FCoE networks. HP, for one, is looking to leverage the protocol in its bid to extend FCoE to the core, where it will rely on TRILL's multipath and multihop capabilities for massively scaled environments.
As a new standard yet to be finalized by the IETF, it's hard to predict how successful TRILL will be in the open market. But one thing is clear: the multi-tiered structure of the past will not suffice for the fluid networks of the future. Something will have to take its place, and it looks like TRILL is next up to bat.