- A TechNote on The Next Generation
- Jim Metzler
- Distinguished Research Fellow and Co-Founder
- Webtorials Analyst Division
The Interop Session
As mentioned, there were four panelists on the STP session. One of the panelists was from Juniper and he advocated that if IT organizations implement Juniper's Q-Fabric that they could eliminate the STP. The session attendee was correct in pointing out the Q-Fabric is not standards based.
Extreme networks was on the panel and they advocated that IT organizations should implement Multi-Chassis Link Aggregation Group (MC-LAG). While much of basic link aggregation is detailed in the IEEE 802.1AX-2008 standard, that standard does not cover MC-LAN and hence MC-LAG implementations are vendor dependent. That said, most vendors support MC-LAG and in many cases it is possible to get MC-LAG solutions from disparate vendors to work together. As a result, it is possible to look at MC-LAG as a defacto standard.
Cisco was on the panel and they advocated that if IT organizations wanted to eliminate STP that they should implement TRILL (Transparent Interconnection of Lots of Links). TRILL is the result of an Internet Engineering Task Force (IETF) project to develop a Layer 2 shortest-path first (SPF) routing protocol for Ethernet. The process to develop a new standard is always lengthy. In the case of TRILL, the TRILL working group has been working on the standard for several years. The process is virtually completed and the working group is expected to either establish a new charter or disband the working group in July 2012.
Avaya was also on the panel and they advocated that if IT organizations wanted to eliminate STP that they should implement shortest path bridging (SPB). SPB is the IEEE alternative to TRILL and after a multi-year development process by the IEEE 802.1aq working group and the development of RFC 6329, SPB is a standard.
The Limitations of Standards
One of the limitations of standards is that they sometimes are not as explicit as they need to be. For example, OpenFlow 1.0, which is a key component of most SDN implementations, is often criticized as not clearly distinguishing between the protocol features that are required and those that are optional. Another limitation of standards is that vendors can implement proprietary extensions to the standard, which means that two versions of a standard will not interoperate.
One of the biggest limitations of standards is that, as previously mentioned, they typically take several years to fully develop. So, one of the key questions related to standards is "Is it better for vendors to wait until a standard is completed before implementing a new protocol or should they implement a pre-standard version of an emerging protocol?" Using both Cisco and TRILL as an example, Cisco implemented a pre-standard version of TRILL which of necessity, is non-standard. The good news is that if an IT organization implements a pre-standard version of a protocol such as TRILL, they will get some of the benefits of that protocol years before the standard is finalized. The bad news is that they will have a protocol that will not inter-operate with other versions of the protocol unless the vendor provides an upgrade path from the pre-standard version to the standard version.
Not Standards - Interoperability
One important fact that came out of the Interop session highlights a limitation of standards that is of growing importance in the current environment. That limitation is that vendors are beginning to choose which standards they will support and which ones they won't. For example, during the Interop session Cisco stated that they have no intention of implementing SPB and Avaya stated that they have no intention of implementing TRILL. Most likely it was the fact that, because they are implementing different protocols, the products from Cisco will not interoperate with the products from Avaya that was disturbing to the previously mentioned attendee at the Interop session on alternatives to STP. While that is clearly disturbing, both Cisco and Avaya can clearly claim to be developing standards based products.
Unfortunately, TRILL vs. SBP is not the only battle of protocols in which vendors are lining up to support only one protocol. Another protocol battle is brewing between VXLAN and NVGRE, both of which are intended to make it easier to move virtual machines between servers. In this case, a number of vendors including Cisco and VMware are lining up behind VXLAN while another group of vendors, including Microsoft and HP are lining up behind NVGRE.
The bottom line is that the industry does not really want standards, what the industry wants is interoperability. Unfortunately, we are in an era when achieving interoperability will likely be more difficult than it has in a long while.
This TechNote is brought to you in part due to the generous support of:
There have been 4 public interoperability tests for Shortest Path Briding. The link below talks about a interop test where SPB control and data plane interoperability were successfully demonstrated between 5 venodrs in a network of 200 nodes and 400 links.
http://www.convergedigest.com/voip/voiparticle.asp?ID=34093