Is Spanning Tree Protocol in its Twilight Years?
Spanning Tree Protocol (STP) and its modern variants (RSTP/MST) have been the standard for layer 2 (switched/bridged network) loop prevention in the enterprise network for quite a while. Spanning Tree Protocol is not only a requirement for providing path redundancy, but also for protecting the LAN from accidental loop introduction.
Recently I was working on a data center design and the vendor recommended that the client use EAPS (Ethernet Automatic Protection Switching) instead of RSTP. While I won’t pontificate on the pros and cons of this design, I was quite surprised at this recommendation since EAPS is typically most useful in the Metro Ethernet arena where the ring topology of SONET make it a great fit.
As we look at the layer 2 domain, there are a few interesting technologies emerging and lead me to the question: is traditional STP in its twilight years?
Let’s look at some of the problems with STP. Please note: when I say STP I am lumping STP/RSTP/MSTP into one category.
- STP cannot multipath on the same VLAN across all paths. In other words, in a redundant topology, which is STP’s purpose, certain links will be in a blocking/discarding state.
- Convergence times in networks where ultra-low latency is a requirement, such as the trading/financial industry, can be unacceptably slow.
- Poorly or misconfigured STP is a huge security and performance risk.
So what are some alternatives to STP? The short answer is that the industry has absolutely not settled on anything. However there are a number of emerging technologies that better align with the goals of the modern data center. Here are just a few:
- TRILL (Transparent Redundant Interconnect of Lots of Links): Wow that’s a mouthful! In short, TRILL uses a link state protocol (IS-IS) to find the best paths around the network and provide a map so that a loop-free layer 2 multi-pathing can be achieved.
- Virtual Port Channel (vPC): A Cisco proprietary technology that creates a link aggregation group (vPC Domain) similar to a LACP/LAG bond except that the link may terminate to separate bridges. In other words, the LAG group need not be limited to a set of links between ONLY two switches.
- SPBM: Shortest Path Bridging can be thought of as STP on steroids. It offers significant scalability gains and, unlike traditional STP, can scale to a network span of hundreds of bridges.
So now what? We understand STP to be increasingly less and less relevant in the modern data center, but the industry has not quite settled on a replacement. My advice is to sit tight. If at all possible leverage a properly tuned and configured STP and implementation. Don’t dive into any of these emerging technologies unless you keenly identify one that meets your business needs.