From my time working on various networks, I have come to think of Ethernet as having different flavours. Although the actual technology is more or less the same, the features of various switches and how it applies Ethernet can vary, sometimes significantly.
The first flavour of Ethernet should be the most familiar. The fundamental purpose of Enterprise Ethernet is to provide a platform for corporate applications and information systems. Enterprise LAN networks are essentially where Ethernet gained its momentum to become the dominant data link protocol that it is today. The approach to Enterprise LAN switching design has become standardised, despite new features being rolled out by vendors keen to innovate. However, as Gartner notes in their 2011 Enterprise LAN Magic Quadrant:
“A significant portion of the market hasn’t used the sophisticated features available in many vendors’ products, and now faces upgrades as their installed bases age.”
Interestingly, vendors still command a 60-65% gross margin on Enterprise Ethernet for what is essentially a mature market.
Data Centre Ethernet
The use of Ethernet in the Data Centre has evolved into its own flavour in recent years. Vendors such as Cisco with their Nexus platform and specialists such a Force 10 have targeted the Data Centre as a separate entity to Enterprise Ethernet. Data Centre cores consist of high speed and high port densities and in my opinion will see adoption of 40GE sooner than service providers. Edge switches, or top-of-rack/end-of-rack tend to feature hot swappable redundant power supplies and fan units not normally seen in less expensive Enterprise switches. There are also more subtle differences in the hardware, such as faster non-blocking backplanes and deeper buffers.
There is also a focused effort to tweak Ethernet in order to overcome some of its inherent issues. The asynchronous, loss prone nature of Ethernet does not inspire confidence when the focus is on high speed and high capacity service delivery. Some interesting areas include:
- Priority-based Flow control (802.1Qbb) seeks to introduce flow control into Ethernet so that congestion can be managed without the need for higher layer protocols.
- Shortest Path Bridging (802.1aq) selects the best path in a redundant layer-2 topology without being hamstrung by spanning-tree or relying on the network layer.
Finally, there is Fibre Channel over Ethernet, the much flaunted yet elusive promise of a unified data and storage network. Although slow on the uptake, momentum is slowly building on the back of the Data Centre Bridging (DCB) tweaks outlined above.
Service Provider Ethernet
Ethernet has also revolutionised Service Provider networks. Legacy technologies such as ATM and SDH/SONNET today are far less attractive in terms of cost and flexibility than Ethernet based MPLS cores. At the distribution level, Ethernet based MANs (or more commonly Metro-Ethernet) have become an easy, low cost transport method for high speed services. In fact, the service provider industry has been so uprooted by Ethernet that the Metro Ethernet Forum (MEF) was created to try to standardise design approaches.
Ethernet is also present at the Service Provider access. It is often encapsulated via PPP, especially in xDSL based networks. In legacy ADSL networks it is common to have Ethernet, encapsulated in PPP, transported by ATM. However, newer access platforms have dropped the ATM transport in place of Ethernet, where 802.1Q in Q tags can be used to create point-to-point VLANs. When incorporated with some additional traffic shaping, this allows Ethernet to look and act similar to ATM PVCs.
Industrial Ethernet focuses on providing a platform for infrastructure based information systems. With the ubiquity of both the Internet and Ethernet, many infrastructure based information systems are migrating from proprietary systems to TCP/IP stack with an Ethernet network interface. Although this allows these systems to be networked easily, their unique characteristics have given rise to an Ethernet flavour of their own.
Industrial Ethernet devices are environmentally hardened to operate in outdoor environments. They often have an extended operational temperature, in the range of -40C to 75C. Some devices are also water and dust proof with rating of IP67 and higher. Devices that are deployed on main roads or rail sometimes use vibration resistant M12 connectors in place of the default RJ-45.
The physical topology of devices in industrial networks is also different. Instead of having star like core, distribution and access topology with redundant links, Industrial (and for that matter Service Provider) Ethernet is often built on interconnected backbone and sub-rings. This means that industrial switches often support layer-2 redundancy protocols more suited to ring topologies such as ITU-T G.8032 Ethernet Ring Protection Switching.
Some vendors are innovating further, providing enhanced protocol level security for SCADA based systems. This allows for a distributed layer-7 security policy to be implemented down to the port level. Consider a SCADA water control network that had a distributed security policy which only allows write level commands from specific Ethernet terminals. This level of security it vital on information systems that control infrastructure. It is doubly so when you consider that these systems, once proprietary and stand-alone are being connected with enterprise, corporate or government networks and even the Internet.