So welcome to 2011. It has been a couple of months since I last blogged, and there are a number of activities underway that have me professionally pumped. You know what they say— enjoy what you do and you will never work a day! Let's do a quick run-down on all of the activities in play:
IEEE 802.3 100Gb/s Backplane and Copper Cable Study Group: First study group meeting was in January in lovely Fort Lauderdale, Fla. It should be an interesting project, as the typical channel versus modulation debates are emerging. How far will the reach be? What type of board materials will be considered? It is almost a repeat of the first generation Backplane Ethernet project.
IEEE 802.3 Ethernet Bandwidth Assessment Ad hoc: An ad hoc effort to gather information from the industry regarding future bandwidth needs. During the 2006 incarnation of the Higher Speed Study Group this took a considerable amount of time. Given all the on-going industry debate on the need for Terabit Ethernet, gathering the actual data demonstrating future industry needs should be exciting. The initial response has been encouraging, as end-users express interest in the effort, perhaps realizing the need to drive the industry for their future needs.
The final area I want to talk about is the next generation of 100Gb/s optics efforts. For multi-mode fiber specifications, the goal will be to reduce the current number of parallel fibers used in 100GBASE-SR10 from 10 (in each direction) to four (in each direction). The biggest controversy I expect here is the "reach" debate. In the single mode fiber (SMF) world the controversy is somewhat more intense as the first generation of 100GBASE-LR4 modules (SMF for at least 10km) is not meeting the cost and power expectations of some in the industry. A number of these individuals feel that a shorter-reach solution would be less overkill for their needs, and could consume less power and be less expensive. While "reach" is often the number that everyone can relate to, the optical budget is really the key issue. For this shorter-reach need, some are proposing a new optical solution based on 10 lambdas of 10G. Others are looking to the next generation of 100GBASE-LR4 modules, which look to pull the 10:4 gearbox out of the module by leveraging work underway in the industry on 25Gb/s electrical signaling. This could help to reduce the power and cost and improve port density.
The electrical interfacefor such an implementation, however, also has been a source of debate, as groups have begun to debate non-retimed versus retimed versus partially retimed solutions. Some think that the development of the electrical interface will drive the decision of whether a new optical specification is even needed. This would help to prevent market fragmentation, and help lead to lower costs by leveraging economies of scale.
The debate related to the SMF solution has been interesting. At an informal gathering of industry experts, individuals voiced their opposition to a project to target 2km over single mode fiber, based on a 10x10 solution. However, they were willing to consider starting a project to explore a new solution that could yield lower power and cost while improving port density over the first generation CFP implementations of the 100GBASE-LR4 physical layer specification. Note that in the first instance individuals were being asked to start a project assuming a given solution, while in the second instance the problem was being studied, and all potential solutions (such as reducing the reach, developing a new electrical interface or examining the number of lambdas, 10x10 vs. 4x25, to transport 100G) were left on the table for consideration until the project was started.
I see these decisions as a reflection of the industry's preference for the open IEEE 802.3 process, where problems are identified for study and then once a project is authorized, an open and fair discussion of all proposals is brought forward. What some fail to realize though is that the IEEE is not making any decisions, but is providing the open forum and rules for the industry to gather and debate the problems and potential solutions. These participating individuals are a reflection of the industry, and the speed at which decisions will be reached will mirror the consensus of the industry itself. When there is consensus, the standardization process can move quite quickly. But the IEEE process itself cannot guarantee that consensus will be created if none exists within the industry.
It should also be noted that the debate and scrutiny that happen within these groups drive solutions that work and are reliable. Those involved in the IEEE 802.3 Working Group are very conscious of the strength of their procedures and are zealous in seeing that they are followed. This is at the very core of the strength of the Ethernet specification. I look back at the IEEE P802.3ba Task Force that developed 40Gb/s and 100Gb/s Ethernet with pride. I know all of the hard work that was put into this specification by so many people, and want the industry to know and appreciate the effort and dedication shown by these individuals.
I did not mean to imply this. The reality is that to support the higher port capacities for line rate 40G and 100G front panel I/O, backplane capacities will need to increase. This will drive the jump from 10G to 25G.
thank you John, I presume that your answers means that there is a battle in the marketplace between people who can do 10x10 only for now and those who can do 4x25...which companies are on which side? Kris
It wasn't that long ago that 10G was "bloody expensive circuitry" and people were skeptical of it. However, industry involvement and investment has now made it a reality. Simply put, if we are to support the densities that end users are looking at, we will need to increase the speeds and feeds inside the box. Some argue that you don't need 4x25 at the board edge, however, if you look at a simple chip and consider the number of I/O it takes to get in and out, the simple mathematics of 8 pair for 4x25 versus the 20 pair for 10x10 start to add up. Then consider the backplane. You have to consider the entire system.
As Moore’s Law reverses beyond 28nm, consider network-on-chip (NoC). While more and more content in SoC designs is coming from third-party IP providers, interconnect-fabric is one area that is still in transition.
Join our online Radio Show on Friday 11th July starting at 2:00pm Eastern, when EETimes editor of all things fun and interesting, Max Maxfield, and embedded systems expert, Jack Ganssle, will debate as to just what is, and is not, and embedded system.