Ethernet Alliance

Ethernet Alliance Blog

TEF 2014: The Rate Debate

By John D'Ambrosia

The recent request to do a “Call-for-Interest” in the IEEE 802.3 Ethernet Working Group at the upcoming IEEE 802 July Plenary to address 25 Gigabit Ethernet is an interesting development in the on-going saga of Ethernet.  Its success can be attributed to a number of factors.  “Good enough.  Cheap.  It works.”  These are words and phrases typically employed to describe the success of the Ethernet recipe. 

Personally, I have always noted that I feel Ethernet’s success has been its ability to evolve to meet the needs of the market.  That evolution, however, has been fueled by the broad deployment Ethernet enjoys, where individuals have noted the needs of their particular application spaces.  The legacy of the IEEE P802.3ba 40GbE and 100GbE project may not be that it forever altered Ethernet’s 10x incremental improvement legacy.  Instead, it is the recognition by the market that this legacy is no longer optimal to support the breadth of the application space that Ethernet now services, each with their own unique cost structures and timing requirements.  Consider, for example, that the IEEE 802.3 Ethernet Bandwidth ad hoc forecasted that bandwidth growth is expected on average to experience a 58% compound annual growth rate (CAGR), but for the applications studied, the actual range was from 32% to 95% CAGR.[i]    

The justification for the 25GbE CFI is to form a study group to explore providing a single lane 25 Gigabit Ethernet (25GbE) specification for server interconnects to support server-ToR implementations based on the new 25GbE interface.   This justification is based on the development of the underlying 4x25Gb/s architecture that supports 100GbE.  Those familiar with the development of 40GbE may recall that 40GbE was identified as the next server rate back in 2007 by the then IEEE 802.3 Higher Speed Study Group.  While 40GbE has found much market success in data center networking, the 2007 forecast that the market would be shipping 40GbE based servers in volume by 2014 was optimistic, which is leading to speculation of the validity of the entire conceived rate scheme from the 40GbE / 100GbE project.   It should be pointed out that such debate now has the benefit of hindsight of what happened, as opposed to the situation at the time of the Higher Speed Study Group, which was required to make the best decisions it could, based on the information it had at the time.

Debating the decisions of the past can be somewhat amusing.  Looking to the future is another matter entirely.  If 25GbE can be justified, based on the 4 x 25 Gb/s architecture of 100GbE, then can 50GbE be justified if an 8 x 50Gb/s architecture is selected for 400GbE?  If that happens, then the following rate progression for Ethernet occurs after 10GbE: 10GbE, 25GbE, 40GbE, 50GbE, 100GbE.  And 40GbE and 50GbE are fairly close in rates – should both be introduced?  If so, can the industry develop technology that can be used to support both?   It needs to be remembered that the industry has finite resources, and will be taxed to support a multitude of projects and new technologies. 

Rate debates like this resonate throughout the entire eco-system…..

In my conversations with individuals, I hear people talking to the need for an industry roadmap.  The Ethernet Alliance tackled this topic in the past – see http://ethernetalliance.org/subcommittees/roadmap-subcommittee/.  Given the rate of change that the industry has been experiencing in the past 12 months, it is time to take a fresh look at the Ethernet Roadmap.  To that end, the Ethernet Alliance will be hosting a Technology Exploration Forum on October 16, 2014 at the Santa Clara Convention Center.   It will focus on the value of introducing new rates of Ethernet into the campus and data center application spaces for computing and networking.  This forum will bring together experts and key players in the Ethernet industry to discuss and explore the “Rate Debate.”  In addition to its own broad membership, which includes component vendors, system vendors, and end users, this event is open to non-Ethernet Alliance members.  Together, these two constituencies will provide a diverse set of backgrounds, ensuring lively discussion and debate.

I look forward to seeing you there.  For more information on the Ethernet Alliance TEF2014: The Rate Debate, please see http://ethernetalliance.org/events/technology-exploration-forums/.



[i] [i]      IEEE 802.3 Ethernet Bandwidth Assessment Ad Hoc, http://www.ieee802.org/3/ad_hoc/bwa/BWA_Report.pdf, July, 2012.

 

Share

The views and opinions expressed in this blog are solely that of the individual(s) and should not be considered the views or positions of the Ethernet Alliance.

Share

The Industry Discussion – 100GbE over a Single Lambda!

By John D'Ambrosia

This week the Ethernet Alliance, along with its partner the OIDA, is hosting a workshop on June 12-13 at the San Jose Convention Center to explore 100 Gigabit Ethernet per lambda for data center applications.  (For more information on the agenda, please see http://bit.ly/1fNDC8q.  To register for the event, please see http://bit.ly/1kQIOab.)

The genesis of this workshop began last year as the Ethernet Alliance was in the formative stages of planning for its Technology Exploration Forum 2013, which explored the future of Ethernet.  Part of that event’s agenda was a panel discussion on the development of the next optical eco-system, which it was argued, would be based on 100 Gb/s lambdas.  After our experts gave a number of excellent presentations (find those at http://ethernetalliance.org/tef-2013-the-future-of-ethernet-presentations/), I posed the following question, “How many of you feel this is possible – regardless of the timing?”  Every member of that panel raised their hand in agreement but as discussion continued two things became evident.  First of all, there were a number of technical challenges and hurdles that needed to be addressed, which included further analysis on signaling / modulation schemes.  Second was the need for this solution, and how the development would be paid for.  The conversation elevated even higher, as people realized that this was an industry level discussion and decision.

The need for this week’s industry workshop was clear.

Such discussion however cannot occur in a vacuum and must take into account the state of 100GbE today.  Currently, 100GbE is mainly being serviced by 100GBASE-LR4 that multiplexes four 25 Gb/s lambdas over one single-mode fiber up to 10 km.  Soon, 100GBASE-SR4, with individual 25Gb/s VCSELs, will send single lambdas over four multimode fibers to 100 meters.  At the same time, a plethora of optical multi-source agreements, or MSAs, have come into existence as the industry looks to accelerate the adoption of 100GbE.  While many of us involved in the standardization process realize that we can’t second guess the market, I find the lack of a single clear direction to drive volume up and cost down for the data center space troublesome.

I remembered something that Bob Metcalfe said last year during one of the many of the 40th anniversary of Ethernet celebrations, “Sometimes you need transitional technologies – but that’s ok.”  Words of wisdom from the Maestro. 100GbE did need a 4×25 Gb/s configuration to progress the industry forward.  However, historically the industry has seen optical solutions in the Ethernet space transition to a serial, single lambda solution.  Therefore we do need to start considering this evolution. 

Some may say that this is already being accomplished, as we see 100GbE technologies being supported via a single wavelength in optical transport networks (OTN) today.   This is done on the OTN line side however, not the Ethernet side or client side connection, which is supported via a number of Ethernet solutions that are based on either parallel lambda or fibers.  Furthermore, we see 40GbE being aggressively deployed in data centers today.   I have always said that the economics of an application will drive the solution that gets deployed, and the cited deployments of 40GbE and 100GbE are perfect examples of this concept.

The need for the Voice of Ethernet to be raised is clear.  Discussions targeting the development of 100GbE over a single lambda for the data center space must begin as this will provide the industry with the long-term solution it needs.  So come be part of the discussion this week at the ‘100GbE per Lambda for Data Center Workshop’! 

Share

The views and opinions expressed in this blog are solely that of the individual(s) and should not be considered the views or positions of the Ethernet Alliance.

Share

The 100GbE Lambda Workshop – Minimizing the Risk

By Tom Hausken

OIDA is pleased to co-host with Ethernet Alliance the upcoming joint workshop on 100GbE per lambda for data centers, and it promises to be different from other workshops that OIDA has hosted in recent years.

Note that is very specific: 100GbE, single-wavelength, technology, data centers.  We couldn’t describe it any other way.  “Per lambda” (or single wavelength) means it’s not about CWDM or parallel optics.  It means single wavelength.  “Technology” is not in the title, but could be, and means that it’s not about standards.  “Data centers” means it’s not about telecom.  And “100GbE” means just that—don’t miss the “E” in 100GbE, by the way.  This workshop is about how to make this particular technology happen.

We will talk about the end-user and system drivers, and the optical and electronic technology.  We will also talk about some of the market and non-market challenges to commercialization. 

A key challenge is the rationalization and stratification of the supply chain.  The days of vertical integration of AT&T, British Telecom, France Telecom, and others is long gone.  Today, the increasing investment required to commercialize new technologies like 100GbE is mostly pushed to the layer of component suppliers, companies that are not in a position to take on a lot of risk. 

That’s the real reason for the workshop: the cost to commercialization is high, so let’s get together and discuss what we can to minimize the risk for everyone.  We hope to see you there!

Share

The views and opinions expressed in this blog are solely that of the individual(s) and should not be considered the views or positions of the Ethernet Alliance.

Share