The telecom industry is facing growing bandwidth demand due to 3G and 4G wireless as well as video traffic.  To meet this demand, the industry is preparing to move from today’s 10G systems to higher-bandwidth 40G platforms.  Although the 40G AdvancedTCA (ATCA) standards are not yet finalized, designers can—and should—start designing for 40G today.  The shift from 10G to 40G ATCA is expected to begin this year, with full-fledged deployments starting in 2011.  Given that ATCA rollouts typically require a year or more, network equipment designers must start 40G designs now if they are to meet these timelines.


We will show how to get started on 40G in a moment, but let’s first review 10G ATCA technology.  Today’s 10G ATCA backplanes are based on the PICMG* 3.1 Option 9 standard.  As shown in Figure 1, the PICMG 3.1 Option 9 backplane implements a 10GBase-KX4 link, which is comprised of four serial lanes running at 2.5 Gbps each.



Figure 1. The 10GBase-KX4 backplane link is comprised of four serial lanes running at 2.5 Gbps each.


Moving forward, ATCA backplanes will be upgraded to meet the 10GBase-KR standard, which achieves a throughput of 10 Gbps per lane.  With the upgrade to 10GBase-KR, the backplane will offer four 10 Gpbs lanes.  The final step in the transition to 40G is to tie the four lanes together to create a single 40 Gpbs 40GBase-KR4 link.  These steps will take some time.  The 10GBase-KR and the 40GBase-KR4 technologies are expected to be ratified as part of the PICMG 3.1 R2 ATCA specifications in 2010.  10GBase-KR technology is expected to come to market in 2010, while 40GBase-KR4 systems are expected to appear in 2011.


As should be clear by now, 40G backplanes will largely resemble today’s 10G backplanes.  The primary difference is that the serial lanes will run at 10 Gbps instead of 2.5 Gbps.  The other key difference between 10G and 40G is power.  40G cards will need increased processing power, and that means more heat.  Thus, 40G shelves must be rated for higher heat dissipation.  Among other things, this means leaving more space between cards for better airflow and larger heat sinks.


While these are not trivial issues, the fact remains that 10G and 40G platforms have much in common.  In fact, some of today’s 10G platforms have been verified as “40G-ready”—that is, the platforms have been designed and tested for 40 Gbps traffic.  Examples of these platforms include the Emerson Centellis* 4440 and the RadiSys ATCA 4.0 Platform, which is based on the Promentum* SYS-6016.  Both of these platforms feature backplanes that support 10GBase-KR and 40GBase-KR4.  These platforms have also been designed with the increased cooling needs of 40G in mind. For example, the Centellis 4440 provides cooling that conforms to the CP-TA B.4 thermal profile, the highest level currently defined. This cooling ability dovetails with the shelf’s power distribution to allow blade power levels up to 350 Watts per board.


Of course, the shelf is not the only thing you need to get to 40G—you also need 40G switches and payload cards.  These cards are not available today, so you will need to start development with 10G components.  Once 40G switches are released—which is expected to happen later this year—you can replace your 10G switches.  Then when 40G payload cards are released in 2011, you can upgrade those cards as well.  Figure 2 illustrates this upgrade process.  Note that the chassis, cabling, and system installation remain intact throughout this process, allowing much of the integration and verification efforts to be re-used. The result is a reduction in the cost, complexity, and risk (as compared to starting a 40G design from scratch).

































Figure 2. The 40G upgrade path, using the RadiSys platform as an example.


To follow this path, you must first select a 40G platform that is compatible with existing 10G cards.  You’ll also want to look for platforms with pre-validated 40G and 10G components to avoid interoperability issues.  This compatibility is obviously essential for near-term deployments using existing 10G cards.  It will also serve you well over the long run, as many deployments have elements with lower bandwidth requirements.  Using 10G cards for these less-demanding elements is a great way to keep costs down.


It is also a good idea to look for vendors with scalable solutions.  For example, Emerson offers the 2-slot Centellis* 2000 in addition to its 14-slot Centellis 4440.  (See Figure 3.)  Having compatible platforms in different sizes makes it easier to meet the needs of different customers and different applications.  Many elements of an ATCA platform core are essentially unaffected by the platform’s size.  Thus, it is possible to re-use development efforts when moving between solutions of different sizes when using compatible platform bases.



Figure 3.  Emerson offers two 40G-ready platforms: the 14-slot Centellis 4440 (left) and 2-slot Centellis 2000 (right).


Finally, you should consider the partner ecosystem when selecting a 40G solution.  A comprehensive, tightly-integrated ecosystem can shave months off your development time by providing system management software, control and data plane software, and more.


The bottom line is that you can get a head start on 40G by making wise use of today’s 10G platforms.  For more reading on this topic, I highly recommend Emerson’s white paper Get Ready for 40G ATCA (registration required) and the RadiSys paper A Smooth Transition to 40G.  I also suggest Emerson’s recent blog 10G Deployments, 40G Available and new applications for ATCA.


So… are you ready to start designing for 40G?  If you’ve already started, what challenges have you encountered?  What solutions can you share with the community?  Let us know by providing feedback in the comments section below.


Emerson Network Power and RadiSys are Premier members of the Intel® Embedded Alliance.



Kenton Williston

Roving Reporter (Intel Contractor)

Intel® Embedded Alliance


Embedded Innovator magazine