OIF Gives 40 Gig a Boost

What could turn out to be a milestone in the development of 40-Gbit/s transport equipment is about to be announced by the Optical Internetworking Forum (OIF).

The achievement -- the result of a ballot of OIF members concerning its so-called SerDes-Framer Interface 5 (SFI-5) -- may sound a little arcane. But it paves the way for system vendors to mix and match chips from different suppliers in their 40-gig systems; and that promises to encourage semiconductor manufacturers to charge ahead with developments, knowing there's a free and open market for their 40-gig products.

"It takes the risk away," says Tom Palkert, a member of the board of the OIF and a systems architect with Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC). "Whether the market is there is another matter," he cautions.

Some vendors have already developed products based on the SFI-5 spec, on the assumption that it would gain the approval of OIF members in last week's ballot.

Agilent Technologies Inc. (NYSE: A), for instance, already has a tester for evaluating SFI-5 interfaces. The so-called ParBERT (Parallel Bit Error Rate Tester) measures the way in which the 40 Gbit/s of traffic is split into 16 channels, each running at 2.5 Gbit/s, and synchronized using two extra channels of signaling info.

"The SFI-5 interface is a key specification for the creation of industry standard 40G transponders," writes Larry Desjardin, Agilent's high-bandwidth program manager, in an email to Light Reading. "This will help the economies of scale, a key success factor for 40G."

Agere Systems (NYSE: AGR) "has multiple devices currently in development that will support SFI-5," writes Glen Haley, Agere's media relations manager, in another email. "Agere absolutely believes that the OIF SFI-5 interface will be very relevant to next-generation solutions for multiport 10G applications that become migrateable to OC768C applications."

Not everybody agrees that SFI-5 will be important, however. Bill Woodruff, vice president of marketing for Velio Communications Inc., sees only "very narrow applications" for it. One of his reasons for saying this is that most system vendors won't want to use a standardized FEC (forward error correction) chip, in his opinion, because proprietary FEC technologies have far better performance. "People kill for a half decibel in this business," he hyperbolizes.

To put things in perspective, the OIF is standardizing four interfaces between the chips found in a typical transponder.

. The current status is as follows:

  • Very Short Reach (VSR) interfaces, between SerDes and optical devices, have already been standardized (see OIF Promotes VSR Specs).
  • SerDes Framer Interface (SFI), between SerDes and framer (FEC or PHY) chips: The OIF has already standardized a 10-Gbit/s version, called SFI-4 and is working on a more compact version called SFI-4, Phase 2. Now it's settled on a 40-Gbit/s version called SFI-5.
  • System Packet Interface (SPI), between network processors and framer chips: Standards already exist for a 10-Gbit/s version, SPI-4, and a compact 10 Gbit/s version, SPI-4 Phase 2.
  • TDM Fabric-to-Framer Interface, between switches and framer chips, is still a work in progress.

    — Peter Heywood, Founding Editor, Light Reading
  • AutoDog 12/4/2012 | 10:17:03 PM
    re: OIF Gives 40 Gig a Boost Either the Bill Woodruff of Velio has no clue what SFI-5 is, or he was quoted out of context... Yes, folks will kill for an extra 0.5dB of FEC coding gain, but that doesn't mean SFI-5 is thrown out the window.

    On the contraty, SFI-5 is FEC agnostic... Vendors can implement whatever FEC algorithms they like and still use SFI-5, since it is specifies an electrical interface over a relatively wide speed range. SFI-5 doesn't know or care what data is actually being carried over it.

    Suffice to say, when (not if) 40G takes hold, SFI-5 will be a cornerstone of most implementations.
    woodruff 12/4/2012 | 10:17:02 PM
    re: OIF Gives 40 Gig a Boost I appreciate Autodog's comment, "Either the Bill Woodruff of Velio has no clue what SFI-5 is, or he was quoted out of context." One or the other will be right ;-)

    SFI-5 will be critical for optical module to FEC or framer interface. When (I agree it is when, not if) 40G takes off, SFI-5 will be a critical enabler. But as data moves around inside boxes, SFI-5 will be used at the module interface, while other standards, such as SPI-5 or TFI-5 will be used to connect devices on the line card, and to connect to the switching cards. Thus, the narrow applications comment. Systems with 40G data paths are being implemented today, which will have more SPI-5 and TFI-5 connections than SFI-5.

    AutoDog 12/4/2012 | 10:17:02 PM
    re: OIF Gives 40 Gig a Boost Fair enough... I suspected you were not implying that SFI-5 wouldn't be used at the serdes/FEC interfaces.

    Another important note that the article didn't mention was the SxI-5 spec that should be adopted at or near the same time as SFI-5. SxI-5 will be the common electrical spec that both SFI-5 and SPI-5 will be based upon.
    lacrosse 12/4/2012 | 10:16:54 PM
    re: OIF Gives 40 Gig a Boost 40 Gig
    bad boy 12/4/2012 | 10:16:38 PM
    re: OIF Gives 40 Gig a Boost Bill,

    What's the big difference between SFI-5 and TFI-5? I Understand that Velio is major force behind TFI-5 at OIF but not very clear on what specifically differ between these two.

    Sign In