Does OIF have to do everything?
Ransom Stephens- May 22, 2012
Is there a paranoid conspiracy among the producers of technology standard specifications? Was I once a co-conspirator? Did they give me a mind wipe when I stopped working for The Man?Why do spec’s leave gaping holes? Is there some political or legal reason that they don’t want to commit? Is this why a 3rd party has to come in? Is it a liability thing?
The Optical Internetworking Forum has just started three new electrical interface projects:
• The Ultra Short Reach Electrical Interface defines a link of 10 mm or less between an ASIC (application specific integrated circuit) and an optical engine at data rates up to 56 Gb/s.
• The Close Proximity Electrical Interface defines a link up to 5 cm for low power chip-optical engine board connections at up to 56 Gb/s.
• The Very Short Reach modulation format (CEI-56G-VSR) will determine optimal modulation formats for single lane electrical I/O at rates above 28 Gb/s for chip-to-module applications targeted at 40 Gb/s modules.
It looks to me like the OIF-CEI (Optical Internetworking Forum - Common Electrical Interface) committee is picking up loose ends, bless them.
Reading specifications is sort of like reading the terms and conditions of a credit card. It takes more than expertise to divine their meaning; it takes perseverance. In their zeal to avoid misinterpretation, authors of specs are repetitive, self-referencing, redundant, rarely clear, and always boring. There are gems to be found but mining them requires a sharp pick and tolerance for lots of dust.
Though we might hope specs guarantee that compliant components will interoperate seamlessly, what we really expect is guidance and if you put in the hours and work through the document you should get it.
I read the 40/100 GbE spec in November of 2011, some 14 months after its release. At the IEEE 802.3ba website, a banner says “The work of the P802.3ba 40 Gb/s and 100 Gb/s Ethernet Task Force is now complete with the approval of IEEE Std 802.3ba-2010 at the June 2010 IEEE Standards Board meeting.”
Complete? Not exactly. Don’t get me wrong, there are some shiny objects in it - I ‘m particularly fond of the gearbox that shuffles problems with skew up a layer in the stack - but one thing stood out.
The most intriguing 40/100 GbE configuration is four lanes at 25 Gb/s each. However, there is no guidance in the spec for developing electrical channels at 25 Gb/s. I suppose the omission implies that ASIC-to-OE should remain proprietary; that those pieces of the system have to comply as a unit and maybe I’m a techno-socialist whiner in this regard (I’m working up to a blog entry arguing that the very existence of standards is techno-socialism), but they said it was finished.
So where do you turn? A year after the release of 802.3ba (the 40/100 GbE spec) the OIF-CEI put out “agreements” for propagation of 25 Gb/s over 686 mm (or less) of PCB.
The OIF mission, from oiforum.com, includes: “The Optical Internetworking Forum (OIF) promotes the development and deployment of interoperable networking solutions and services through the creation of Implementation Agreements (IAs) for optical networking products… [and] will provide feedback to worldwide standards organizations to help achieve a set of implementable, interoperable solutions.”
Agreements are fine. We don’t care what they’re called as long as there’s hope that your widget can work in the same system as that guy’s over there.