The Co-Packaging Framework IA (Implementers Agreement) effort just announced by the OIF is a strong endorsement for co-packaging; a technology platform whose star is in the ascendant. Co-packaging was – as recently as two years ago — considered a futuristic platform, but much attention has been brought to it as the hyperscale data center managers start thinking seriously about the optical engines that will take them into the 800G+ era.
Signs of the times are that the OIF itself has run a conference on co-packaged optics as has EPIC and the co-packaged technology platform will also be featured at the upcoming ECOC event. CIR is already engaged in putting together a market assessment of co-packaged optics (See: HERE) covering a broad swath of applications and approaches to co-packaging. Still this is very early days for co-packaging and nothing has yet been pinned down on the standards, technical or application sides of the co-packaging story.
With this in mind, here are a few of CIR’s thoughts on open question co-packaging standardization. These thoughts are inspired by the list of priorities set out in the OIF press release about the launch of its co-packaging efforts.
Putting applications first: The fact that OIF began its list of important standardization efforts with “applications” is encouraging. There have been many communications standards in the past that don’t focus on real world applications. Sometimes if the applications for new technologies aren’t considered early on, these technologies are soon confined to the proverbial “dustbins of history.”
Similarly, CIR is encouraged by the appearance on the scene of the Co-packaging Collaboration, which is an end user organization and therefore, by definition, applications oriented. No doubt some industry groups in the semiconductor industry will also want to get into this party. Applications will always be given top priority in standardization in principal – it is the thing to do. But unlikely this is an area that OIF is going to go into in depth. It is just not what OIF does.
How far will standardization go? Big questions remain. Will the OIF group really extend its interests to all potential applications for co-packaging or will it just cover applications in hyperscale data centers and HPC?
OIF is a communications-oriented group, but we note that the literature on co-packaging suggests that there are applications for co-packaging in metro networks, edge computer sensors, storage, aerospace and the video industry. Will the OIF standardization group consider these applications and, if not, who will consider them as far as standardization goes?
We think that it is most likely that at first the OIF will focus almost entirely on 800G/Terabit networking – there is so much work to do. In doing this we expect the OIF to work with the Collaboration and perhaps with the IEEE 802.3 in some way. Other applications for co-packaging are probably get no more than a mention at first. But as co-packaging finds its way into other areas, other standards may be developed based to some extent on the work that the OIF will be doing.
Interoperability and industry politics: Interoperability is specifically mentioned in the OIF release, but is often easier to demand than achieve. This is in part because interoperability is promoted by specifying as much as possible, but this is easier said than done. Two modules that are fully and deterministically designed to be the same will by definition be fully interoperable.
But commercial vendors don’t usually think much of that idea. It prevents them from distinguishing their products in the market space. In addition, vendors have some pretty fixed ideas about using their favorite technologies and embodying them in the standards. Compromises are made along the way that can make interoperability more difficult to achieve. In the case of co-packaged optics, for example, where the light sources are placed (external or internal) may not be specified in any future co-packaged standards.