Convoluted 20/80 Rule in Data Center Optics

Avatar photoPublished on
blog 1486467627 33

Actually, it is probably a lot closer to a 90/10 relationship in which there is unbelievable amount of attention given to high-end optical development spurred on by a small number of mega-Data Center (DC) operators, which may not represent much higher than 10 percent of the total DC infrastructure market. At least the dot-com bubble initially came about with a legitimate bandwidth driver – the Internet. This time we have a cloud bubble based on relatively limited volume being promoted by a tiny minority of people in the industry with a vested interest including from small optical component vendors – of course, in some cases, being funded by those large end-users.

The hype surrounds the new cloud model supposedly being a game changer, and that it will result in a ramping up to higher optical speeds quickly in the next few years. There is also the assertion that the different set of economics compared with the enterprise or service provider markets is an important consideration. However, the amount of investment in these niche suppliers is rather modest, and the cost pressure being more intense in this cloud sub-sector only means that it will be harder for any componentry company to make any money.

Microsoft wants the industry to get enthusiastic over its projection of 1,000 to 10,000 leaf-to-spine links (typical reach of 500 meters) a year, while it will be throwing out the optics every three years, and it is looking for an obnoxiously low $1 per gigabit. Even assuming five new data centers annually, and the maximum number of connections, how is a volume of 50,000 devices supposed to thrill the more than 10 existing component vendors in the space? Reinforcing the notion of a bubble is the current trend of these large enterprises to go with fewer and larger data centers.

In spite of their rhetoric of the great support for the large components firms, such big enterprises do not appear to be concerned at all about their potential role in the possible disintegration of the optical ecosystem. There seems to be a lack of respect for these corporations as there is apparently this belief that they can magically devise a sufficient business case regardless of the fact that each data center environment evidently wants its own individual standards.

One suggestion that can be heard in the market to the problem of tailor-made requirements is to look at the DSL modem business for guidance. Under this scenario, there would be a whole slew of players (a lot of them in Asia), providing customized solutions. In a country like China, in which there is a tremendous amount of low cost labor, it is asserted that it could be a solution for optics as well.

Yet, in the long term, this kind of business structure is bound to have calamitous consequences. As we recently pointed out, the government of China still has not learned the standard lesson of controlling every facet of life. Specifically, easy access to credit, allowing anyone to create a company, has resulted in a lot of entities facing inadequate consumption and excessive capacity – thus, the regime is looking for a graceful way out of its situation.

[written by Mark Lutkowitz]

(For more insights on Microsoft’s investments in optical component companies, please click here.)