The Lightwave Blog

Posted by Stephen Hardy

Here's a belated update to the last post on SFP+ optics for Google's DIY 10GbE switch. (No, I'm not going to link to it -- just scroll one item down!)

As I mentioned in the post, I shot an email to Google to ask if the stories about the existence of the switch were accurate and, if so, what they could tell me about the optics. I didn't have a lot of hope of getting a response, particularly when my query, directed at the media relations department, elicited a canned email from the Help Desk in reply.

However, lo and behold, I received an email from Google's Sonya Boralv (who is a frequently quoted Google media relations person) the very next day. Her response:

We can't share details on our infrastructure but would encourage any interested engineers to check out our job listings at www.google.com/jobs. We currently have job openings for hardware and software engineers with networking backgrounds that might be of interest to your readers.

So I've struck out so far on the optics story -- but I may be the gateway to your next career. 10% of your first paycheck there will do just fine as thanks.

Labels: Google, SFP+

Posted by Stephen Hardy

As you've no doubt read elsewhere, Andrew Schmitt of Nyquist Capital broke a story on his blog (and if his blog isn't on your regular visit list, it should be) about Google building it's own 10GbE switch. (Read it here.)

Andrew and I have been going back and forth over the past couple of days about his description of the optics used in the switch, which he says are nonstandard implementations of SFP+ transceivers:

What is interesting about Google's approach is that it has eschewed traditional 10GBASE optical standards and instead adopted off-standard solutions that better suit its needs for time-to-market, power and port density, and cost. While Google makes use of the SFP+ cage format, it does not use the receive dispersion compensation (EDC) function typically associated with SFP+. Instead Google is looking to employ a combination of twinax cabling for short reach (<10m) intra-rack cabling and a motley 850nm SR-like standard. Off the shelf SR optical modules appear to work well up to 100m over without receive equalization. Ironically, Finisar (FNSR) proposed such a solution several years ago.

It sounded to me that either the assumption is that Google is using SR transceivers with 62.5-micron cabling (which indeed would be nonstandard -- and curious) or that SFP+ has gotten a bit too closely associated with LRM, to the point that SR versions might be considered nonstandard. Andrew tells me that "(t)hey are using standard pre-emph/receive EQ but not the full LRM EDC spec," so that's not the explanation.

Unfortunately, Andrew has been too busy to continue our discussion beyond that last bit of information. I have a query into Google on the subject, but while I was holding my breath and turning blue waiting for them to respond, I thought I'd throw the question out to you: What am I missing here?

Labels: Google, Schmitt, SFP+