Google (Nasdaq: GOOG) has built and deployed its own 10-Gbit/s Ethernet switches -- a move that could have implications for the systems, chip, and optical transceiver markets.
Analyst Andrew Schmitt of Nyquist Capital was trying to trace shipments of SFP+ optical components and found the trail led to a Google-built switch. Schmitt, an industry analyst who has financial positions in Broadcom, Vitesse, Finisar, and AMCC, reported his findings in a column earlier today.
"This decision by Google, while small in terms of units purchased, is enormous in terms of the disruptive impact it should have on 10-GigE switching equipment providers and their component supply chains," he writes.
Google took the same DIY approach with servers a few years ago. In this case, its work has implications for switch vendors like Force10 Networks Inc. and Woven Systems Inc. , since Google not only couldn't find what it needed, but also built its boxes from off-the-shelf chips.
Moreover, Schmitt believes Google is taking an unconventional approach with its choice of optical interfaces.
The events that led to the high prices have really opened a unique window to make this sort of product possible. Wall Street expects margins well over 50%, and businesses generally bow to Wall Street's whims, no matter how pointless. Of course the company does also get the benefit of gobs of operating capital. Cisco has used optics modules in particular as the means to increase their margins, by jacking up the prices on components that are equivalent to those available on the street. Since they are clearly the same optics, every IT guy thinks -- hey, I'm getting ripped off for these things, generating resentment against Cisco.
The concept of building your own ethernet switch is a daunting one for a real world business which isn't in the networking business. It isn't that the design of the switch is all that difficult -- I could put together a small team to build any product that Cisco makes, and the resulting product would be better and cheaper. But, it's a much better idea if I focused on my own business because it would take too much time just to reinvent the wheel.
To make it even remotely viable, you'd need large enough quantities to recoup the up front costs (probably something in the low 7 figures $, maybe less if the system is small & dumb enough). You'll also need to make it simple, so you don't need to spend a lot on maintenance, repairs, bug fixes, internal tech support. This is really the difference between a "workgroup hub" and an "enterprise switch" -- once the amount of services running over that switch get large or critical, you start having to worry about remote network management and integrating it into your OSS, etc. This introduces significant amounts of software and complexity, and it spins out of control.
If that is what you need, then the cost becomes too high, and it always makes sense to just buy it. My guess with Google, is that they have lots of small clusters which only need a few very fast ports, and are not using this for aggregation points. That would give them the high quantity of switches needed to recoup one-time costs, and a port count which is >2 but much fewer than 24 would mean that they wouldn't need to populate many of the I/O modules, which would give them a significant advantage in cost of goods in addition to avoiding the IOS tax. If it is on a controlled access net, then the filtering is all done at the access points and the switches don't need to have that feature enabled either.
I challenge whether end-users request the features rather that vendors have painted themselves into a corner and tout their margins (to wall street) and justify their higher prices (to their customers).
Google, as a company, is in a unique position to drastically change business models - I repeat myself when stating look at what they've done to the server DIY market as an example of their purchasing power.
If they hit on a magic formula for an inexpensive (and what they need as fully featured) commmodity switch, you can bet its impact will be felt right down to those start-ups you talk about.
>>how many of those value-add features do >>>customers really use? Whether customers use it or no, they look for lot of features, without those features essentially networking vendors cannot get those accounts. Especially startups.. Google is very frugal when it comes to buying equipment. A failed search is not the end of the world. It is really no different than a consumer who wants all the great features in a digital camera which never really get used, and only uses his camera for basic point and shoot 95% of the time! But without those features there is no purchase.
Here's an interesting question that maybe an interesting story angle.
Of all the value-add features the networking vendors position to justify their price point, how many of those value-add features do customers really use?
Acording to the Nyquist Capital story, they make it look like a streightforward implimentation of Broadcom's BCM56800 20 port 10/1G Etherenet switch chip. The only special sauce is using SFP+ and twinax. Between Broadcom's demo systems and a contract manufacturer not a rocket science.
They built their own ethernet switch which only supports 1/10 of the features of everybody else. I don't think any networking equipment companies need to worry. Google probably is spending too much on hiring the engineers to build this, but they have plenty of money.
There's no rocket science associated with standard switching technology - it's all commodity functionality.
In the same vein that Google builds their own off-the-shelf servers, they most definitely can (and probably will) build their own off-the-shelf 10GigE switch.
Building this has nothing to do with intelligence. It's all about saving $$$, which IS something that they do know quite a lot about having done it with computers, and appear to be ready to repeat again.
I'd get ready to start hearing marketing flunky spin from networking companies on how/why they justify their 50% plus product gross margins for value added features.
The discussion around Google's ventures into just anýthing oozes with lack of understanding of the halo effect.
Google has a tremendously successful and profitable search engine. As for the rest of Google's acitivitites, who knows whether they are smart?
The comparison with ATT during the monopoly years cannot be easily dismissed. Google has so much money, and have been expanding its number of employees so rapidly, that they pretty much have to expand into every neighbouring area. How else can they keep all these people busy? My own connections into Google tell me stories of absolutely ridiculous projects, in the line of designing their own coffee machines.
There is very little pressure on them to proof that they are doing the right things, as Google is still considered generally flawless.
Building their own equipment might work, but I suspect mostly because of the classic case of not including everything in the calcualtions. Google compares the hardware and build costs with the complete cost from the vendors. In the long run they will have massive support costs, but internal costs are always harder to quantify.
Actually, the lower the volumes the less attractive it is for Google to try to roll their own, unless there is some super function that they can't get a vendor to implement.
My guess is that the volumes are high, but the vendors aren't willing to drop their pants and bend over. Once you drop your pants for one customer, word gets around and everyone thinks you are easy.
As I used to say when I was in the business, we will bend over backwards to make our customers happy, but we will not bend over forwards.
The blogs and comments are the opinions only of the writers and do not reflect the views of Light Reading. They are no substitute for your own research and should not be relied upon for trading or any other purpose.
To save this item to your list of favorite Light Reading content so you can find it later in your Profile page, click the "Save It" button next to the item.
If you found this interesting or useful, please use the links to the services below to share it with other readers. You will need a free account with each service to share an item via that service.
Digg
Del.icio.us
Reddit
