Google this week introduced its first Wi-Fi router and my initial reaction was "Why?" Wi-Fi access points and home routers tend to be low-margin commodity products that could only hurt financial results for the search giant. What made it worth the pain on Wall Street, then, for Google to introduce this gizmo? And then I realized it is Google’s best hope to save the Internet… and itself.
Wi-Fi is everywhere and it generally sucks. Wi-Fi has become the go-to method of networking homes and even businesses. I remember product introductions in New York back in the 80s and 90s when we were told over and over again that it cost $100 per foot to pull Ethernet cable in Manhattan (a price that was always blamed on the local electricians union by-the-way). Well the lesson must have stuck, because more and more Ethernet is for data centers and Wi-Fi is for everything else. Even my old friend, Ethernet inventor Bob Metcalfe, has started claiming that Wi-Fi is Ethernet, which it isn’t and he knows that.
The contrast between these technologies is stark. My three sons have been developing a computer hardware product they’ll be throwing up on KickStarter in a couple weeks and the lab they’ve built next to the foosball and pool tables in their man-cave (that’s what they call it) is entirely hard-wired with gig-Ethernet and what a joy that is. Lights flash and things happen exactly the way they are supposed to while the rest of the house is wireless and in constant networking turmoil. We have an 802.11ac network with no parts more than a year old yet still the access point in Mama’s room (one of five) loses its mind at least twice a day.
Wi-Fi is a miracle and it’s not going away, but what we have today is generally a hodgepodge of technologies and vendors that kinda-sorta work together some of the time. Every Wi-Fi vendor claims interoperability while at the same time making the point that if you buy your equipment only from them and stick with the latest version (replace everything annually) it will work a lot better. A single old 802.11b device, we’re told, can bring much of the network back to 1999 speeds.
The problem with Wi-Fi, as I understand it, is that 802.11 is a LAN technology developed with little thought to its WAN implications. How many of us run local servers? So nearly everything on a Wi-Fi network has to do with reaching-out 30 or so hops across a TCP/IP network that wasn’t even a factor when Wi-Fi was being developed 20 years ago by electrical (rather than networking) engineers. As a result we have queuing and timing and buffering problems in Wi-Fi that make bufferbloat look simple. These problems exist right down to the chip level where the people who actually know how to fix them generally have no access.
So what does this have to do with Google introducing a Wi-Fi router? Well Google’s continued success relies on the Internet actually functioning all the way out to that mobile device or (shudder) xBox in your son’s bedroom. xBox, if you didn’t know, is a particularly heinous networking device, especially over Wi-Fi. If Wi-Fi is the future of the Internet then Google’s future success is dependent on making Wi-Fi work better, hence the router, which I expect will become something of a reference design for other vendors to copy.
Google’s $199 OnHub, which you can order now, does a lot of things right. It supports every Wi-Fi variant, has 13 antennas, and switches seamlessly between 2.4-GHz and 5-GHz operation on constantly varying channels trying to get the best signal to the devices that need it. This is all from the wireless LAN best-practices playbook and so of course Google says that an all-Google Wi-Fi network is the best way to go. I’m guessing my ramshackle home network will require three of the things and wonder how all that shaking and baking will function on a multi-access point environment?
But those 13 antennas and the 1.4-GHz Qualcomm processor don’t inherently address the problems Wi-Fi brings to the Internet. That’s where OnHub is potentially even more radical, because it’s the first such device that’s likely to be managed by the vendor and not by you. One huge problem with Wi-Fi is the firmware in these devices is difficult to upgrade and impossible to upgrade remotely, but OnHub promises to change that with a continuous stream of tweaks to its GenToo brain straight from the Googleplex. If we forget privacy considerations for a moment this is a brilliant approach because it makes each Wi-Fi network a dynamic thing capable of being optimized beyond anything imagined to date.
I know, having looked deep into the soul of my own Wi-Fi network, that there’s the potential to increase real networking performance (measured not just by bitrate, but by a combination of bitrate and latency) by at least 10X, but to make that happen requires constant tuning and updates.
Whether Google is the best outfit to trust with that tuning and those updates is another story.