Thread: Is It Going To Fly?

Last week Google / Nest unveiled some specs of the Thread IoT protocol announced back in July. My take on this: I am disappointed.

First, both Tony Fadell and Chris Boross made several good points about the need for a 2nd network ant home (assuming WiFi is the 1st one).

"Wi-Fi is not perfect for all devices in the home and it's not perfect for all use cases," said Boross in front of an audience. "We feel there should be a second network in the home."

They were both pitching multi-year lifespan of battery powered devices. We all know this cannot be done with WiFi today. So they propose Thread, which really is a (yet another) 6LoWPAN stack running on top of a single - channel 802.15.4 ZigBee at 250kbps. Being deeply involved in Bluetooth Smart for more than a year now, I opened my eyes wide... WHAT? They are going to push 100+ byte packets over a 250kbps network? This is crazy. Or may be is enough for a couple of thermostats but not for a smart home!

They admitted the Thread network saturates itself at about 250 nodes. Well... this thing was supposed to be for the future. Gartner says we will have 500 devices on average few years from now (meaning - for instance - I will have 2500 at my home. I had 250 back in 2008).

Then they need a border gateway to use a mobile device to connect and configure any Thread appliance. "How do you explain to a customer buying a ceiling fan, they need a gateway?" - somebody asked.

802.15.4 being a single channel solution, working today in the extremely crowded frequency band falls short when it comes to resilience to radio interference. Bluetooth Smart for example hops around on 40 channels, dynamically skipping the noisy ones. ZigBee just fails when a channel is not available.

Energy efficiency? Give me a break. To transmit 100 bytes (a typical 6LoWPAN frame) over a 250kbps raw data rate link (Thread) it takes about 20x more time compared to transmitting 20 bytes over a 1Mbps raw data rate link (Bluetooth Smart). Based on the transmitter activity alone, it is safe to assume Bluetooth Smart is an order of magnitude more energy efficient... Of course a temperature sensor that sends one packet every 5 minutes will last several years. But what if you have an occupancy sensor that sends data every second? Can you do 5 years at 1Hz on a battery with Thread?

Thread promises 100ms latency. This is again fine for HVAC, but way too much for interactive applications like a touch dimmer / color picker plate controlling lights. When you are sliding your finger on a color wheel light scene picker, you want to stream a new value to the LED lamps every 10-20ms to achieve a smooth, responsive effect. Thread just can't handle such load.

Then there is no indoor location / proximity feature at all. Phones are separated from Thread devices by a gateway. So they can't be aware of  their physical whereabouts relative to Thread beacons (even if they existed). Thread keyfobs? Who will be carrying them when all wearables have already gone the Bluetooth Smart path...

My impression is the Thread protocol is very much thermostat and smoke detector - oriented. But when it comes to interactive lighting and having a plethora of sensors spread around the house, it is just not capable of delivering the experience expected today. Not to mention tomorrow, when appetite grows. It all looks like they took what was the mainstream radio technology back in 2010 when Nest was founded (read: 802.15.4) and decided to build on top of that. Also the fact Silicon Labs (who acquired Ember, the #1 ZigBee player) is a founding member of the Thread Group may explain why Thread is based on the 802.15.4 technology. But the world moves on and now we have much better options available, namely Bluetooth Smart.

We will see how this initiative plays out. It seems the underlying radio standard is too old and too heavy to reach the market objectives and customers' expectations. But not everything is lost. Silicon Labs has also recently acquired Energy Micro, together with Geir Førre, who was the co-founder of ChipCon, a company acquired by Texas Instruments back in 2006 (ever wondered why the most popular Bluetooth Smart silicon is called "CC2541" ?). And there is no fundamental reason why Thread as an upper layer protocol couldn't run on top of Bluetooth Smart - since the 4.1 release Bluetooth Smart supports connection oriented channels capable of carrying IPv6 traffic. So may be there will be a marriage?

We are adding Thread to the list of protocols supported by our software, but at the same time, in the interest of the best performing products and the happiest user experiences, I am so glad that we, at Seed, have put our smart home bet on Bluetooth Smart. For all who want a live experience : we will have some super cool Bluetooth Smart demos at the dreamforce 14 show next week in San Francisco. You're invited!