Category Archives: Uncategorized

The Problem of Power

Listening to MIT’s Sanjay Sarma present a few weeks ago led me to think a bit about how we’re going to power all these connected devices in the Internet of Things. Mr. Sarma’s primary point is that RFID didn’t become ubiquitous until the lifetime cost of tags was reduced to pennies. Had the industry stuck with the conventional thinking around battery-powered devices with more powerful processors, it is unlikely RFID would have become as far-reaching.

The cost of batteries is dropping, and the lifespan we’re able to stretch out of them is increasing, but I don’t foresee lifetime costs (which combines the cost of the battery and the cost of replacing it when it fails) dropping to the under 10 cents range that Mr. Sarma was targeting. Even the best lithium battery technology can only be reliably stretched to a few years under constant load (even under low power) – and these batteries are still relatively expensive.

Some of the newer wide band communication protocols like 6LoWPAN and mesh networking like that proposed by Iotera can certainly help by reducing the energy requirements needed for long range communication to well below what is required by things like WIFI, Bluetooth, or Zigbee. But the reduced power needs don’t change the equation enough to meet Mr. Sarma’s goal for lifetime costs.

Energy Harvesting offers promise as a battery booster. Ambient energy can be harvested from heat, light, motion, pressure, chemical reaction, or other sources. These military boots that harvest energy from marching soldiers are a good example, but this technology is in its infancy and still relatively expensive, well outside of Mr. Sarma’s range.

But what Mr. Sarma proposed as a solution is actually quite interesting. He went back to his roots, and asked us to consider passive tags. Passive tags have no inherent power source. They are powered by a signal coming from a nearby “reader”. When the reader passes by, the tag is powered up and sends a response to the reader. The data transmitted from the tag can be anything it is able to sense.

At first blush, you may think (as I did), “didn’t we already discount Near Field Communication in IoT?”

Well on second thought, perhaps we really didn’t, and perhaps it does deserve some more consideration. If we are to get to 50 billion connected things in the next 5 years, we’ll need to expand beyond the consumer market. Businesses and governments will need to deploy lots of stuff, in all likelihood mostly sensors, and they’ll need to do it fast.

Some quick math, if you believe the 50 billion number, and then you assume 25% of those will be consumer devices (almost two for every human on earth), that leaves 38 billion for businesses. If you assume that 80% of those will be deployed by the Global 3000, that would mean each of these companies would be on average deploying almost 10 million connected things. At $10 each in five year cost, that would be $100 million. That seems to me to be on the high side (of course my assumptions are based on little science, but you get the point).

Passive tag technology reduces this number to $1,000,000 (assuming 10 cents per tag). And now it seems quite reasonable all of a sudden… in fact 50 billion devices now seems kind of low.

So could passive tags be the answer? Mr. Sarma raised the example of passive tags being used to detect termites, using one regular antenna on the tag and one antenna fashioned from wood. When only a single signal is detected from the sensor, you can deduce that the wooden antenna has been eaten and therefore you have a termite. Since the tags are so cheap, you can afford to put them everywhere, which enables you to average out the anomalies, and in the case of termites, pinpoint exactly where they are likely hiding out. Passive tag sensors now exist for everything from heat sensors, to chemical sensors, to moisture sensors.

I do believe that Mr. Sarma is on to something – at least for a specific class of sensor. It makes sense to me that if there is a cheaper sensor, that is what businesses will use. That said, there are many IoT scenarios that need much more than what passive sensors can provide, so there will likely be a mix of passive tags and more advanced applications with MCUs and batteries (and likely as much harvesting as is economically sensible).

So I take away two things from this: 1) Don’t discount the power of the passive tag. I believe it will play a role in IoT. 2) If passive tags do begin to dominate the sensor space, 50 billion is likely way too low of an estimate for the number of connected things.

Advertisements
Tagged , , ,

Merry Christmas!

Merry Christmas, everyone! I hope you all take this opportunity to spend time with friends and family and remember the true meaning of Christmas.

While you are taking some much needed time off, you might try having some fun with APIs. Luckily ProgrammableWeb found a Santa Speak option on the Open MapQuest API. And in the spirit of the holidays, Orange API is offering a holiday special for credits. And my personal favorite,  snowify.me – which uses the Google Maps API to let you create a white Christmas even if the weatherman isn’t delivering one to your home.

 

That should keep you busy for awhile – Have a Merry Christmas!!

The lowest of latencies

I spent some time at SIFMA last week talking to customers about their low latency messaging requirements. For those of you who don’t know, low latency messaging is a specialized niche of the messaging space designed for requirements where latency is measured in the nanoseconds, and systems require up to 100 million messages per second. There aren’t many applications that require this kind of speed and throughput, but in Financial Markets, things like high frequency algorithmic trading and exchange execution require that kind of horsepower.

Generally speaking, the low latency messaging space looks very different from the normal messaging space. Players like Informatica (which is not in the messaging middleware space at all otherwise), and Tibco thrive here, along with IBM with its WebSphere MQ Low Latency Messaging (LLM) product. Beyond that, there are a bunch of smaller niche players. Tibco has long been a player in low latency with their Rendezvous product line, though they had fallen way behind competitors in terms of speed until their recent release of FTL. The new release puts Tibco in the same ballpark as IBM and Informatica in terms of pure speed, with recent STAC results showing IBM and TIBCO in the lead and Informatica trailing slightly behind.

However, it isn’t just speed that matters. After all, this is messaging… Reliable delivery, guaranteed message ordering, high availability, and durable subscriptions are all important – and you don’t want to sacrifice performance to get those. The vendors vary widely on their support for these without performance penalties (something that doesn’t even show up on benchmarks).

In addition, skills are a serious concern. This is a very small niche, so finding people skilled enough to squeeze the optimal performance out of the products is always a challenge. IBM has addressed this by leveraging the WebSphere MQ base, which dominates the messaging space with over 10,000 customers and around 80% of the market. IBM has also provided development productivity accelerators like built in monitoring and congestion management, and a versatile API which allows applications to be coded once and reused across Infiniband, 10GE, Shared Memory or a mix.

Tibco likewise has cultivated a strong skill base with 1,000 Rendezvous customers. However, with FTL they are introducing an entirely new technology with a separate API, so existing skills will need to be retooled.

That said, the strategies of IBM and Tibco are likely to be the most successful, since they can leverage large install bases and complementary messaging middleware portfolios to deliver more value than other niche vendors can provide.