ktsdesign - Fotolia
Around the middle of the 2000s, car makers had a problem. In-dash infotainment systems were becoming a popular...
option on many models, but the systems lacked much-needed compute power and storage, making the systems slow and clunky -- a situation customers would not tolerate for long. Adding more power and storage was an option, but that would drive up costs. Instead, car makers turned to WindSpring Inc. and its unique data compression technology.
Unlike other data compression technologies (then or now), WindSpring's tech has a number of advantages over standard file- or image-compression algorithms. First of all, it is selective, only decompressing data that needs to be decompressed to perform an operation of some sort, such as updating a navigation system. Automakers liked this because it makes their systems' speed competitive with smartphones without the added costs of more compute and storage.
But that is only one use case. Even though WindSpring's data compression technology is rolling around in millions of cars today, its killer app may well be in IoT devices. Billions are expected to make their way online in the coming years and almost all of them will be battery powered -- and this is IoT's Achilles' heel.
IoT: It's all about power
To be useful and ubiquitous, IoT devices and sensors have to conserve battery power at all costs. This means most devices will spend most of the time "asleep" and have very limited onboard compute or storage. Because there is so little storage, these devices cycle on and off constantly to phone home and dump data. Waking up takes power. So does sending data -- even over emerging low-power local area networks like Bluetooth Low Energy, Sigfox or ZigBee, and then, in many cases, to carriers' 2G (yes, they still exist and are used mostly for machine-to-machine connections like vending machines phoning home), 3G and 4G networks. To make everything even more complex for manufacturers and operators, each of these networks has different bandwidth restrictions and power demands.
If the amount of data can be significantly reduced, device manufacturers can design their products from the start to be more efficient. This not only makes their devices more useful and long lived, but also allows device operators like farmers, facility managers or utilities running smart meters to cut down on the costs of sending data over networks. These costs can be significant. According to Tom Hunt, who was WindSpring's CEO as this article was prepared, this can save IoT device operators a lot of money given that data transfer cost can be as high as $15 per Mbps.
20:1 data compression technology
WindSpring claims its SpringBoard Intelligent Compression APIs reduce data by a factor of 20:1 by using a multistage approach that compacts the data, compresses it and finally converts it. In real-world terms, that equates to a 98% reduction. Its nearest competitor, CAST Inc., claims a compression ratio of just 3:1. But the use cases are different, said Paul Teich, principal analyst with Tirias Research. This is because CAST's offering seems to be aimed at IoT devices that may interact or be acted upon, like an actuator turning a value off or on, and controlled via a web interface.
"I think we're over the first wave of IoT architecture where every device is supposed to have its own webpage and/or communicate directly to a cloud service," Teich said. "That's good for configuring my home router, but do agricultural IoT sensors need webpages? Each one of them? No, that's just not appropriate architecture or design."
Likewise, older Ethernet compression protocols like HTTP, SSH and PKZIP don't work well on the small amounts of data IoT sensors send. These chatty protocols were designed for much larger file sizes and, therefore, introduce a lot of packet overhead into a data stream. The end result is they actually make small data much larger.
"Going just from two times to four times compression is very, very difficult," Teich said. "So, you have to understand the data you are working with and have a context for it."
Lossy vs. Lossless
Another factor to consider in IoT data compression is lossless versus lossy algorithms. Because there's not much of it, sensor data needs to be compressed in a lossless fashion. According to the Institute of Electrical and Electronics Engineers (IEEE), "The compression performance to realize significant gain in processing high-volume sensor data cannot be attained by conventional lossy compression methods."
Kevin CurranIEEE senior member
If you compress an image and a few pixels get literally lost in translation, it's not a big deal. But if a temperature sensor in a refinery sends incomplete data, there could be problems.
Another benefit of WindSpring's IoT data compression is encryption. SpringBoard compresses and encrypts data at rest in the same pass. Therefore, if a device is hacked or stolen, the data portion, at least, is safe from tampering. For a humidity sensor in the field, this may not be a big deal, but as IoT devices become more common and interwoven into everyday life -- think smartwatches, fitness trackers, lightbulbs and pacemakers --this could matter more and more.
"Compression algorithm design can be tricky due to the small code, data memories and the energy constraints," said Kevin Curran, IEEE senior member and cybersecurity professor at Ulster University. "This is because the energy constraint limits everything from node size, data sensing rates and link bandwidth to actual weight. Often, the radio is the main energy consumer in a constrained IoT device and can be responsible for anywhere from 40 to 70% of the overall energy budget. Therefore, given the high amount of energy spent on communications, it is a valid aim to reduce radio energy and the most sensible technique is to concentrate on efficient data compression algorithms."
This may be why, according to Teich, there is no one else doing what WindSpring is doing -- at least, not publically. This means the company pretty much has the IoT data compression technology market to itself. And it has a few other things working in its favor: patents, substantial first-mover advantage and a licensing model that could make its tech a de facto standard sooner than later.
"It took [Hunt] years to build this experience base around this technology, so what he is selling essentially is a quick start," Teich said. "I don't know of anyone else who is licensing it."
Beyond IoT: Data compression for MVNOs
Beyond IoT device makers and operators is the lesser-known mobile virtual network operator (MVNO) market. These are resellers of network bandwidth bought in bulk from major network backbone carriers like AT&T and Verizon. Players in this space include Consumer Cellular, Boost Mobile and Cricket Wireless -- there are hundreds of these companies worldwide.
WindSpring's data compression technology enables these carriers to make better use of their leased bandwidth by selling, say, 1 Mb of bandwidth to a customer but, through compression, only using 20% or less of that amount in actual transmission. The customer still gets 1 Mb of service, but the MVNO can reduce that megabit by 80% before sending it, allowing the MVNO to reduce end-user costs and carry more customers on the same amount of leased bandwidth.
Toppling the Tower of Babel
Finally, WindSpring offers a network protocol adaptor, Any2Any Protocol Connector, which allows any network protocol to talk to any other network protocol, enabling manufacturers and operators to interact with their data regardless of its source. For example, this allows IoT communication protocols like ZigBee, Thread or Apple's HomeKit to send data to the same cloud-based service provider of home automation services. Or a farmer can have many different devices from multiple manufacturers all sending data via different networks to the same cloud server for analysis.
The data center of the future needs data compression to handle IoT