Spiraling Data Costs Imperil IoT
来源： English News 发布者：English News
热度14票 时间：2016年4月15日 17:46
MADISON, Wis.—The widespread implementation of Internet of Things (IoT) devices — highlighted by Cisco’s prediction of “50 billion devices and objects connected to the Internet by 2020” — has gone from pipe dream to survival strategy for the electronics industry.
As the dream morphs toward reality, IoT system designers ponder a number of yet-to-be-sorted-out deployment challenges. These include cost [of IoT devices], limited battery life, a bandwidth-constrained network, the lack of standard protocols and a security deficit.
But what about the cost of data? IoT is all about connected devices designed to gather data for intelligent analysis.
Given that 50 billion IoT devices and their contents need to live for a long time, the cost of data — associated with the transmission, storage and mechanisms for efficiently retrieving it — can’t be ignored, Tom Hunt, president and CEO of WindSpring, told EE Times.
WindSpring (San Jose, Calif.) is a software technology licensing company with hopes to enable cost-effective solutions for IoT deployments by licensing a suite of IoT tools — such as data compression and protocol conversions — to device vendors and service providers.
Noting that IoT has so many uses, Lee Doyle, principal analyst at Doyle Research, said the amount of IoT data generated is accumulating quickly. Further, IoT devices need to operate under varying situations — including constrained network bandwidth and latency. “You need IoT compression tools like what WindSpring is proposing.”
Spiraling out of control
Many IoT device developers and service providers don’t realize until much too late in the design cycle that their data costs are “quickly spiraling out of control,” Hunt observed.
IoT devices generate frequent bursty messages. When sent via traditional cellular network such as 4G/LTE, data gets expensive, said Hunt.
Assume that you have a wearable device that sends out data in 166 bytes. Once it’s sent via IP network, the data balloons to 403 bytes (including the HTTP header in 237 bytes). The longer it takes to send a message and the chattier its protocol gets, “it costs you more,” explained Hunt. Additional effects are radio usage and battery life.
Click here for larger image
Some cellular network operators are turning to Low-Power Wide-Area Networks (LPWANs) such as Sigfox and LoRa, said Hunt. They’re looking at ways to dump IoT data traffic onto those separate networks.
But LPWANs are designed to run at extremely low data rates. Restrictions imposed by Sigfox, for example, are a 12-byte payload for uplink and 140 messages per day per device. Lora sets two-way communication at 0.3 to 50kbps data rate, with maximum payload at 256 bytes. This, in turn, requires longer air time.
IoT device vendors need “a new way of compressing and optimizing data,” said Hunt, so that devices can efficiently operate regardless of network constraints, and over any existing or new protocol that enters the market.
But wait. There’s more. Is Windspring the only company doing this? Aren’t there other vendors who are also offering “data compression”?
The Windspring CEO acknowledged that some offer compression when moving data between servers. “But our solution is capable of looking at different types of information before we compress it and turning that into a data format that makes the best fit for particular applications. Further, we convert it to destination networks. If it is 4G, we convert it into 500 bytes. If it goes to Sigfox, we make it into 12 bytes.”
Paul Teich, principal analyst at Tirias Research, agrees.
“General compression techniques are really widespread. But general lossless compression algorithms can't compress data much past three- or four-to-one ratios, maybe high single digits,” he said.
The problem with such general compression techniques is that “the algorithms have no context about the data. If a service provider knows what data structures are being sent from one point to another, then they can use that context to drive much better compression,” he explained. The catch, however, is that “they sacrifice the general purpose compression capability — that context is only good for a single system,” he added.
Taking advantage of its patented compression algorithms, Windspring has developed “Intelligent Compression APIs” that use a multi-stage approach. It compacts, making the data smaller before compression. It encodes as little data as possible and converts data to maximize optimization.
The result, according to Hunt, is compression as high as 20x and data that can be sent appropriately over any kind of network.
Tirias Research’s Teich said, “I haven't run across anyone trying to market this [type of compression tools].”
Windspring's primary competition is those embedded system engineers who are trying to develop “do-it-yourself” solutions, he noted. “I used to be an embedded systems programmer in a previous decade and I know that where bandwidth is limited — embedded designers have been doing this for a very long time.”
He said, “NASA for instance is compressing as much data as possible to save deep space probe power and cram as much data into very low bitrate channels as they possibly can. But it takes some knowledge of compression to do it well." To do it well in the IoT space, "It takes knowledge of the sensors and data structures to eke out every bit of possible compression,” he added.
Teich believes that WindSpring should get customers close to optimal compression without spending as much time on it as NASA does.”
Beyond data compression, Windspring also offers what the company calls an “Any-to-Any (A2) Protocol Connector.” Hunt said, “We can swap out protocols to make more efficient end-to-end network communications.”
Windspring says that the new tool “seamlessly receives device data in any protocol and translates it on the fly into the required network protocol, then sends it to the server or any client device in real time.”
Given multiple protocols out there, the protocol conversion is obviously necessary. So, how’s Windspring’s protocol conversion tool different from others?
Tirias Research’s Teich said, “At face value, not much different.”
But here’s the thing.
In Teich’s view, “Gateways are the new high ground in IoT competition.” He noted, “If WindSpring compression is used from end-point to gateway and then from gateway to service, then it will be more efficient to use WindSpring's protocol conversion in the middle.”
He added, “Context still matters in gateway aggregation and upstream compression. Simply passing through compressed data from end-point to service through the gateway won't optimize gateway to service bandwidth, which will be important for distant remote gateways that may be solar or battery powered — oilfields, seismic monitoring, all sorts of applications.”
According to Windspring, its A2A Protocol Connector works with dozens of protocols—including HTTP, Apple HomeKit, Nest Weave, Google Brillo, MQTT, AllJoyn, ZeroMQ (0MQ) and CoAP (Constrained Application Protocol). The company says it will continually update the tool “to work with new protocols as they are deployed.”
What happened in India
Today, Windspring has “multiple” undisclosed customers. Without naming names, Hunt shared a story about one of its Indian customers. The customer, described as “a leading global enterprise customer” was trying to deploy 3G-connected smart meters.
Given that a typical smart meter generates as much as 400MB of data per day, “moving that data through wireless networks is incredibly costly,” he explained.
In this client’s case, “these meters generated sixty 40-byte messages per minute. With wireless costs at $15 per MB, that approach simply wasn’t cost effective.” The customer worked with Windspring, integrated Windspring’s compression API on meter and server. The solution used “less than 300bytes of memory and 2K of storage, resulting in data reduction by a factor of 10,” according to Hunt.
Working in a resource-constrained environment
Windspring, a 10-year old company, has grown up in the embedded market.
The genesis of the company’s technology is “a unique brand of compression for embedded systems with limited CPU power and memory size,” explained Hunt.
In particular, Windspring began by developing compression technologies for automotive navigation systems. Maps were delivered via CD, DVD or HDD. It was not only necessary to compress the huge mapping data to fit the storage medium, but it was also important that the embedded system in a car does not run out of memory while booting up a navigation application, Hunt explained.
In particular, the challenge for the system was to be able to randomly access compressed data and decompress only the data it needs, he added.
Windspring’s customers for the navigation systems read like a Who’s Who in the automotive market, including Pioneer, Toyota, Denso, BMW, Honda, Alpine and Tom Tom.
Hunt said, “We developed six patents on compression techniques for the navigation market. In designing our new IoT tools, we’ve used our streaming patent — the technology to access the data you need from multiple sources and reassemble it in a single stream.”
Instead of approaching data not as a file, but as a block, which Windspring did for the automotive navigation system, the company is now handling data on a byte or bit level for its new IoT tools, said Hunt.