January 15, 2015
By John Daane
As we look toward 2015, many of the dynamics that have supported the programmable logic industry in recent years remain in place. ASIC and ASSP design costs spiral, uncertainties about functionality grow, and high-volume opportunities are rare. New dynamics and the evolving Internet of Things (IoT), however, are creating an uncertain environment with unintended consequences. As the industry struggles to make the IoT work, two trends will emerge that play directly into the strengths of programmable logic, creating great opportunities for FPGAs.
The IoT Idea
The concept behind the IoT is simple: if a device is controlled by electronics, you can simply connect the controller to the Internet. With its own IP address, the device has new powers—it can report status, send data, or accept commands with a smart-phone or cloud application.
The Internet connection also triggers a process of migration. Designers can use cloud software to add functionality to the device, even using big-data analysis to enrich the device’s performance. Imagine a portable ultrasound device, for example, that adjusts its imager and guides its user based on heavyweight image processing and image comparisons against a massive database of previous images, diagnoses, and outcomes from around the world.
Designers will also use this connection to pull intelligence out of the device and into the cloud, making the device less expensive and more portable, and allowing easy functionality upgrades. The end result is a device composed of sensors and actuators with network interfaces, all connecting across the Internet to control software.
We are already seeing this process at work. Smart-phone services like Apple’s Siri execute in the cloud, not on the handset. In the networking space, network functions virtualization (NFV) replaces entire boxes of specialized networking hardware with software running on servers in data centers. The future of the IoT appears to be a universe of tiny, inexpensive sensors and actuators connected through the Internet to giant data centers.
This view, however, is too simplistic. Trends in data acceleration and network enrichment—both favorable to FPGAs—are addressing needs for faster and predictable real-time performance, low-power WiFi and Ethernet connections, and rapidly evolving device security.
Data Center Acceleration
In the data center, tasks historically performed in hardware are being transmuted into software running on servers. Sometimes this conversion is easy. Sometimes it only works when you break the task into many threads, and spread the threads over many servers. Sometimes, however, even multithreading doesn’t make the software fast enough, and hardware acceleration is needed to meet the real-time requirements of the task.
Developments from web giants Microsoft Bing and Baidu—supported by Altera—show that with their flexibility and massive parallelism, FPGAs can improve throughput, response time, and energy efficiency working intimately with the server CPUs. A key part of this trend is to make the FPGA programmable—and debuggable—in languages already familiar to data-center programmers. Altera’s SDK for OpenCL™ tool, for example, allows programmers to create and debug a complete CPU-interface-FPGA platform from a high-level language.
Enriching the Network
The other problem is that today’s Internet is not appropriate for the IoT, for three major reasons. First, the WiFi and Ethernet connections at the edge of the Internet require power-hungry, always-on interfaces, the opposite of what all those tiny sensors and actuators—the Things—can provide. So a myriad of short-range, low-power networks is springing up to connect the Things to hubs, which in turn connect to WiFi or Ethernet.
Second, some devices have tasks—like controlling a motor—that must have guaranteed maximum delay. If you move the controller to the other end of the Internet, how do you guarantee that packets will arrive at each end on time? Eventually the answer will be new network functionality, such as the time-sensitive networks being developed. But for now, these control loops will require real-time computing in the IoT hubs.
Third, hackers will certainly attack the IoT—remember the Stuxnet worm of 2010. As the IoT emerges, responsibility for authentication, encryption, and functional safety will fall upon the hubs. The hubs will have to support a plethora of ultra-low-power network protocols, a wide variety of application-dependent, real-time computing loads, and rapidly evolving security and safety tasks. This varied and uncertain environment would be a nightmare for ASICs or standard-product SoCs, but it is the natural environment for fast, computationally efficient FPGAs.
Looking Ahead to 2015
The pendulum swings between investment in shared resources—networks and data centers—and investment in terminal devices. Today we see momentum shifting from the smart phone explosion back toward investment in 4G build-out, creation of the IoT peripheral networks and their smart hubs, and evolution of data centers. As the industry faces these emerging trends, programmable logic will surely be at the forefront of leading these developments. 2015 will be an exciting year!