This is part eleven of a series. Read part ten here.
When thinking about the IoT systems, I conclude that each of the three systems – the cloud, the aggregator, or the sensor – will have a different class of components. For example, the components in the smart sensors will be designed for ultra-low power, accepting the resulting performance levels. The components for the aggregator will be higher performance, but within a power budget. Finally, the components in the cloud will be designed for maximum performance with less emphasis on power dissipation or cost.
Here is a chart I put together many years ago to try to put a definition to the various performance concepts:
To put the chart into words:
- High performance devices are those in which performance is the primary priority, if not the only priority. This is a description of a cloud computer.
- Power efficient devices are those which, given no performance is sacrificed, will prioritize reducing the power dissipation. This aligns with an aggregator system.
- Ultra-low power devices are those in which, given that absolute minimal power dissipation is achieved, will prioritize the performance. This describes the smart sensor.
To take it a step further, Figure 1 gives some perspective of the power requirements for various IoT applications.
[Figure 1 | Example power requirements for various IoT applications.]
You can see a correlation between the three concepts from the above chart and the three systems that make up the IoT system. If you haven’t guessed, the cloud is populated with the high performance devices.
If I take this concept of performance as the primary, if not the only, priority, for cloud computing, the class of devices in the cloud are high performance multi-core processing systems, GPUs, FPGAs, and specifically designed custom devices. In all of these classes of devices, their goal is to maximize performance, maximize communications bandwidth, or guarantee security and privacy.
To complicate things a bit, many aggregators will take on the concept of high performance and will be hard to distinguish from being part of the cloud. This is appropriate, as it can easily be argued that every system within the IoT system has a cloud to connect to.
You might ask at this point why I seem to be focusing this discussion on microprocessors rather than memories or communications systems. One reason is that my background in technology has been in the world of digital signal processors (DSPs) and I didn’t worry about the world of memory, wireless communications, or security unless a DSP was involved. That means there are some stories you can tell that will amaze me and the other readers of this blog. So tell me a story.
How for a few questions for you:
- When you think of cloud computing, which high performance processors do you think about?
- Which is the most important aspect of cloud computing: performance, bandwidth, security, or a different aspect?
- If you have designed a cloud computing element, what was your choice of the processing element? Why did you chose it? (Yes, this is a different question than question one above.)