IIC Journal of Innovation | Page 34

A Horizontal Taxonomy for the Industrial IoT variation or “jitter”. Even a fast server with low average latency can experience large jitter under load. In a distributed system, the most important architectural impact is the potential jitter imposed by a server or broker in the data path. An architecture that can satisfy a human user Figure 4: Added Server Latency annoyed by a wait longer Although the hardware transmit time is often negligible, sending data than 8 seconds for a through a server “hop” requires traversing the sending machine’s transmit website will never satisfy stack, the server’s receive stack, the server’s processing queue, the server’s an industrial control that transmit stack, and finally the destination’s receive stack. Each of these has threads, queues, and buffers that add uncontrolled latency. Worse, the server must respond in 2ms. cannot easily prioritize traffic as easily as the endpoints. Thus, systems that We find the “knee in the are sensitive to maximum latency often cannot use data servers. curve” that greatly impacts design occurs when the speed of response is measured in a few tens of milliseconds (ms) or even microseconds (µs). We choose 100ms, simply because that is about the unpredictable delay of today’s servers. Systems that much respond faster than this usually must be peer-to-peer, and that is a huge architectural impact. Figure 5: IIoT Real-Time Applications To provide quality feel to surgeons, medical robotics distributed control loops must run at rates up to 3 kHz and control the “jitter” to only tens of microseconds. Similarly, autonomous cars must react to fast enough to safely control the vehicle and prevent collisions. These fundamental performance needs imply a system architecture that does not send data through intermediaries. IIC Journal of Innovation - 33 -