Key Criteria to Move Cloud Workloads to the Edge
communication links , and on the order of 10ms for edge node queueing and processing delays , for a grand total of about 15ms round-trip sense-compute-actuate latency ( a twenty-fold improvement ). So , instead of our 100Km / hour vehicle getting 8M closer to a dangerous collision due to system latency , this scenario allows it to only approach about 40cm before the breaks are applied . The increased safety of the lower latency edge computing-based architecture should be obvious .
Other use cases are equally latency critical . Augmented reality / virtual reality applications often compute eye views in the network to reduce the amount of computation hardware needed on the goggles . High network latency can create time lags between head motion sensing and video rendering that induce nausea in some individuals . In haptics ( e . g ., tactical feedback joysticks ) networks where the force feedback is calculated on networked computing , latency exceeding about 15ms can negatively impact the illusion of touch 8 . By moving the computation associated with these critical applications from cloud data centers to edge nodes near the endpoints they serve , the user experience can be greatly improved .
Workloads in closed-loop industrial control systems ( applications like robotics , welding , printing , etc .) can be even more latency critical , some requiring less than 1ms round-trip delay .
Workloads being considered for cloud data center execution should be carefully evaluated for their latency requirements . If the latency requirements are beyond what cloud data centers can reasonably deliver , those workloads ( or at least their latency-sensitive subcomponents ) should move to edge computing nodes .
Network Bandwidth
Network bandwidth refers to the peak or average data rates or data set sizes on the various wireless and wireline interconnect links in the IoT network architecture . A functional partitioning of IoT applications where the processing is all done in cloud data centers results in large datasets being transferred , or high streaming bandwidth between the IoT devices and cloud . This bandwidth can be quite costly , in terms of the network charges for the service use , and also in terms of its impact on other applications that share the same interconnect networks .
Consider a use case where a high bandwidth sensor needs to transport its entire data stream continuously to a cloud data center for analysis , for example , a 4K-resolution surveillance camera
8
Rank M ., Shi Z ., Müller H . J ., Hirche S . ( 2010 ) The Influence of Different Haptic Environments on Time Delay Discrimination in Force Feedback . In : Kappers A . M . L ., van Erp J . B . F ., Bergmann Tiest W . M ., van der Helm F . C . T . ( eds ) Haptics : Generating and Perceiving Tangible Sensations . EuroHaptics 2010 . Lecture Notes in Computer Science , vol . 6191 . Springer , Berlin , Heidelberg .
https :// doi . org / 10.1007 / 978-3-642-14064-8 _ 30
- 6 - June 2021