Key Criteria to Move Cloud Workloads to the Edge
IoT services . Moving critical portions of the computational workloads and storage to edge computing nodes can eliminate some of these vulnerabilities ( perhaps at the expense of introducing a few new ones ). Incidents of security compromise can often be easier to detect , locate , isolate , repair and restore on edge nodes than cloud data centers .
Reliability is a key aspect of trustworthiness , especially for mission-critical or life-critical workloads . The cloud is often difficult to monitor , manage , and recover after faults . Because of the huge scale of efficient cloud data centers , a site-wide outage can impact millions of application instances . Duplication and redundancy are often difficult in these high-scale cloud networks , because of the huge amount of inter-site traffic necessary to ensure that the contexts of redundant elements are geographically distributed across diverse data centers and keep the cloud-distributed databases consistently updated .
Cloud data centers are also susceptible to outages in the data links that interconnect them with the IoT devices and with their peer data centers . Edge techniques can add another dimension to reliability by distributing computation workloads and storage instances across multiple edge nodes that are still relatively physically close to each other and to the IoT devices . Edge nodes themselves can be designed to be fault tolerant , with duplicated processing , storage and I / O modules insulating system availability from single point failures . That highly reliable hardwarebased fault tolerance architecture is impossible to achieve in the cloud using commodity servers .
Resilience is the property of systems to continue operation within spec even in the presence of abnormal conditions . Some cloud-based architectures are pretty brittle , that is a single , relatively small failure , overload or overflow can have a large impact on the operation of the system . A single power outage , fiber cable cut , primary router failure , or natural disaster can destroy a data center ’ s ability to process computational loads .
Modern data center architectures do add some redundancy to the power and data networking infrastructure that supports their servers , but multiple data centers spread out by considerable distances as multiple availability zones 12 are required to provide adequate resiliency for many critical IoT applications . Edge techniques enhance resilience by providing multiple edge nodes , any one of which is capable of providing full service . Edge nodes are often arranged in multiple layers , and computational loads can be moved to an adjacent layer via the north-south links if a node on one layer fails or becomes overloaded . Further , the east-west links that interconnect edge nodes on a given layer can move data between peer edge nodes , providing resilience in the face of single node failures or localized overloads .
Privacy is the final aspect of trustworthiness . There are certain concerns with privacy unique to the cloud , especially if the cloud service is hosted by a web-scale company with significant financial stake in understanding the patterns of your data . On a public data center that is shared
12
- 10 - June 2021