Edge Computing That Utilizes In-Network CPUs to Achieve High Capacity and Interruption Tolerance With Fewer Edge Servers

We have proposed AMec (Access-Metro edge computing), an edge computing system that utilizes surplus computing resources in access metro networks. Although the use of CPUs on network devices was expected to reduce the number of edge servers, there were concerns about frequent interruption and pod's relocation due to executing their main job such as information gathering and routing reconfiguration, unlike MEC (Multi-access Edge Computing), which only uses dedicated edge servers. To avoid the pod's relocation, in this paper, we propose an allocation method that calculates the interruption occurrence rate when each pod of the application is placed in each node by using the usage duration characteristics of the application and allocates the pods in a way that the value is below a threshold. Under conditions where a certain level of interruption can be acceptable, it is confirmed that AMec with the proposed method can reduce the number of edge servers by 62.5 % ∼ 66.7 % compared to MEC.