A novel Latency-Guaranteed based Resource Double Auction for market-oriented edge computing

Edge computing, taking of distributed architecture and edge-servers being close to end-devices, has been widely attended to provide extra computation resources to assist smart end-devices in completing computation tasks with low latency. Although considerable efforts on resource allocation have been developed to reduce energy consumption and computation latency in edge computing, the profits of edge-servers in market-oriented edge computing have not been investigated. In addition, few efforts considered the combination of multiple constraints (such as bandwidth, latency, etc.) in resource auctions for the edge-servers with limited communication and computation resources. To this end, in this paper we propose a novel Latency-Guaranteed based Resource Double Auction (LGRDA) scheme for the market-oriented edge computing system with resource-constrained edge-servers existed. The proposed scheme can maximize the profits of edge-servers and achieve the great efficiency of resources utilization, as well as guarantee acceptable latency for computation tasks of end-devices. Particularly, LGRDA conducts a resource-task matching model to match the limited communication and computation resources of edge-servers to computation tasks of end-devices with considering the combination of multiple constraints, such as bandwidth, latency, etc. To determine the effective resource-task pairs from all available matches with objective of achieving maximum profits for edge-servers, a multi-users double auction mechanism is proposed to determine the effective matches (also namely winning resource-task pairs) through bidding. Via analysis and simulation experiments, our data show that LGRDA can achieve individual rationality, budget balance and truthfulness in auctions. Additionally, our data also demonstrate that our LGRDA can significantly increase the system efficiency of market-oriented edge computing in terms of resource utilization rate and average profits for edge-servers, as well as the number of winning matches, in comparison with existing schemes.