Hetero-Edge: Orchestration of Real-time Vision Applications on Heterogeneous Edge Clouds

MPCSM: Microservice Placement for Edge-Cloud Collaborative Smart Manufacturing

“…2) Placement of One Multi-hop Service: These studies discuss the modularization and placement of a single service at the microservice level. For example, Zhang et al [14] proposed an orchestration framework that broke down an edge application into multiple Storm tasks with a directed acyclic graph (DAG) representation. Such tasks were then mapped to heterogeneous edge servers for efficient execution.…”

“…2) Mathematical Model based Methods: These studies construct mathematical models to calculate the transmission, processing, and waiting time depending on various metrics collected from services and the environment [4], [12]- [14], [18]- [23]. For example, for the energy-delay optimization of multiple collaborative edge intelligence applications, Zhang et al [21] intuitively calculated the transmission time through dividing the transmission load d by the link bandwidth b, the processing time through dividing the service computation requirement c by the device capacity v, and the waiting time through the M/M/N queueing model.…”

“…Generally, the service placement issue is formalized as a latency-minimal combinatorial optimization problem [10], where a specific placement strategy (i.e., the mapping between each service and the device where the service should be placed) has to be determined under heterogeneous constraints in terms of device resource, service requirements, etc. For such an optimization problem, conventional algorithms (e.g., programming [17] and searching [14]) can be applied to achieve the optimal or suboptimal solution, once the parameters (e.g., the end-to-end (E2E) latency of different services under different placement strategies, will be explicitly discussed later) are determined. Since it is quite difficult to profile each service's E2E latency 1 under numerous viable placements in real-world collaborative systems, existing approaches [4], [12]- [14], [18]- [23] usually use mechanism-driven solutions that construct different mathematical models to estimate the E2E service latency.…”

“…For such an optimization problem, conventional algorithms (e.g., programming [17] and searching [14]) can be applied to achieve the optimal or suboptimal solution, once the parameters (e.g., the end-to-end (E2E) latency of different services under different placement strategies, will be explicitly discussed later) are determined. Since it is quite difficult to profile each service's E2E latency 1 under numerous viable placements in real-world collaborative systems, existing approaches [4], [12]- [14], [18]- [23] usually use mechanism-driven solutions that construct different mathematical models to estimate the E2E service latency. However, as the development of smart manufacturing services (e.g., multi-hop services composed of several microservices with inherently complex dependencies [16]), such idealized estimations cannot accurately depict the non-linear E2E latency of services based on edge-cloud collaborations.…”

Integrated security of cyber-physical vehicle networked systems in the age of 5G

“…如果在数据的存储过程中引入第三方审查机构, 审查 机构在审查数据完整性、正确性的过程中, 不能限制合法用户对数据的操作, 也不能为数据安全带来 新的威胁 [112] . [113,114] . 基于异常事件检测结果…”