An official website of the United States government
Mobile devices such as smartphones and autonomous vehicles increasingly rely on deep neural networks (DNNs) to execute complex inference tasks such as image classification and speech recognition, among others. However, continuously executing the entire DNN on mobile devices can quickly deplete their battery. Although task offloading to cloud/edge servers may decrease the mobile device’s computational burden, erratic patterns in channel quality, network, and edge server load can lead to a significant delay in task execution. Recently, approaches based on split computing (SC) have been proposed, where the DNN is split into a head and a tail model, executed respectively on the mobile device and on the edge server. Ultimately, this may reduce bandwidth usage as well as energy consumption. Another approach, called early exiting (EE), trains models to embed multiple “exits” earlier in the architecture, each providing increasingly higher target accuracy. Therefore, the tradeoff between accuracy and delay can be tuned according to the current conditions or application demands. In this article, we provide a comprehensive survey of the state of the art in SC and EE strategies by presenting a comparison of the most relevant approaches. We conclude the article by providing a set of compelling research challenges.
more »
« less
An edge-based architecture to support the execution of ambience intelligence tasks using the IoP paradigm
In an IoP environment, edge computing has been proposed to address the problems of resource limitations of edge devices such as smartphones as well as the high-latency, user privacy exposure and network bottleneck that the cloud computing platform solutions incur. This paper presents a context management framework comprised of sensors, mobile devices such as smartphones and an edge server to enable high performance, context-aware computing at the edge. Key features of this architecture include energy-efficient discovery of available sensors and edge services for the client, an automated mechanism for task planning and execution on the edge server, and a dynamic environment where new sensors and services may be added to the framework. A prototype of this architecture has been implemented, and an experimental evaluation using two computer vision tasks as example services is presented. Performance measurement shows that the execution of the example tasks performs quite well and the proposed framework is well suited for an edge-computing environment.
more »
« less
- Award ID(s):
- 1816379
- PAR ID:
- 10310801
- Date Published:
- Journal Name:
- Future generation computer systems
- Volume:
- 114
- ISSN:
- 0167-739X
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
With distributed communication, computation, and storage resources close to end users, edge computing has great potentials to support delay-sensitive industrial applications involving intelligent edge devices. Cognitive portable ground penetrating radars (GPRs) are expected to achieve high-quality sensing performance in a variety of industrial environments by operating intelligently and adaptively under varying sensing conditions. Although edge computing makes it very promising to develop cognitive portable GPRs, both strict performance requirement and trade-offs between communication and computation pose significant challenges. This paper presents an edge computing framework for cognitive portable GPRs. Specifically, the system architecture of an EC-enabled cognitive portable GPR is developed. Based on the identification of various involved computation tasks, an offloading policy was proposed to determine whether computation tasks should be executed locally or offloaded to the edge server. Experimental results show the efficacy of the proposed methods. The framework also provides insight into the design of cognitive Internet of things (IoT) supported by edge computing.more » « less
-
In the era of pervasive digital connectivity, intelligent surveillance systems (ISS) have become essential tools for ensuring public safety, protecting critical infrastructure, and deterring security threats in various environments. The current state of these systems heavily relies on the computational capabilities of mobile devices for tasks such as real-time video analysis, object detection, and tracking. However, the limited processing power and energy constraints of these devices hinder their ability to perform these tasks efficiently and effectively. The dynamic nature of the surveillance environment also adds complexity to the task-offloading process. To address this issue, mobile edge computing (MEC) comes into play by offering edge servers with higher computational capabilities and proximity to mobile devices. It enables ISS by offloading computationally intensive tasks from resource-constrained mobile devices to nearby MEC servers. Therefore, in this paper, we propose and implement an energy-efficient and cost-effective task-offloading framework in the MEC environment. The amalgamation of binary and partial task-offloading strategies is used to achieve a cost-effective and energy-efficient system. We also compare the proposed framework in MEC with mobile cloud computing (MCC) environments. The proposed framework addresses the challenge of achieving energy-efficient and cost-effective solutions in the context of MEC for ISS. The iFogSim simulator is used for implementation and simulation purposes. The simulation results show that the proposed framework reduces latency, cost, execution time, network usage, and energy consumption.more » « less
-
In edge computing deployments, where devices may be in close proximity to each other, these devices may offload similar computational tasks (i.e., tasks with similar input data for the same edge computing service or for services of the same nature). This results in the execution of duplicate (redundant) computation, which may become a pressing issue for future edge computing environments, since such deployments are envisioned to consist of small-scale data-centers at the edge. To tackle this issue, in this paper, we highlight the importance of paradigms for the deduplication and reuse of computation at the network edge. Such paradigms have the potential to significantly reduce the completion times for offloaded tasks, accommodating more users, devices, and tasks with the same volume of deployed edge computing resources, however, they come with their own technical challenges. Finally, we present a multi-layer architecture to enable computation deduplication and reuse at the network edge and discuss open challenges and future research directions.more » « less
-
In edge computing use cases (e.g., smart cities), where several users and devices may be in close proximity to each other, computational tasks with similar input data for the same services (e.g., image or video annotation) may be offloaded to the edge. The execution of such tasks often yields the same results (output) and thus duplicate (redundant) computation. Based on this observation, prior work has advocated for "computation reuse", a paradigm where the results of previously executed tasks are stored at the edge and are reused to satisfy incoming tasks with similar input data, instead of executing these incoming tasks from scratch. However, realizing computation reuse in practical edge computing deployments, where services may be offered by multiple (distributed) edge nodes (servers) for scalability and fault tolerance, is still largely unexplored. To tackle this challenge, in this paper, we present Reservoir, a framework to enable pervasive computation reuse at the edge, while imposing marginal overheads on user devices and the operation of the edge network infrastructure. Reservoir takes advantage of Locality Sensitive Hashing (LSH) and runs on top of Named-Data Networking (NDN), extending the NDN architecture for the realization of the computation reuse semantics in the network. Our evaluation demonstrated that Reservoir can reuse computation with up to an almost perfect accuracy, achieving 4.25-21.34x lower task completion times compared to cases without computation reuse.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1016/j.future.2020.08.001
