چكيده لاتين
In edge computing, repetitive computations are common. The concept of "computation reuse" has been proposed as an approach to enhance the efficiency of edge computing. This approach involves reusing the results of previous computations to prevent them from being executed again. This practice increases efficiency, reduces redundant computations, and improves response times to requests. In TCP/IP architecture, there is no mechanism for detecting repetitive computations, making the implementation of the "computation reuse" approach practically unfeasible. Therefore, utilizing ICN-based edge computing can significantly improve this issue. By using ICN, it becomes possible to identify repetitive computations and computation reuse results. This approach helps improve the efficiency and productivity of resource utilization in edge computing and reduces the execution time of computations. Most of the research conducted so far has primarily focused on how to route packets for computation reuse, and only a limited number of studies have utilized in-network caching for this purpose. One of the issues in research that focuses on packet routing is that it can lead to an imbalance in load distribution among computing nodes. Therefore, this thesis proposes a new approach that considers computational requests with similar input data that lead to the same results as identical. This effectively takes a step towards computation reuse through ICN in-network caching. In this context, a new method for refining the Content Storage (CS) table is initially proposed, which takes input parameters into consideration. This method is specifically suitable for input parameters with small volumes. For computations where the inputs are images, a new metric called "Similarity Index" has been introduced to evaluate the degree of computation reuse. The goal of the Similarity Index is to effectively regard images as similar with minor changes in the angle of photography. As a result, this can create opportunities for computation reuse. The Similarity Index is provided through an algorithm called HNSW, which ultimately leads to the delivery of computation requests through caching instead of the main edge. Subsequently, computation reuse is achieved through a forwarding algorithm, in which the best compute node is selected based on the reuse criteria, load balancing, and capacity of edge nodes. Next, an analytical model is proposed for computing request transfer considering computation reuse in ICN-based edge computing. The proposed method has been simulated using the ndnSIM tool, and its performance has been evaluated. Simulation results show that the proposed method can reduce completion time by up to 8.88 times compared to the no-reuse case, while in previous works, this reduction was only 3.37 times. These findings highlight the efficiency and potential of the proposed method in optimizing edge computing performance.
