Skip to main content
Fig. 9 | EURASIP Journal on Wireless Communications and Networking

Fig. 9

From: Partitioning multi-layer edge network for neural network collaborative computing

Fig. 9

Response time changes as the number of batch task requests increases. One advantage of multi-layer edge network collaborative computing is pipelined collaborative computing. When there are batch requests to be processed, the logic layers are processed in parallel like the factory pipeline. network collaborative computing. We still use LeNet5, AlexNet, VGG network in this experiment and set the device bandwidth as 100 KB. It shows the experimental results of the response time of each computing method when processing requests in batch. When the number of requests in each batch is small, the gap between computing methods is not obvious. With the increase of each batch of requests, the processing time of cloud computing and single-edge device computing increases faster. This is because these two computing modes cannot rely on pipeline for parallel processing, and the request accumulation is serious. The edge-cloud collaborative computing and multi-layer edge network collaborative computing process faster than the first two methods

Back to article page