Comparison of analytical and ML-based models for predicting CPU–GPU data transfer time

Secure data stream transmission method for cell pathological image storage system

“…The proposed method was compared with References [4,5] methods to mainly compare the node energy consumption, transmission ratio, and transmission bandwidth consumption. The target network range was 200 m × 200 m, the number of test nodes was 200, the initial energy of nodes before data transmission was 4 J, and the packet transmission length was uniformly set as 400 bytes.…”

“…Taking network data packets actually collected as transmission objects, the proposed method, References [4,5] methods are used to transmit the same data group. The advantages of the proposed method are proved by comparing the results.…”

“…By comparing and analyzing the transmission efficiency of different methods given in Table 4, it can be seen that the transmission efficiency obtained by Reference [4] method for network data with a size of 50 MB is  50 512.32 0.097 ∕ MB/s. The transmission efficiency of Reference [5] method is  50 254.25 0.796 ∕ Mb/s for the network data with a size of 50 Mb, and the corresponding transmission efficiency of the proposed method is  50 4.25 11.76 ∕ Mb/s.…”

“…For the size of the same network packets, three methods of the corresponding transfer efficiency gap are larger, and the proposed method by constructing the network data transmission model to analyze the actual data, when the transmission of the packet size is 5 MB, transfer rate of 2 s, that is to say, needs to transfer at least 2.5 MB/s to meet the requirements of network data realtime transmission. However, both References [4,5] methods are difficult to meet the requirements of real-time transmission, which will cause data accumulation and occupy a large amount of node transmission bandwidth, and are not suitable for big data network environment.…”

“…By comparing with the Machine Learning (ML) model in reference, an analytical model for predicting the safe transfer time of Central Processing Unit-Graphic Processing Unit (CPU-GPU) data stream is proposed. 4 The accurate prediction of data stream transfer time is quite effective for improving the performance of GPU analytical modeling, the prediction accuracy of kernel performance, and the composition of CPU and GPU to solve computing problems. To estimate the data transfer time between the CPU and the GPU, the present research employs three models based on ML and a new analytical model called the λ-model.…”