Improving Performance of Docker Instance via Image Reconstruction

Improving Performance of Docker Instance via Image Reconstruction

연구 분야: Software Development

학회: International Conference on Big Data Intelligence and Computing

Docker has become a key paradigm for application deployment in cloud computing. A running Docker instance need to access its corresponding docker image for searching the required file from time to time. The more layers a docker image has, the longer the search time is. The number of image layers has significant impact on the efficiency of the application. However, reducing the number of image layers via layer merging will reduce the shareability of image layers and increase the storage occupation. To address the above issue, we exploit image reconfiguration to achieve a trade-off between the operation efficiency and storage occupation. We formulate the image reconstruction problem as a knapsack problem, and design a dynamic programming-based offline image reconstruction algorithm named DPOIRA to solve it. We conduct experiments on the busybox and crate, and the experiment results show the effectiveness of our algorithm.