Abstract:
Nowadays, each websites and e-commerce sites are designed in a manner that pages can be loaded faster. For faster page loading through optimizing websites, one of the major problems related to images are absence of having images in their intended sizes. Hence, a browser needs to resize the images every time a page gets loaded. Thus, while loading a website, most of the time overhead pertains to image related tasks such as its loading and resizing. Here, pre-availability of smaller-size images can substantially increase the loading speed. Accordingly, in this study, we propose a new clouds framework to enable faster image loading from a cloud through ensuring the pre-availability. To ensure the pre-availability, we enable necessary pre-processing of images going beyond the conventional online processing. Accordingly, we expand cloud file sharing capabilities from the conventional “only storing images” to also performing pre-resizing along with security enforcing. We propose to do it through moving and executing user-defined programs near the data inside a private object storage cloud, which is significantly different from traditional public clouds having well-known security vulnerabilities. As arbitrary separation of storage and computation can increase latency and decrease performance of cloud file sharing, our approach is to first realize all application requirements and then to move images after processing them based on the realized requirements (rather than moving images first and then process them where the application is located). Here, we perform the first task of resizing images according to several dimensions for covering diversified remote user devices available now-a-days. Then, we perform encoding and encryption of images using P-Fibonacci transform of Discrete Cosine Coefficients (PFCC) algorithm. Afterwards, we resize images using Bicubic interpolation method to both JPEG and PJPEG types through adding a new middleware named iBuck for the purpose of faster and smooth retrieval. We adopt the necessary algorithms and image types after careful evaluation on performances of the existing alternatives in our intended framework. Here, our evaluation rigorously subsumes diversified techniques covering both objective and subjective evaluations including QoE metrics such as Mean Square Error, Structural Similarity Index, etc. Furthermore, we conduct detailed experimentation over real setups to evaluate
v
performance of our implemented middleware in a cloud le sharing environment with both custom and standard data sets. Our evaluation reveals substantial performance improvement while using our proposed cloud framework compared to that using the conventional cloud framework.
