The Research And Implement Of Distributed High Performance Image Server

As we have entered the age of web 2.0, Internet contents are created by users instead of the web server itself nowadays. As the number of users becomes larger and larger, the scale of storage and processing is enlarged as well. Single server strategy can no longer meet the requirements like high throughput, high concurrency, and strong reliability. Distributed system is born to solve such problems. Distributed systems adopt the "divide and conquer principle", distributing storage and processing to a group of servers to gain high performance.Distributed image service becomes more and more important for Internet as social networks like Facebook, Tencent and e-commerce systems like Taobao, JingDong become more and more popular. Hundreds of billions of images are saved at the backend of these systems. How to store and process such a large scale of images becomes a tough problem. On one hand, images in internet are always tiny and massive, which are not suite for storage systems like GFS to store. On another hand, because image applications always need to use different size types of a same image, how to build a system that can server any size types of images become a practical issue to solve.Our paper is to research and implement a distributed high performance image storage and processing system- WIMG. WIMG adopts the traditional master/slave architecture, distributing file storage and image processing. WIMG uses a cache layer and a save-new strategy to gain high performance. Beyond testing, WIMG is a stable, reliable and extendible system. WIMG also provide a client for other applications or systems to use WIMG as the backend system.