(In collaboration with Kaiyuan Zhang and Xi Wang)
ShoeBox
Images are among the most popular forms of data stored and transferred on the internet. In many applications–especially social media services, users upload and retrieve images using a wide variety of devices with different screen sizes, resolutions, and pixel densities. To accomodate these use cases, modern image storage must efficiently handle serving images at multiple quality levels or resolutions. Previous approaches have sacrificed storage efficiency for performance, storing a (redundant) version of each image for each quality level. ShoeBox is a project designed to enable efficient multi-quality image storage by only storing a single version of each image. To improve performance, ShoeBox only reads the required image data (depending on the specified quality) for each request.