Use-case-optimized data storage of scalable media files

In professionel movie post-production, various components are pushed to their performance limits. Former processors for example were not able to playback the JPEG 2000 compressed image-sequences which are used for distribution in Digital Cinema as well as long term storage in digital archiving - in real time. While today´s processors are able to decode such files a new bottleneck became part of the processing chain: In many cases conventional Hard Disk Drives (HDD) are not able to deliver the requested data - which is limited to a maximum value of 250 Mbit/s in Digital Cinema - in real time. In this paper we propose an algorithm for increasing the data throughput of conventional HDDs by utilizing the progression-order of scalable media files, called UCODAS (Use-Case-Optimized DAta Storage). The motivation is given by the architecture of conventional hard drives, and finding that the file structure of scalable media, such as JPEG 2000, can be (re )arranged in such a way that the throughput of the disk can be significantly increased - especially if subsequent access patterns to the image-sequence are known a-priori. The advantages and disadvantages of today´s hard drives are summarized before we show how scaling is achieved within JPEG 2000. Then, various methods for improving the performance of HDDs - taking advantage of the scalability - are proposed and the data sets used for the measurements are described. We performed tests using a collection of common files systems including FAT32, NTFS, ext2 and ext3 as well as RAW data access without a file-system in order to prove our implementation of the UCODAS algorithm. In particular, we show that UCODAS can increase the data-throughput of a conventional HDD by more than a factor of 3 and thus overcome the bottleneck introduced by conventional HDDs.