Damien Le Moal

Tempo: Disk drive power consumption characterization and modeling

Accurately modeling a hard disk drive performance and power requirements is of significant interest to designers of Consumer Electronic systems that include a disk drive as a storage device. In this paper, a method to create a detailed power and performance model of a disk is presented. The proposed method is used to accurately estimate the disk power consumption using a trace of all requests issued to the disk with a real-time streaming workload.

Cost-effective streaming server implementation using Hi-tactix

High performance and high quality for continuous media stream delivery needed by streaming server systems cannot be achieved efficiently using general-purpose operating systems, due to the overhead of the I/O mechanism implementation generally used. Special OS combined with powerful hardware can deliver better performance and quality but increases development complexity and deployment costs. The External I/O Engine Architecture adopts a hybrid approach, implementing streaming engines using the streaming-oriented Hi-Tactix operating system on inexpensive hardware, in combination with existing stream servers. The evaluation results of a QuickTime video server implemented with the existing Darwin Streaming Server using the External I/O Engine Architecture shows that Hi-Tactix can deliver 5 times the performance of a conventional OS and better stream quality, while keeping the amount of code necessary low.

Power efficient real-time disk scheduling

Hard-disk drive power consumption reduction methods focus mainly on increasing the amount of time the disk is in standby mode (disk spun down) by implementing aggressive data read-ahead and caching at the operating system and/or application level. However, these methods cannot be applied efficiently to systems with limited memory and high bit-rate requirements such as digital video recorders handling high-definition video. In this paper, we introduce the Audio/Video File System (AVFS), composed of a file system and a disk I/O scheduler. Compared to traditional methods, the proposed scheduler reduces seek overhead by processing real-time requests to video files using batches built dynamically depending on the requests deadlines. Evaluation results show an important reduction in disk utilization rates and a reduction of up to 20 % of the disk power consumption with only 4 MB of data buffer per video stream.