CONTINUOUS MEDIA PRIORITY AWARE
STORAGE SCHEDULER
REFERENCE TO RELATED APPLICATION
[0001] This application claims priority to U.S. Provisional Patent Application Ser. No. 60,496,529 which was filed Aug. 20, 2003, entitled CONTINUOUS MEDIA PRIORITY AWARE STORAGE SCHEDULER.
TECHNICAL FIELD
[0002] The present invention relates generally to computer systems, and more particularly to a system and method that facilitates scheduling data requests to a disk in an efficient and bounded manner.
BACKGROUND OF THE INVENTION
[0003] To reduce the performance penalty incurred by seeking a drive head, operating system software (and drive firmware) traditionally reorder I/O packets to minimize seek time. Much research has been done to address long seek times while still maintaining fair access to the disk drive. Traditional disk scheduling algorithms generally reorder requests destined for a disk to maximize data throughput. One common algorithm implemented by many operating systems is the C-LOOK algorithm, often incorrectly referred to as the C-SCAN algorithm. The C-LOOK algorithm reorders any number of requests to achieve the highest throughput to a device that can be attained. However, in this type reordering, the C-LOOK algorithm can introduce large latencies between the time a request is begun and when the request is eventually completed.
[0004] As noted above, the most common algorithm that is implemented is the C-LOOK algorithm which orders requests in order that the disk head sweeps in one direction across the disk performing Input/Output operations (I/Os), then returns to the other side of the disk to perform additional I/Os. This is often referred to as logical block address ordering or LBA ordering. Unfortunately, traditional disk scheduling algorithms trade-off higher throughput to the disk, for longer access times (or latencies) over simple first-in-first-out (FIFO) queuing. This provides better throughput, but at the cost of potentially and dramatically higher latencies for I/O requests. For traditional operating systems, this trade-off is well worth the cost. For newer systems that need to support audio-visual streaming applications, in one example, merely trading latency for throughput is often not acceptable.
[0005] Many alternative disk-scheduling algorithms have been implemented. In particular, the first-come-first-serve (FCFS, also sometimes referred to as FIFO) algorithm and earliest-deadline-first (EDF) algorithms have been proposed respectively for scheduling in multimedia applications. The FCFS algorithm bounds the time to perform an I/O as a function of the current queue length and the EDF algorithm guarantees that requests with earlier deadlines will be executed before requests with later deadlines.
[0006] Unfortunately, both the FCFS and EDF algorithms suffer from poor I/O throughput. On special purpose and real-time operating systems, this poor utilization of the disk may not be a drawback, but on general purpose operating
systems that also host multimedia applications, for example, significantly decreasing throughput to bound the latency to perform I/O requests is generally unacceptable.
SUMMARY OF THE INVENTION [0007] The following presents a simplified summary of the invention in order to provide a basic understanding of some aspects of the invention. This summary is not an extensive overview of the invention. It is not intended to identify critical elements of the invention or to delineate the scope of the invention. Its sole purpose is to present some concepts of the invention in a simplified form as a prelude to the more detailed description that is presented later.
[0008] The present invention relates to systems and methods that facilitate dynamic scheduling of data requests to a storage media in an efficient, and timely manner. The data requests are processed as Input/Output (I/O) reads and writes to the storage media such as a disk drive or other type media. Scheduling is achieved in an architecture that reorders a fixed number of requests per scan of the disk drive, thus providing a fixed upper bound on the latency of the request wherein each scan of the disk drive is referred to as a round. Since the requests are reordered into an efficient pattern for the disk drive to execute, high throughput to the disk drive is also achieved. Thus, a disk scheduling system is provided that includes at least one scheduling component. The scheduling component employs a predetermined number of requests within a round to provide a particular latency guarantee for the requests while maintaining high throughput level in connection with disk updates.
[0009] In one aspect an admission controller is provided that categorizes I/O requests according to a determined category of I/O. For example, I/O associated with multimedia applications could be categorized as periodic I/O whereas other I/O could be characterized as aperiodic I/O. Aperiodic I/O is further characterized by priority where there are any fixed number of priorities. In one example implementation, the number of priorities may be limited to 32, however this can easily be changed to more or less values in other implementations.
[0010] Periodic I/O requests can be queued to the periodic I/O queue and maintained in earliest-deadline-first (EDF) order, if desired. Aperiodic I/O requests can be queued to a separate aperiodic I/O queue based upon the priority of the request; wherein the aperiodic queues can be maintained in FCFS order, for example. During each round of the algorithm, a fixed number of requests from the periodic queue and the aperiodic queues are removed and placed on a sweep queue. The sweep queue is maintained in C-LOOK order. The requests are removed from the sweep queue in C-LOOK order and scheduled to the disk. Since the elements are scheduled to the physical drive in C-LOOK order, high performance is maintained to the drive.
[0011] To the accomplishment of the foregoing and related ends, certain illustrative aspects of the invention are described herein in connection with the following description and the annexed drawings. These aspects are indicative of various ways in which the invention may be practiced, all of which are intended to be covered by the present invention. Other advantages and novel features of the invention may become apparent from the following detailed description of the invention when considered in conjunction with the drawings.
