Electrical computers and digital processing systems: memory – Storage accessing and control – Specific memory composition
Reexamination Certificate
1999-01-15
2002-06-18
Kim, Matthew (Department: 2186)
Electrical computers and digital processing systems: memory
Storage accessing and control
Specific memory composition
C711S136000, C711S129000, C711S171000, C711S173000
Reexamination Certificate
active
06408357
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to data storage devices and methods. More particularly, the present invention relates to a disk drive having a cache memory and to a method of optimizing the operating of such a cache memory.
2. Description of the Prior Art and Related Information
Disk drives typically communicate with host computers via standardized interfaces. One such standardized interface is the Integrated Drive Electronics (IDE) interface, sometimes referred to as the AT attachment (ATA) interface. The IDE interface was originally based upon the IBM PC AT 16-bit bus standard, although most disk drives today utilize an enhanced version of the standard, which may be called EIDE or ATA-n where n represents a generation of the standard which has been documented in a standards committee. In the IDE or EIDE or ATA interfaces (hereafter collectively IDE), the disk drive controller is built into the logic board of the disk drive. The IDE standard allows for the presence of a cache memory in the disk drive controller to temporarily store data being transferred to and from the host computer system. Cache memories allow disk drives to temporarily store data, likely to be requested by the host computer, in a semiconductor memory. The latencies associated with semiconductor memories are orders of magnitude smaller than the latencies inherent in mechanically accessing data stored on a rotating disk. Therefore, retrieving data from the cache memory rather than from disk allows for a faster retrieval of needed data and thus leads to more efficient use of processor resources.
In IDE and other interface standards, the smallest indivisible unit addressable on a disk is a logical block (512 bytes), and each such logical block is assigned a Logical Block Address (LBA). When the operating system of the host computer sends a command to the disk drive, the nature of the command is specified (read, write), along with the LBA and the count, specifying the number of contiguous logical blocks to be transferred. To keep track of where data is stored on the disk or disks, the file system of the operating system often maintains a logical structure specifying the location of each file stored on the disk. This logical structure is often called a File Allocation Table (FAT), each entry of which points to the start of a file stored on the disk or to the next cluster of the file if the file is larger than one cluster. Typically, the disk is written to in groups of blocks by specifying the LBA of the first block and the count, or number of subsequent contiguous blocks. Different operating systems specify a different number of blocks to update the FAT. For example, the Microsoft Windows 95 and 98® operating systems specify a 1-block write to update the FAT, while the Windows NT® operating systems specifies an 8-block FAT update.
To more efficiently utilize the available cache memory, it is known to divide a disk drive cache memory into one or more logical segments and to further set up one or more of these logical segments as ring (circular) buffers for temporary storage of read or write data. Also known are adaptive cache management methods wherein cache logical segments are allocated and managed according to whether the data is sequential, random or repetitive, as disclosed in commonly assigned U.S. Pat. No. 6,092,149, entitled “DISK DRIVE CACHE SYSTEM USING A DYNAMIC PRIORITY SEQUENTIALLY STREAM OF DATA SEGMENTS CONTINUOUSLY ADAPTED ACCORDING TO PREFETCHED SEQUENTIAL, RANDOM, AND REPORTING TYPES OF ACCESSES” (as amended)” issued on Jul. 18, 2000, in the names of M. Hicken; S. Howe; D. Sokolov; T. Swatosh and J. Williams, the disclosure of which is hereby incorporated herein by reference.
However, such segmentation, although highly effective in enhancing (i.e., speeding up) drive operations under certain operative conditions, may nevertheless be somewhat less effective under certain other conditions. One such condition, for example, is that of repetitive FAT updates. In the Windows 95 (using a 16-bit FAT or FAT16) or 98® (using a 32-bit FAT or FAT32) environment, the FAT updates manifest themselves as 1-block writes, whereas in a Windows NT® environment, such FAT updates consist of 8-block writes. Such FAT updates may not make the most efficient use of available cache memory resources. For example, when a 1-block write is performed to update the FAT, an entire multi-block logical segment is typically used, resulting in wasted cache space. Indeed, as the logical segments typically are greater than 1 logical block, a 1-block write to a logical segment renders the remaining blocks within that logical segment inaccessible and thus unavailable to subsequent writes. Moreover, in the case of FAT updates, the actuator on which the disk read-write heads are disposed must repetitively sweep across the radial width of the disk to write the data and to update the FAT, as the FAT is typically located at one end of the disk whereas the corresponding data is typically spread over a wide area. Such repetitive FAT updates limit the performance of the drive by causing the actuator to repeatedly sweep across the disk, and may decrease the performance of the host processor as it sits idle, waiting for requested data or for confirmation that the data has indeed been written to disk.
One commonly accepted means of measuring the performance of computer systems is by testing the drive against industry-standard benchmark programs, such as the Winbench series of programs developed by Ziff-Davis, Inc. Such benchmark programs attempt to simulate the demands of typical applications and to measure the disk performance during the simulation. In the process of analyzing the disk access patterns during the operation of such benchmark programs, numerous instances of single-block writes were observed, interspersed with other operations pertinent to the operation of the benchmark. Further investigations revealed that the vast majority of the single-block writes were made for the purpose of updating the FAT and that many overwrites of the same logical block occurred. Such repetitive single block writes to update the FAT, moreover, were found to degrade the performance of the drive, as measured by the benchmark program. As the benchmark programs are designed to simulate the normal operation of typical software applications, it is expected that such repetitive FAT updates observed during the operation of the benchmark programs similarly degrade disk performance in real world situations. Moreover, the problems associated with repetitive single-block writes are not confined to the Windows 95 or 98® operating systems, but are also present in the eight-block writes to update the FAT in the Windows NT® operating system. Other operating systems may also experience similar degradation in performance due to repetitive low block count writes. Such repetitive, low block count writes may be carried out for the purpose of updating a file system or for some other wholly unrelated purpose. There is, therefore, a need for methods and devices to mitigate the effects of such performance degrading repetitive single-block and low block count writes both during the execution of benchmark programs as well as during normal operation of the drive.
SUMMARY OF THE INVENTION
An object of this invention, therefore, is to enhance the performance of disk drives by providing means and methods to mitigate the effects of such performance degrading repetitive single-block and low block count writes to cache memory.
Accordingly, this invention can be regarded as a method of operating a disk drive having a cache memory wherein the disk drive is connectable to a host computer for reading and writing data on a disk. The method comprises the steps of defining a length parameter specifying the length of a data segment to be written to the disk; defining a first portion of the cache memory for storing a data segment whose length is equal to the length parameter and defining a second portion of the cache memory for storing a data segm
Hanmann Jonathan Lee
Kellerman Marcus C.
Anderson Matthew D.
Kim Matthew
Shara Esq Milad G
Western Digital Technologies Inc.
LandOfFree
Disk drive having a cache portion for storing write data... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Disk drive having a cache portion for storing write data..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Disk drive having a cache portion for storing write data... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2933650