Electrical computers and digital data processing systems: input/ – Input/output data processing – Input/output process timing
Reexamination Certificate
1999-10-01
2001-05-22
Lee, Thomas (Department: 2182)
Electrical computers and digital data processing systems: input/
Input/output data processing
Input/output process timing
C714S764000, C714S770000
Reexamination Certificate
active
06237052
ABSTRACT:
TECHNICAL FIELD
The present invention lies in the field of digital data storage and more specifically is concerned with disk drive controllers for multiple disk drives, generally known as disk drive arrays.
BACKGROUND OF THE INVENTION
Hard Disk Drives
Hard disk drives are found today in virtually every computer (except perhaps low-end computers attached to a network server, in which case the network server includes one or more drives). A hard disk drive typically comprises one or more rotating disks or “platters” carrying a magnetic media on which digital data can be stored (or “written”) and later read back when needed. Rotating magnetic (or optical) media disks are known for high capacity, low cost storage of digital data. Each platter typically contains a multiplicity of concentric data track locations, each capable of storing useful information. The information stored in each track is accessed by a transducer head assembly which is moved among the concentric tracks. Such an access process is typically bifurcated into two operations. First, a “track seek” operation is accomplished to position the transducer assembly generally over the track that contains the data to be recovered and, second, a “track following” operation maintains the transducer in precise alignment with the track as the data is read therefrom. Both these operations are also accomplished when data is to be written by the transducer head assembly to a specific track on the disk.
In use, one or more drives are typically coupled to a microprocessor system as further described below. The microprocessor, or “host” stores digital data on the drives, and reads it back whenever required. The drives are controlled by a disk controller apparatus. Thus, a write command for example from the host to store a block of data actually goes to the disk controller. The disk controller directs the more specific operations of the disk drive necessary to carry out the write operation, and the analogous procedure applies to a read operation. This arrangement frees the host to do other tasks in the interim. The disk controller notifies the host, e.g. by interrupt, when the requested disk access (read or write) operation has been completed. A disk write operation generally copies data from a buffer memory or cache, often formed of SRAM (static random access memory), onto the hard disk drive media, while a disk read operation copies data from the drive(s) into the buffer memory. The buffer memory is coupled to the host bus by a host interface, as illustrated in
FIG. 1
(prior art). Disk data is often buffered by another static ram cache within the drive itself. The drive electronics controls the data transfers between this cache and the magnetic media.
Disk Drive Performance and Caching
Over the past twenty years, microprocessor data transfer rates have increased from less than 1 MByte per second to over 100 megabytes per second. At the current speeds, hierarchical memory designs consisting of static ram based cache backed up by larger and slower SRAM can utilize most of the processor's speed. Disk drive technology has not kept up, however. In a hard disk drive, the bit rate of the serial data stream to and from the head is determined by the bit density on the media and the RPM. Unfortunately, increasing the RPM much above 5000 causes a sharp drop off in reliability. The bit density also is related to the head gap. The head must fly within half the gap width to discriminate bits. With thin film heads and high resolution media, disks have gone from 14″ down to 1″ diameter and less, and capacities have increased from 5 MBytes to 20 GBytes, but data transfer rates have increased only from 5 to about 40 MBits per second which is around 5 MBytes per second. System performance thus is limited because the faster microprocessor is hampered by the disk drive data transfer “bottleneck”.
The caching of more than the requested sector can be of advantage for an application which makes repeated accesses to the same general area of the disk, but requests only a small chunk of data at a time. The probability will be very high that the next sector will already be in the cache resulting in zero access time. This can be enhanced for serial applications by reading ahead in anticipation before data from the next track is requested. More elaborate strategies such as segmenting and adaptive local cache are being developed by disk drive manufacturers as well. Larger DRAM based caches at the disk controller or system level (global cache) are used to buffer blocks of data from several locations on the disk. This can reduce the number of seeks required for applications with multiple input and output streams or for systems with concurrent tasks. Such caches will also tend to retain frequently used data, such as directory structures, eliminating the disk access times for these structures altogether.
Various caching schemes are being used to improve performance. Virtually all contemporary drives are “intelligent” with some amount of local buffer or cache, i.e. on-board the drive itself, typically in the order of 32K to 256K. Such a local buffer does not provide any advantage for a single random access (other than making the disk and host transfer rates independent). For the transfer of a large block of data, however, the local cache can be a significant advantage. For example, assume a drive has ten sectors per track, and that an application has requested data starting with sector one. If the drive determines that the first sector to pass under the head is going to be sector six, it could read sectors six through ten into the buffer, followed by sectors one through five. While the access time to sector one is unchanged, the drive will have read the entire track in a single revolution. If the sectors were read in order, it would have had to wait an average of one half revolution to get to sector one and then taken a full revolution to read the track. The ability to read the sectors out of order thus eliminates the rotational latency for cases when the entire track is required. This strategy is sometimes called “zero latency”.
Disk Arrays
Despite all of the prior art in disk drives, controllers, and system level caches, a process cannot average a higher disk transfer rate than the data rate at the head. DRAM memory devices have increased in speed, but memory systems have also increased their performance by increasing the numbers of bits accessed in parallel. Current generations of processors use 32 or 64 bit DRAM. Unfortunately, this approach is not directly applicable to disk drives. While some work has been done using heads with multiple gaps, drives of this type are still very exotic. To increase bandwidth as well as storage capacity, it is known to deploy multiple disks operating in concert, i.e. “disk arrays”. The disk array cost per MByte is optimal in the range of 1-2 GBytes. Storing larger amounts of data on multiple drives in this size range does not impose a substantial cost penalty. The use of two drives can essentially double the transfer rate. Four drives can quadruple the transfer rate. Disk arrays require substantial supporting hardware, however. For example, at a 5 MBytes per second data rate at the head, two or three drives could saturate a 16 MByte per second IDE interface, and two drives could saturate a 10 MByte per second SCSI bus. For a high performance disk array, therefore, each drive or pair of drives must have its own controller so that the controller does not become a transfer bottleneck.
While four drives have the potential of achieving four times the single drive transfer rate, this would rarely be achieved if the disk capacity were simply mapped consecutively over the four drives. A given process whose data was stored on drive
0
would be limited by the performance of drive
0
. (Only on a file server with a backlog of disk activity might all four drives occasionally find themselves simultaneously busy.) To achieve an improvement in performance for any single process, the data for that process must be distributed ac
Du Thuan
Lee Thomas
NetCell Corporation
Rives LLP Stoel
Stolowitz Micah D.
LandOfFree
On-the-fly redundancy operation for forming redundant drive... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with On-the-fly redundancy operation for forming redundant drive..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and On-the-fly redundancy operation for forming redundant drive... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2511640