Electrical computers and digital processing systems: memory – Storage accessing and control – Specific memory composition
Reexamination Certificate
1997-09-16
2004-12-14
Portka, Gary (Department: 2188)
Electrical computers and digital processing systems: memory
Storage accessing and control
Specific memory composition
C711S113000, C714S006130
Reexamination Certificate
active
06832291
ABSTRACT:
CLAIM OF PRIORITY
This application makes reference to, and claims all benefits accruing under 35 U.S.C. §119 from an application for MEMORY SYSTEM FOR IMPROVING DATA INPUT/OUTPUT PERFORMANCE AND METHOD OF CACHING DATA RECOVERY INFORMATION earlier filed in the Korean Industrial Property Office on the 16
th
of Sep. 1996, and there duly assigned Serial No. 40202/1996.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a memory system such as a redundant array of inexpensive disks (RAID) and, more particularly, to a redundant arrays of inexpensive disks capable of providing high data input/output performance and a method of caching data recovery information using the redundant array of inexpensive disks.
2. Description of the Related Art
A high technology computer system depends considerably on its central processor unit (CPU) and input/output subsystem to increase overall system performance. While the information processing speed of the CPU has been dramatically improved in recent years because of VLSI technology, the performance of the input/output subsystem has not improved as desired. This increases the time required to access data in the memory system. Furthermore, since the cost of restoring data has increased when an error is generated in the input/output subsystem, an input/output subsystem having excellent performance and reliability is needed. As a solution to this, a disk array system known as a redundant array of inexpensive disks (RAID) constructed of a number of relatively small capacity disk drives has been proposed as a low cost alternative to a single large expensive disk for storing digital information.
RAID systems are now commercially available as cost effective mass storage providing reliable and continuous services to a host computer or network file server. The theory of RAID is to use relatively inexpensive disks, which may individually have a higher chance of failure than expensive disks, and compensating for this higher failure rate by adding redundancy by creating and storing parity blocks to facilitate recovery from a disk failure. Reports on the performance and reliability of disk arrays are presented in “
A Case For Redundant Arrays Of Inexpensive Disks
(
RAID
)” by D. Patterson, G. Gibson, and R. H. Kartz, at Report No. UCB/CDS87/89, December 1987, Computer Science Division (EECS), University of California, Berkeley, Calif. 94720. Exemplars of contemporary RAID systems are disclosed in U.S. Pat. No. 5,257,367 for Data Storage System With Asynchronous Host Operating System Communication Link issued to Goodlander et al., U.S. Pat. Nos. 5,367,669 and 5,455,934 for Fault Tolerant Hard Disk Array Controller issued to Holland et al., U.S. Pat. No. 5,418,921 for Method And Means For Fast Writing Data To LRU Cached Based DASD Arrays Under Drivers Fault Tolerant Modes issued to Cortney et al., U.S. Pat. No. 5,463,765 for Disk Array System, Data Writing Method Thereof, And Fault Recovering Method issued Kakuta et al., U.S. Pat. No. 5,485,598 for Redundant Disk Array (RAID) System Utilizing Separate Cache Memories For The Host System And The Check Data issued to Kashima et al., U.S. Pat. No. 5,522,032 for RAID Level 5 With Free Blocks Parity Cache issued to Franaszek et al., U.S. Pat. No. 5,530,948 for System And Method For Command Queuing On RAID Levels 4 And 5 Parity Drives issued to Islam, U.S. Pat. No. 5,579,474 for Disk Array System And Its Control Method issued to Kakuta et al., U.S. Pat. No. 5,640,506 for Integrity Protection For Parity Calculation For RAID Parity Cache issued to Duffy, and U.S. Pat. No. 5,636,359 for Performance Enhancement System And Method For A Hierarchical Data Cache Using A RAID Parity Scheme issued to Beardsley et al.
As generally discussed in the Patterson report and subsequent contemporary RAID systems, the large personal computer market has supported the developement of inexpensive disk drives having a better ratio of performance to cost than single large expensive disk systems. The number of input/outputs (I/Os) per second per read/write head in an inexpensive disk is within a factor of two of the large disks. Therefore, the parallel transfer from several inexpensive disks in a RAID system, in which a set of inexpensive disks function as a single logical disk drive, produces better performance than a single large expensive disk (SLED) at a reduced cost.
Unfortunately, when data is stored on more than one disk, the mean time to failure varies inversely with the number of disks in the array. In order to correct for this decreased mean time to failure of the system, error recognition and correction is characteristic of all RAID systems. Generally, each RAID system is organized in six structures commonly referred to as six levels each having a different means for error recognition and correction as described hereinbelow.
In a RAID structure of level 0, data is distributed and stored in all drives in the disk array, taking interests in performance rather than data reliability.
In a RAID structure of level 1, the mirroring, a conventional method of improving the disk performance, has a high cost since all contents of the disk must be stored in a reproduction disk without change. Accordingly, in a database system requiring a large-capacity disk space, only the fifty percent of the disk space can be used. However, the mirroring is the best way to enhance the data reliability because identical data is stored in the reproduction disk. In a RAID structure of level 2, this is used to minimize the cost required to enhance data reliability. The RAID structure of level 2 distributes and stores data in each disk array in bites, and has several test disks using a Hamming Code, besides the data disk, in order to recognize and correct errors.
In a RAID structure of level 3, data is input/output in parallel to/from the drive when input/output is requested once, and parity data is stored in a separate drive. Furthermore, disk spindles are synchronized so as to make all drives simultaneously input or output data. Accordingly, rapid data transmission can be carried out even if parallel input/output is not performed fast. If one drive has an error, the erroneous data can be restored by using the currently operated drive and parity drive even though the total data rate is decreased. The RAID structure of level 3 is used in an application which requires very fast data transmission rate, super computer and image manipulation processors. That is, the RAID of level 3 has a efficiency in a long data block transmission but has a lower efficiency in a short data block transmission which requires fast input/output request. Furthermore, since the data drive is used together with a single drive for redundancy, a device which is smaller than that used in the RAID of level 1 is used but its controller becomes more expensive and complicated.
In a RAID structure of level 4, the parity data is calculated and stored in a separate drive, and data is striped across. The data can be restored when it has error. Its reading performance is similar to that of a RAID of level 1 but its writing is much poorer than the single drive because the parity information must be provided to the single drive. Thus, the RAID structure of level 5 having improved writing performance is supplemented to the RAID of level 4.
In a RAID structure of level 5, data is striped across in each drive array, and parity data is distributed and stored in all drives in order to remove the bottleneck phenomenon when data is written. In this RAID structure, since the data written in all drives must be read in order to calculate the parity when the data is written, its speed is slower. However, it is possible to process the data input/output transmission and to restore data stored in a drive having an error. Accordingly, the RAID structure of level 5 is effective in the recording of long pieces of data, and is also effective in the recording of short pieces of data if an application program gives weight to the data reading or the array design is improved in order to incr
Bushnell , Esq. Robert E.
Portka Gary
Samsung Electronics Co,. Ltd.
LandOfFree
Memory system for improving data input/output performance... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Memory system for improving data input/output performance..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Memory system for improving data input/output performance... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3280599