Electrical computers and digital processing systems: memory – Storage accessing and control – Specific memory composition
Reexamination Certificate
2000-12-26
2003-08-12
Ellis, Kevin L. (Department: 2186)
Electrical computers and digital processing systems: memory
Storage accessing and control
Specific memory composition
Reexamination Certificate
active
06606683
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an information recording apparatus and an information recording method using the information recording apparatus, and more particularly, to an information recording apparatus and an information recording method using the information recording apparatus that stores un-updated data in duplicate in more than two physically independent memories and thereby continues the updating processing, when data processing by one processing apparatus is stopped due to a power failure or memory failure, etc., by using un-updated data duplicated on the other processing apparatus.
2. Description of the Prior Art
A disk array apparatus writes data distributed into a plurality of disks, creates redundant data based on this data, writes this data in separate disks and thereby prevents loss of data due to disk failures. Even if a disk fails, such a disk array apparatus can restore the data as far as the level of redundancy permits and the data is not lost unless disks exceeding in number the level of redundancy fail. However, data may be lost in the case where data is not properly written in the disk, which will be described later.
Unlike a single disk, the disk array apparatus cannot perform a data write in one access. That is, the disk array apparatus requires processing in a plurality of stages such as reading old data which is originally written in the area in which data is to be written and an old parity on this old data, generating new parity data by using this data and data to be newly written and writing this data in a disk. Therefore, since the number of times in which the disk is accessed increases, which reduces the performance, a method of incorporating a cache memory in the disk array apparatus and notifying a host device of the completion of data write when a data transfer from the host device to the cache memory is completed.
In this method, the completion of data write is notified when the data transfer from the host device to the cache memory is completed, and therefore protection of data being written is an absolute condition. However, this method requires un-updated data to be held in the cache memory for a long time, which is likely to lead to data loss due to a memory failure or power failure, etc. For this reason, two control apparatuses are provided, data, which is to be processed between these apparatuses, is copied and thus data is controlled in duplicate.
BRIEF SUMMARY OF THE INVENTION
Object of the Invention
Thus, providing duplicated data reduces the possibility of data being lost due to a memory failure, power failure, etc., but maintaining consistency between the data recorded in the disk and redundant data such as parity related to this data is an important element for the disk array apparatus and losing this consistency will cause wrong data to be sent to the host device.
RAID (Redundant Array of Inexpensive Disks) uses a data distribution method having several levels. For example, there are various methods, like a method called “level 3” (RAID 3) in which data received from the host device is always divided into a plurality of disks, redundant data is created simultaneously with this division of data and these data items are written simultaneously as a data group whose consistency should be maintained or another method called “level 5” (RAID 5) in which data is divided into sectors and written and parity information is added as another sector as in the case of normal data.
FIG. 10
illustrates an example of data arrangement on disks that make up the RAID 5. The following two methods are available as data update methods in the disk array apparatus with such disk arrangement:
(1) Method of generating parity data by reading all data groups; whose consistency should be maintained
(2) Method of generating parity data using old data Here, the case where a write to a block (Data #
01
) from the host device is executed will be taken as an example. In the method (1), block data (Data #
09
), (Data #
11
) and (Data #
19
) are read from hard disks #
2
, #
3
and #
4
, respectively and new parity data is generated from these block data and content of new (Data #
01
) received from the host device, and the new (Data #
01
) and new parity data are written in hard disks #
1
and #
5
, respectively. Furthermore, in the method (2), old data (Data #
01
) and old parity data are read from the hard disks #
1
and #
5
, respectively, and new parity data is created from an exclusive OR between this old data, old parity data and new data (Data #
01
) received from the host device and the new (Data #
01
) and new parity data are written in the hard disks #
1
and #
5
, respectively.
The same result can be obtained by using either method. According to the method (1), in the case where more new data sent from the host device is found in the data group whose consistency should be maintained, the number of times the disk is accessed can be a small number. However, in the case of access to small blocks, the greater the number of disks that make up RAID, the greater the required number of times the disk is accessed to update one block. On the contrary, in the method (2), the number of times the disk is accessed can be a small number even when access is made to a small block.
When the method (2) is used, creating new parity data requires the old (Data #
01
) to be finally overwritten with new data and the old parity data to be read. In this case, even if un-updated data is duplicated using the cache memories of the two control apparatuses, if processing is stopped by a power failure, etc. while one control apparatus is rewriting the old (Data #
01
) or old parity data, there is a possibility that wrong new parity data will be created when the other control apparatus takes over the disk data updating processing using the duplicated un-updated data. That is, when processing is stopped while un-updated data or new parity is being written in the disk and the other control apparatus takes over the processing using the duplicated data, there is a possibility that the old data and old parity necessary to generate new parity may have already been overwritten. Once these old data and old parity have been overwritten, it will no longer be possible to generate new parity using the above-described method.
According to the method (2), processing of data updating in the disk becomes irreversible upon starting a write of un-updated data or the new parity in the disk and attempting to continue the updating processing using the data duplicated by the backup control apparatus without realizing that this point of entering into an irreversible state is passed will end up losing data consistency.
It is an object of the present invention to provide a disk array apparatus that is equipped with a plurality of independent memories, protects un-updated data by maintaining duplicated data in these memories, capable of carrying out data updating without data getting garbled when one control apparatus stops processing due to a power failure, etc. during a data communication and the other control apparatus takes over this processing properly. It is another object of the present invention to provide a disk array apparatus capable of allowing logically consistent parity data to be used in the disk arrays preventing wrong data from being restored and transferred to the host device even if the disk array apparatus loses redundancy when the processing is stopped.
SUMMARY OF THE INVENTION
The information recording apparatus of the present invention is an information recording apparatus configured in such a way as to comprise a plurality of controlling means for recording data received from a host device on an information recording medium with each controlling means having memory, temporarily store the data in memory of one controlling means and at the same time store the data in duplicate in memory of the other controlling means and, when data recording processing car
Ellis Kevin L.
NEC Corporation
Whitman Curtis & Christofferson, PC
LandOfFree
Information recording apparatus and control method thereof does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Information recording apparatus and control method thereof, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Information recording apparatus and control method thereof will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3113528