This invention relates in general to redundant arrays of disks, such as RAID (Redundant Array of Independent Disks) sets. More specifically, the invention relates to an apparatus and method for providing virtual storage volumes, particularly very large virtual storage volumes (e.g., 100 Gigabytes (GB) or more), using redundant arrays of disks, such as RAID sets.
Some computer software applications are organized according to what is referred to as a “single volume architecture,” meaning that they store data in a single data file that resides on a single volume. This “volume” may be a physical volume, such as a disk drive, or it may be a virtual volume, such as a RAID set. The Exchange® e-mail program provided by Microsoft Corporation of Redmond, Wash. is an example of such a single-volume-architecture application.
In some cases, the single volume architecture of a particular software application can be problematic, because the size of the data file the application needs to store on a single volume exceeds the capacity of the volume. For example, implementations of Microsoft's Exchange® e-mail program in large organizations having many e-mail users can require a single-volume storage capacity exceeding 100 GB, which is greater than many conventional volumes, physical or virtual, can provide. Although it is possible to solve this problem by changing a single-volume-architecture application into a multi-volume-architecture application so that it saves data in multiple files spread across multiple volumes, such efforts can be prohibitively time-consuming and expensive.
Accordingly, there is a need in the art for a very large virtual storage volume having the storage capacity necessary to meet the needs of a single-volume-architecture software application, such as Microsoft's Exchange® e-mail program. Preferably, such a storage volume should have built-in disaster tolerance capabilities through the use of remote mirroring or other techniques in order to ensure the integrity of its stored data. In addition, such a storage volume should preferably have cloning capabilities so that data backup can occur off-line without interrupting read/write access to the data.
An inventive apparatus for providing a very large storage volume includes a plurality of disks and a local back-end controller that organizes and presents the disks as redundant arrays of disks (e.g., RAID-5 sets). Also, a local front-end controller stripes the redundant arrays of disks and presents the striped arrays as a very large storage volume.
To provide local redundancy, another plurality of disks and an associated back-end controller can be provided, in which case the local front-end controller forms mirror sets from the redundant arrays of disks presented by both back-end controllers. In addition, a further plurality of disks and an associated back-end controller can be provided to enable off-line backup of the data stored on the volume by cloning the data onto the disks, and then using the disks as the data source for off-line backup. Also, a still further plurality of disks and an associated back-end controller can be provided at a remote location to protect against disasters occurring at the primary location (commonly referred to as “disaster tolerance”). The disks and back-end controllers providing cloning capabilities and disaster tolerance can be incorporated into the mirror sets formed by the local front-end controller. Further, spare disks can be provided on any or all of the back-end controllers to allow restoration of redundancy after the loss of any particular disk.
If, for example, the disks each have 9.1 GB of storage capacity and the local back-end controller organizes the disks into eleven, six-member RAID-5 sets, then the very large storage volume has a storage capacity in excess of 500 GB, which should be adequate for most single-volume architecture programs. In addition, the redundancy restoration capabilities provided by the spare disks, the parity associated with RAID-5 sets, and the mirroring ensures the integrity of the data stored on the very large storage volume.
In another embodiment of this invention, the apparatus described above can be incorporated into an electronic system that also includes a host computer.
In a further embodiment of this invention, data is stored on a plurality of disks by organizing the disks into a plurality of redundant arrays of disks. The redundant arrays of disks are striped together to form a virtual volume, and the data is then written to the virtual volume.
In still another embodiment of this invention, data is again stored on a plurality of disks by organizing the disks into a plurality of redundant arrays of disks. Mirror sets are formed from the redundant arrays of disks, and these mirror sets are then striped together to form a virtual volume. The data is then written to the virtual volume.
As shown in
As used herein, a “RAID-0 stripe set” will be understood to refer to a virtual volume comprised of two or more member disks or volumes across which “stripes” of data are stored. Also, as used herein, a “RAID-1 mirror set” will be understood to refer to a virtual volume comprised of two or more member disks or volumes, each of which contains an identical copy of the data stored in the mirror set.
The mirror set M1, for example, comprises five, 45.5 GB RAID-5 sets PL1, RL1, C1, PR1, and RR1 as members. Similarly, the mirror set M11 comprises five, 45.5 GB RAID-5 sets PL11, RL11, C11, PR11, and RR11 as members. For purposes of clarity, the RAID-5 set members of the mirror sets M2–M10 are illustrated but not labeled.
Of course, it will be understood that the members of the mirror sets M1–M11 can be other than RAID-5 sets (e.g, RAID-3 or RAID-4 sets). Also, as used herein, a “RAID-5 set” will be understood to refer to a virtual volume comprised of three or more independently accessible member disks or volumes having redundancy protection through parity information distributed across its members.
The RAID-5 sets PL1–PL11 comprise the primary local storage copy of the data stored in the very large volume 10, which means that they are the primary location to which the data is written and from which the data is read. Also, the RAID-5 sets RL1–RL11 comprise a redundant local storage copy of the data, which provides mirroring-type redundancy for the stored data. In addition, the RAID-5 sets C1–C11 comprise a cloning storage copy of the data, which is convenient for use in performing off-line data backups without interrupting read/write activities to the very large volume 10. Disaster tolerance is provided by the RAID-5 sets PR1–PR11, which comprise a primary remote storage copy, and the RAID-5 sets RR1–RR11, which comprise a redundant remote storage copy. Of course, it should be understood that embodiments of this invention may exclude the redundancy provided by the RAID-5 sets RL1–RL11, the cloning capability provided by the RAID-5 sets C1–C11, or the disaster tolerance provided by the RAID-5 sets PR1–PR11 and RR1–RR11.
The RAID-5 set PL1, for example, comprises six, 9.1 GB physical disks 12 distributed across six SCSI busses bus1–bus6 of a back-end controller (see
As described herein, the very large volume 10 has a high degree of redundancy. If the bus3 physical disk 12 fails, for example, it is replaced by the bus3 spare disk 16 by using parity data to rebuild the data stored on the failing bus3 physical disk 12 onto the replacement bus3 spare disk 16. If bus3 itself fails, for example, the parity redundancy in the RAID-5 sets PL1–PL11 regenerates the data stored on the failing bus3. If the back-end controller (see
As shown in a block diagram in
The primary local storage copy comprises physical disks 24 (which include disks 12, 14, and 16 of
In addition, the very large volume 10 is connected to a remote host computer 44 that reads data from, and writes data to, the volume 10 via a remote front-end controller 46 that acts as a mirroring and striping engine for the primary and redundant remote storage copies. The local and remote host computers 20 and 44 are connected via a network interconnect 48, such as the internet or a dedicated network line.
In an alternative embodiment, the front-end controller 22 can be configured to present a 273 GB unit and a 227.5 GB unit to the local host computer 20, rather than the single 500.5 GB unit described above. Of course, it should be understood that the front-end controller 22 can, in fact, be configured in a multitude of ways to group the mirror sets M1–11 into between one and eleven total units, each potentially ranging in size from 45.5 GB to 500.5 GB. Further, it should be understood that the front-end controller 22 can be configured (typically using software) to partition the striped mirror sets M1–11 into an infinite number and size of units.
In addition, it should be understood that although this invention has been described with reference to an embodiment having two levels of stacked controllers, the invention is not limited to the two levels described. Rather, the invention includes within its scope any number of levels of stacked controllers.
Although this invention has been described with reference to particular embodiments, the invention is not limited to these described embodiments. Rather, the invention is limited only by the appended claims, which include within their scope all equivalent devices and methods that operate according to the principles of the invention as described.
Number | Name | Date | Kind |
---|---|---|---|
5611069 | Matoba | Mar 1997 | A |
5678061 | Mourad | Oct 1997 | A |
5991844 | Khosrowpour | Nov 1999 | A |
6115788 | Thowe | Sep 2000 | A |
6148383 | Micka et al. | Nov 2000 | A |
6178521 | Filgate | Jan 2001 | B1 |
6182198 | Hubis et al. | Jan 2001 | B1 |
6256748 | Pinson | Jul 2001 | B1 |
6282610 | Bergsten | Aug 2001 | B1 |
6330687 | Griffith | Dec 2001 | B1 |
6425052 | Hashemi | Jul 2002 | B1 |
Number | Date | Country | |
---|---|---|---|
20020035667 A1 | Mar 2002 | US |