Electrical computers and digital processing systems: memory – Storage accessing and control – Control technique
Reexamination Certificate
1998-12-09
2001-07-10
Verbrugge, Kevin (Department: 2185)
Electrical computers and digital processing systems: memory
Storage accessing and control
Control technique
Reexamination Certificate
active
06260125
ABSTRACT:
The present invention relates to clustered computer systems and, more particularly, to methods for mirroring disk drives within a network.
BACKGROUND OF THE INVENTION
The increased reliance by business on On-Line Transaction Processing and Decision Support Systems has increased the demand for high availability systems since these systems are critical to the functioning of day-to-day activities in many businesses. These systems are not only essential for the support of normal daily operations, but they also store critically important customer and corporate data. Continuous availability is no longer an ideal; it is a necessity for many companies. Longer work days, expansion into new markets and customer demand for more efficient service create an expanded requirement for increased system availability. Users are demanding a means of ensuring very high availability of their applications and the access to data that permits them to accomplish their tasks and provide the highest levels of customer service. Interruption of workflow due to system failure is expensive and it can cause the loss of business. The need to increase computer system availability is becoming one of businesses key concerns.
Implementation of client/server computing is growing throughout today's businesses—for key business applications as well as electronic mail, distributed databases, file transfer, retail point-of-sale, inter-networking, and other applications. It is possible for companies to gain competitive advantages from client/server environments by controlling the cost of the technology components through economies of scale and the use of clustered computing resources. There is a boost in productivity when businesses have high availability and easy access to information throughout the corporate enterprise.
Computer system availability and reliability are improved when multiple servers are utilized together with a “fail-over” scheme such as provided by NCR Corporation's LifeKeeper product. In such a system, should one server fail the functions and applications associated with the failed server are transferred to one or more of the remaining operational or standby servers.
An important component of a high-availability client/server system is a reliable, fault-tolerant data storage system. In some networked or “clustered” multiple server arrangements, the physical data storage system may be a shared RAID (Redundant Array of Inexpensive Disks) disk array system, or a shared pair of disk drives or disk arrays operating in a mirrored arrangement.
A computer system including multiple servers and a pair of shared disk drives is shown in FIG.
1
.
FIG. 1
provides a diagram of clustered or networked computers having a primary server
101
and a secondary server
103
in a fail-over pair arrangement. Primary server
101
is the preferred application server of the pair, and secondary server
103
preferably provides fail-over protection for the primary server. The primary and secondary servers are coupled through a network bus system
105
to a plurality of client computers
107
though
109
. The primary and secondary servers
101
and
103
each shares access to a pair of disk storage devices
111
and
113
. Disk storage devices
111
and
113
are SCSI (Small Computer Systems Interface) disk drives or disk arrays connected to servers
101
and
103
through a pair of SCSI busses
115
and
117
.
Disk storage devices
111
and
113
are two equal-capacity storage devices that mirror each other. Each storage device contains a duplicate of all files contained on the other storage device, and a write or update to one storage device updates both devices in the same manner. In the event that either storage device fails, the data contained therein remains available to the system from the operational mirror storage device.
In other client/server arrangements, the physical storage devices for the primary and secondary servers may be separate, non-shared, physical storage devices. A network based file system volume replication scheme where the contents of the file system stored on a primary physical storage media are also copied to a secondary physical storage media is commercially available from NCR Corporation, assignee of the present application, under the product name “Extended Mirroring”.
Most disk mirroring procedures utilizing shared drives will write or update both mirror drives synchronously. In systems performing disk mirroring with non-shared drives over a network, writes directed to a primary drive are received and forwarded to the secondary mirror drive. Upon receipt from the secondary drive of an acknowledgement signal indicating a successful update of the secondary drive, the write to the primary is completed. Although the updates to the primary and secondary drives do not occur simultaneously, this process will also be referred to as a synchronous write in the following discussion.
Disk mirroring applications that provide synchronous writes are inherently limited in what they can do. Being synchronous, each writer blocks until the write is complete. This limits performance and flexibility.
OBJECTS OF THE INVENTION
It is therefore an object of the present invention to provide a new and method and mechanism for mirroring storage devices over a network.
It is another object of the present invention to provide such a method for performing asynchronous updates to mirrored drives.
It is yet another object of the present invention to provide a new and useful method and mechanism for performing asynchronous updates to mirrored drives through implementation of an asynchronous write queue.
It is still a further object of the present invention to provide a new and useful method and mechanism for performing log-based reconstruction of a mirror drive in a network disk mirroring application.
It is also an object of the present invention to provide a new and useful ability to “check point” source and target volumes within a disk mirroring application.
SUMMARY OF THE INVENTION
There is provided, in accordance with the present invention, an asynchronous disk mirroring system for use within a network computer system. The disk mirroring system includes a first storage volume connected to receive write requests from the computer system; a write queue connected to also receive the write requests directed to the first storage volume; and a second storage volume connected to the write queue to receive the write requests. The write queue operates to delay the time of receipt of the write requests by the second storage volume.
In the described embodiment of the invention, the write queue comprises a plurality of serially connected write buffers, wherein the write requests pass through the write queue in a first-in, first-out (FIFO) sequence. Additionally, a mechanism is provided to stop the transmission of additional write requests to the first storage volume and the write queue when the plurality of serially connected write buffers are full. A log file connected to receive the delayed write requests from the write queue is also included within the disk mirroring system to provide for log-based mirror reconstruction and check-pointing of the mirrored volumes.
The above and other objects, features, and advantages of the present invention will become apparent from the following description and the attached drawings.
REFERENCES:
patent: 5513314 (1996-04-01), Kandasamy et al.
patent: 5537533 (1996-07-01), Staheli et al.
patent: 5555371 (1996-09-01), Duyanovich et al.
patent: 5701516 (1997-12-01), Cheng et al.
patent: 5737514 (1998-04-01), Stiffler
patent: 5742792 (1998-04-01), Yanai et al.
patent: 5764903 (1998-06-01), Yu
patent: 5958070 (1999-09-01), Stiffler
patent: 6006307 (1999-12-01), Cherukuri
NCR Corporation
Stover James M.
Verbrugge Kevin
LandOfFree
Asynchronous write queues, reconstruction and check-pointing... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Asynchronous write queues, reconstruction and check-pointing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Asynchronous write queues, reconstruction and check-pointing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2513120