Electrical computers and digital processing systems: memory – Storage accessing and control – Control technique
Reexamination Certificate
1999-06-16
2001-06-19
Kim, Matthew (Department: 2186)
Electrical computers and digital processing systems: memory
Storage accessing and control
Control technique
C709S248000, C707S793000, C711S112000
Reexamination Certificate
active
06249849
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to storage of data on rewritable data storage media which is accessible in data storage libraries, and, more particularly, to updating the redundant data stored in a plurality of data storage libraries.
BACKGROUND OF THE INVENTION
Data processing systems comprising at least one host typically require a large amount of data storage. If the data, typically stored as a data volume, is not immediately required by the hosts, for example, if the data volume is infrequently accessed, the storage of the data volume may be on removable rewritable data storage media, such as magnetic tape or optical disk, and the data volumes may be written and or read by means of a data storage drive.
The data storage drive is typically coupled to the host, or processing unit, by means of a peripheral interface in which commands are directed only from the processing unit to the data storage drive, and the data storage drive responds to those commands, performing the commanded functions. No commands can be sent by the data storage drive to the coupled processing unit. Typically, the commands are performed by a device controller.
Data processing systems having multiple hosts require larger amounts of data storage than can be efficiently handled by single data storage drives.
Data storage libraries typically provide efficient access to large quantities of data volumes stored in removable data storage media, the media stored in storage shelves which are accessed by robots under the control of robot controllers. Due to the large amount of stored data, typically, a plurality of hosts make use of the same data storage library, and a plurality of data storage drives are included in the library to allow access by the hosts. A library manager, which may comprise the same processor as the robot controller, typically tracks each data volume and the data storage media on which it is stored, and tracks the storage shelf location of each data storage media.
Herein, a library manager, either with or without the robot controller, is defined as a “controller” for the data storage library, as is the “controller” for a data storage device as discussed above.
An example of a data storage library system for redundantly storing and accessing data volumes stored on removable data storage media in a plurality of data storage libraries is described in the incorporated coassigned K. F. Day III et al application. The library controller of each library provides an updatable synchronization token directly associated with each data volume. A plurality of directors are provided, each separate from and coupled to the hosts and each separate from and coupled to each data storage library. Each director responds to separate, partitioned data storage drive addresses addressed by the hosts. The responding director supplies each data volume supplied from a host to all of the data storage libraries, and updates each synchronization token directly associated with the supplied data volume. Thus, the directors store duplicate copies of the data volume in the data storage libraries without involvement by the host. In most data processing applications, it is critical to access the most current data. Hence, the currency of the data volumes are each tracked by means of the directly associated synchronization token, and the synchronization token is not tracked by the host.
In the instance where one copy of the data volume is down level and the corresponding synchronization token has not been updated, the responding director compares the tokens from all the libraries to identify the down level data volume, so that the erroneous down level information is not transferred to hosts. The responding director may then update the most current data volume and the corresponding synchronization token and the previously down level volumes are also updated.
If, as a data volume is being updated at one of the libraries, another of the libraries is or goes temporarily out of service, or out of communication with the directors, the data volume at that library cannot be updated. Thus, when the copy of the data volume on the operating library is updated, that volume is exposed, in that one copy will be more current than another. Later access to the data, and conducting the comparison of the tokens, will note the lack of consistency of the data, but if the library having the most current copy is then temporarily out of service, or out of communication with the directors, there will be no ability to note that the accessed data is down level so long as the library is out of service, etc. Further, the down level data may be further updated without knowledge that it is down level, and its token updated, so that the updated data may be lost forever.
SUMMARY OF THE INVENTION
An object of the present invention is to allow a data volume to be updated only when the various updating levels of the volume will be identified to all of the directors or indicated in the token on the libraries with the other copies.
Disclosed are a data storage library subsystem, wherein the data storage libraries are coupled to a plurality of directors, and a method for identifying most recent redundant copies of identifiable data volumes stored in a plurality of data storage libraries, wherein at least one of the redundant copies may be initially updated with respect to other of the redundant copies, the identifiable data volumes accessible in the plurality of data storage libraries. The one of the data storage libraries first updating the copy is called the advancing library, and the later of the data storage libraries updating the copy is each called a lagging library. The identifiable data volumes are each directly associated with an updatable synchronization token, the synchronization token indicating the relative update levels of the directly associated redundant copies. Due to unavailability of a library or of communication between a library and the directors, the copies and the directly associated tokens may be unavailable to the directors at various times.
The present method, in response to a command to update a data volume, attempts to set, by an initiating one of the directors, an “inconsistent” flag to the current token directly associated with potentially down level copies of the data volume on each lagging library. Then, the initiating director receives notification of any successful setting of the current token “inconsistent” flag by the lagging libraries, within a predetermined time-out period. Upon the initiating director, as the result of the lagging library notification receiving step, failing to receive the notification from any of the lagging libraries, the initiating director forwards to each other director available to the initiating director, a request to set the current token “inconsistent” flag for the volume at any lagging library from which no notification has been received. Then, the initiating director receives notification from the other directors of any successful setting and of failure to set the current token “inconsistent” flag of the lagging libraries for which no notification had been received by the initiating director, within a predetermined time-out period. Upon all of the other directors responding to the initiating director in the flag notification receiving step, and upon the other directors indicating failure to set the “inconsistent” flag at the lagging library, the initiating director lists the volume and a new token as an entry in a “hot token” list at the initiating director. The initiating director allows updating only the data volume for which the initiating director has received the notification of posting the “inconsistent” flag for all of the lagging libraries, or the new token has been listed in the “hot token” list at each director.
Additionally, in each of the other directors, upon receiving from the initiating director the forwarded request to set the current token “inconsistent” flag for the volume at any lagging library from which no notification has been received, the director attempts to set th
Day III Kenneth Fairclough
Dewey Douglas William
Bataille Pierre Michael
Holcombe John H.
International Business Machines - Corporation
Kim Matthew
Sullivan Robert M.
LandOfFree
“Fail-Safe” updating of redundant data in... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with “Fail-Safe” updating of redundant data in..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and “Fail-Safe” updating of redundant data in... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2502457