Patent Application
20050182801
This invention relates to data storage library systems, and, more particularly, to reconciling of data in redundant data storage libraries to maintain synchronization of the data for the same volumes between the redundant data storage libraries.
Data storage libraries maintain sets of volumes, each of which can contain data, as backup for information handling systems. Reliability is enhanced by having redundant data storage libraries, each separately maintaining the same volumes.
An example of redundant storage libraries is a peer-to-peer arrangement of International Business Machines Corp. Magstar Virtual Tape Servers. The arrangement is a largely self-managed set of data storage libraries and directors which can provide access and storage for the same data volumes in both libraries (i.e. Peer-to-peer, or P2P). If an operation to recall a data volume from one virtual tape server library fails, then the volume may still be recalled from the other virtual tape server library. Therefore, when a host system writes or updates data that is to be stored, the data will be saved at both virtual tape servers. The data may be supplied at one time to both virtual tape servers, or, in a peer-to-peer arrangement, the data is written to one virtual tape server and subsequently copied from the one virtual tape server (data storage library) to the other virtual tape server (data storage library). Until it is copied, the data of the volume is likely to be different in the one data storage library as compared to the volume of the other data storage library.
Data is transferred to the redundant data storage libraries by at least one director. Reliability is further enhanced by employing a plurality of directors. A function of the directors is to insure that the data of the volumes is synchronized between the redundant data storage libraries, typically comprising two data storage libraries, although additional redundant libraries may be employed. Synchronization of the volumes may be maintained using tokens, which describe at least one version-related characteristic of the volumes.
Tokens allow a comparison of the tokens to be made to determine whether the data of the volumes are synchronized, and if not, which data is more recent. The tokens may have different version-related characteristics, such as different time stamps, and a token representing one copy may have a flag that indicates whether a copy is required, or that the data is inconsistent. Based on an examination of tokens, a director copies the most recent data of a volume from one data storage library to another data storage library, to synchronize the data. An example of a director is an IBM Virtual Tape Controller (VTC) which transparently transfers data and executes commands from one or more host systems with respect to data storage libraries. A plurality of directors may be employed in a Virtual Tape Server P2P system, such as 4 or 8 directors.
The synchronization process can take place immediately or be deferred. For example, a Wall Street firm may desire a high peak host input/output during trading hours, and choose to defer the copying function until after hours. If deferred, the number of volumes to be copied may become large. In addition, should one of the data storage libraries fail or be offline, the director will operate in deferred mode. The process for synchronizing the data of redundant data storage libraries from an extended deferred mode where there are a large number of volumes is called bulk reconciliation. To do a bulk reconcile, a director requests all of the tokens from each data storage library, compares the tokens of all the volumes, and reconciles the volumes that need it. This can be a lengthy process, taking up to about 2 seconds per volume. A data storage library can support a vast number of volumes (although typically only a fraction of the volumes would need reconciliation) so that the bulk reconcile could take hours to complete.
The present invention comprises a data storage system, a director, a computer program product, and a method for synchronization reconciling data to be maintained in redundant data storage libraries, each data storage library having tokens, each token describing at least one version-related characteristic of an associated volume of data stored by the data storage library. The directors are capable of communicating with the redundant data storage libraries.
At least one of the directors establishes a reconcile set of the tokens, comprising tokens of the redundant data storage libraries describing dissimilar version-related characteristics of a same associated volume. A plurality of subsets of the reconcile set of tokens are established based on a first predetermined criteria; and allotted to directors to reconcile the data of volumes of the subsets of the reconcile set of tokens. Thus, the reconcile is split between the plurality of directors.
In one embodiment, the first predetermined criteria is based on the volume identifier of the associated volume. An example is to select subsets based on the last binary digit of the volume identifier, i.e. odd and even numbered subsets. Alternatively, the volume identifiers may be selected to be split into upper and lower halves.
Additionally, in another embodiment, the directors reconciling the data of volumes, conduct the reconcile of the data of the volumes associated with the allotted subset of tokens, in a sequence based on a second predetermined criteria. In one embodiment, the second predetermined criteria is based on the volume identifier of the associated volume, for example, if multiple directors are allotted odd numbered subsets to reconcile, one or more conducts the reconcile in numerical sequence, and another or others conduct the reconcile in reverse numerical sequence, as the second predetermined criteria.
In a further embodiment, a director, when reconciling the data of volumes of a subset of the reconcile set of tokens, additionally senses that tokens for a volume are being re-read and the data of the volume associated with the re-read tokens has been reconciled, and, ceases reconciling the data of the volumes associated with the tokens of the subset.
Still further, the director, in response to ceasing reconciling the data of the volumes associated with the tokens of the subset, reconciles the data of volumes of the reconcile set of tokens.
For a fuller understanding of the present invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings.
This invention is described in preferred embodiments in the following description with reference to the Figures, in which like numbers represent the same or similar elements. While this invention is described in terms of the best mode for achieving this invention's objectives, it will be appreciated by those skilled in the art that variations may be accomplished in view of these teachings without deviating from the spirit or scope of the invention.
Referring to
Referring additionally to
A virtual tape server data storage library has a nonvolatile cache memory 40, which stores active data and comprises virtual data storage drives. When a volume is initially accessed, the library controller 30 identifies the data storage media 18 having the volume, or if the volume is to be written, selects a data storage media for the logical data volume. The library controller then operates the robot 22 to access the data storage media from the storage shelf and delivers the data storage media to a data storage drive 35. The volume is maintained in cache memory 40 for access by the host system, and for immediate future access, if needed. If not needed, eventually, the volume is deleted from cache memory. If the volume is written or updated, the data storage media 18 for the volume is again accessed and the written or updated volume is written to the data storage media.
The volumes are typically accessed using an identifier of the volume, which may comprise a binary identifier, called volume identifier or “VOLID”.
In
Redundant data storage libraries may comprise libraries in different locations for the purpose of disaster recovery. In the example of
The directors 71-74 are typically partitioned to conduct typical storage work for the redundant libraries. For example, volume identifiers (VOLID) are partitioned between the directors. In another example, the data storage drive addresses are partitioned between the directors. Thus, the director having the partitioned address responds to a host command regarding the partitioned VOLID or drive address.
Referring to
The VOLID 81 is the volume identifier of the volume of data that the token describes.
The flag 83 represents that copying is required. As discussed above, when a volume changes, it changes on one data storage library first and then is copied to the other data storage library. This flag indicates that the volume has changed and now needs to be copied to the other data storage library(ies).
The flag 84 represents that the data of the associated volume is inconsistent, meaning that the data is not the latest copy of the volume, and that the data on the other data storage library should be more up to date (and probably has the copy required flag set).
The flags 83 and 84 may comprise a property attribute section of the token.
The data level indication 85 is a timestamp that is incremented each time that the associated volume changes. This helps to keep track of which data storage library has the valid volume data.
Section 87 indicates other attributes of a volume. The attributes of volumes in a data storage library can be employed as an added level of data management. These attribute fields in the token may be used similarly to the data inconsistent and data level fields to assist in keeping the attribute information in synchronization.
Referring to
From a comparison of the tokens of all the volumes, the directors establish 91 a reconcile set of tokens, comprising tokens of the redundant data storage libraries having dissimilar version-related characteristics of a same associated volume. The bulk token read allows the directors to filter out and create a list of all volumes that need to be reconciled.
Alternatively, the tokens can be read individually and compared, but a bulk token read is fast as compared to reading all tokens individually.
In step 93, the director(s) establish a plurality of subsets of the reconcile set of tokens based on a first predetermined criteria. In one embodiment, the first predetermined criteria is based on the volume identifier (VOLID 81) of the associated volume. An example is to select subsets based on the last binary digit of the volume identifier, i.e. odd and even numbered subsets. Alternatively, the volume identifiers may be selected to be split into upper and lower halves. This splits the reconcile set of volumes to be reconciled into two different subsets 94 and 95. Additionally, further subsets may also be established. For example, the volume identifiers may be split into quartiles, or the odd and even sets may also be split into upper and lower halves.
In step 96, the subsets 94, 95 are allotted to different directors 71-74 to reconcile the data of volumes of the subsets of the reconcile set of tokens. Thus, the reconcile is split between the plurality of directors 71-74. The number of directors must be at least two for the split, and may comprise any number in excess of one. The allotting of the subsets of tokens may be accomplished in any of several ways. One director may establish the subsets of tokens, for example, as ranges of volumes, and communicates separate subsets to other directors. One director may establish the subsets of tokens and communicates all the subsets to the other directors, and each director selects its predetermined reconcile subset. The full reconcile set may be provided to all of the directors, and each director selects its predetermined reconcile subset. Alternatively, various combinations may be employed, such as dividing the directors into groups, each group having a “master” director allot or communicate subsets to the other directors a group. Further, it is not necessary that all of the directors of a system conduct the reconcile. A “plurality of directors” may comprise one grouping out of all of the directors.
The established set and subsets of tokens may comprise, for example, a listing of volumes, ordered, for example, alphabetically or, in binary form, numerically.
In one embodiment, the directors conduct the reconcile of the allotted subsets of tokens.
Additionally, in another embodiment, multiple directors 71-74 are allotted the same subset of volumes to reconcile. For example, directors 71 and 72 are allotted the even numbered VOLID subset 94, and directors 73 and 74 are allotted the odd numbered VOLID subset 95, to reconcile between data storage libraries 14 and 15.
In this embodiment, the directors 71-74 reconciling the data of the volumes, in step 97, conduct the reconcile of the data of the volumes associated with the allotted tokens, in a sequence based on a second predetermined criteria. The allotting of the subsets of tokens may be accomplished in any of several ways. Each of the directors may be programmed with its second predetermined criteria, or pattern; may be programmed with all of the criteria, and select the criteria to employ; may be provided with the criteria with the allotment of a subset; etc. In one embodiment, the second predetermined criteria is based on the volume identifier of the associated volume. For example, if multiple directors are allotted odd numbered subsets 95, to reconcile one or more of the directors conducts the reconcile in numerical sequence, shown by arrow 98, and another or others conduct the reconcile in reverse numerical sequence, shown by arrow 99, as the second predetermined criteria. In another example, if multiple directors are allotted the reconcile of even numbered subsets 94, one or more of the directors conducts the reconcile in numerical sequence, shown by arrow 100, and another or others conduct the reconcile in reverse numerical sequence, shown by arrow 101, as the second predetermined criteria. Thus, the reconcile is further split among the directors 71-74.
The directors that are conducting the reconcile continue reconciling the data of the allotted subset 94 or 95, and will, at some point, reach the halfway point 105, 106. In one embodiment, the director completes the bulk reconcile at that point.
In a further embodiment, a director 71-74, when reconciling the data of volumes of a subset 94, 95 of the reconcile set of tokens, may continue beyond the halfway point 105, 106. In doing so, at some point, in step 107, the director will read a token that has already been read by another director which has reconciled the data of the associated volume. When a volume is reconciled by one director, the token is updated on each data storage library to clear token fields 83, 84, 85, 87 that would indicate a reconcile is required. However, the other directors are not necessarily notified when volume tokens have changed and the volume has been reconciled. Therefore, a director should, in step 107, do an individual read of the volume tokens before attempting to reconcile it to be sure to avoid wasting time doing work that another director has already completed. By doing the re-read of a token 80, the director additionally senses that tokens for a volume are being re-read and the data of the volume associated with the re-read tokens has been reconciled, and, the director 71-74 ceases reconciling the data of the volumes associated with the tokens of the subset 94, 95.
Re-reading the tokens takes only a fraction of the time needed to fully reconcile a volume, so once the directors start to overlap in the volumes they reconcile, the process finishes quickly.
Still further, in another embodiment, the director 71-74, in step 108, in response to ceasing reconciling the data of the volumes associated with the tokens of the subset 94, 95, verifies the reconcile of the data of volumes of the full reconcile set of tokens. For example, if directors reconciling allotted subset 94 complete the reconcile before the directors reconciling allotted subset 95 complete their reconcile, one or more of the directors completing the reconcile may then verify the reconcile of the full set and reconcile the data of any unreconciled volume. For example, each director may be provided with the full reconcile set in the form of an alphabetical listing of volumes. As the director reconciles the data of a volume, the volume is removed from the list. Thus, after the reconcile of the subset is completed, the director has a list that is still sorted alphabetically and contains only the volumes that this director has not reconciled. At this point, in step 108, the director starts at the top of the list, confirming that each volume has been reconciled, and reconciling the volume if it has not been reconciled. The order of the second pass may be alphabetical.
The result is a speed-up of the bulk token reconcile process.
The illustrated components of the data storage system(s), directors, and data storage libraries of
While the preferred embodiments of the present invention have been illustrated in detail, it should be apparent that modifications and adaptations to those embodiments may occur to one skilled in the art without departing from the scope of the present invention as set forth in the following claims.
