Database utilities

Details Database utilities Database utilities

2000-04-13

2003-09-16

Shah, Sanjiv (Department: 2172)

Data processing: database and file management or data structures

Database design

Data structure types

C707S793000

Reexamination Certificate

active

06622142

ABSTRACT:

FIELD OF THE INVENTION
This invention relates to databases and database management systems and, in particular, to hierarchical database maintenance.
BACKGROUND OF THE INVENTION
Ideally, when data is stored in a database, it should be stored in physical proximity to other data to which it is related. Such proximal storage will reduce disk traffic and I/O access frequency. Over time, however, as data is deleted and added to the database, data that should be physically proximal or “clustered” becomes dispersed across the database and storage vehicles (DASD, for example) on which the database is resident.
Some database systems, such as IBM's Information Management System (“IMS DL/I” or alternatively, “IMS”), allow construction of data sets with free space distributed through the storage space. IMS provides that ability to specify that a portion of each block or control interval be reserved as free space, during the initial load or reorganization of a database. Every n-th block may also be reserved in entirety. There are two free space parameters that specify the percentage of free space for each block and the other specifies the frequency of completely free blocks.
Free space can be helpful or harmful. It will increase the amount of disk space required and may result in extra I/O's. The challenge is to allocate the right amount of free space during database design so that disk space is minimized while the likelihood of fitting additions in the optimum block is maximized. The volume of additions must be estimated as well as the distribution of those additions. Too much free space is an inefficient use of resources, and too little results in increases in seek time and increased I/O operations.
Databases express relationships between units of data. In a hierarchical database system, such as IBM's IMS, data is organized in a tree-like structure. Each unit of data is known as a segment and related segments are together known as a record. From a root segment, all other segments in the record bear a direct or indirect subordinate relationship. The root segment of a record is established by the database description or definition process (“DBD”). A segment which depends immediately from the root is a child segment and a child segment may be a parent to segments further from the root.
Over time, databases tend to enlarge unevenly so that some groups or “clusters” of related data increase in population more quickly than others. When data is inserted in an IMS data base, IMS uses a documented strategy that tries to place a segment to be inserted as close as possible to segments to which it is related. IMS first tries to place the segment into the block where related segments reside. If that is not possible, IMS tries to place the segment at least in the same track as related segments. If that is not possible, placement in the present, previous or next cylinder is attempted, and so on until it has searched for room both ahead of and behind the placement area. The available placement area is defined by a “SCAN cylinders” statement specified when the data base is generated during the DBD process. If still there is no available room, the segment is placed at the end of the data set in an area known as “overflow.” The overflow area is not contiguous with the root addressable area (“RAA”). If overflow becomes full, IMS will attempt to place the segment anywhere in the data base that room can be found. If there is insufficient free space early in the placement process, data becomes physically dispersed from the data to which it should be proximal. As data becomes dispersed, the read disk head must travel further to access that data and wait longer to complete the random seek on a particular track. Consequently, periodic rearrangement of the no longer clustered data in the database can result in significant improvement in database performance including increased storage efficiency and improved operational speed. Such rearrangement is known in the art as “reorganization.”
Basic IMS access techniques such as Hierarchical Sequential Access Method (HSAM) use sequential access to find a particular segment. The access request starts at the first root, then examines each root sequentially until the destination root is found and then searches up the tree according to certain rules until the target segment is found. Later IMS access techniques developed as part of IMS Version II introduced the hierarchical direct (HD) access methods. Hierarchical direct access methods such as the Hierarchical Indexed Direct Access Method (HIDAM), for example, allow indexed access to any root segment based upon its “key” to its offset from the beginning of the data set to the prefix of the root segment of the target record. This requires that a segment in an HD database never move within a dataset until the data base is reorganized.
Even though physical adjacency between logically related segments improves database efficiency, the functional or logical relationship between segments in an HD access IMS database is not expressed through the physical adjacency of those segments in the database. The segments within a data base record in an HD IMS data base are connected using four-byte Relative Byte Address pointers (“RBA”). A RBA pointer is a four-byte field in a segment that designates the starting position of the destination segment relative to the beginning of the dataset. Fixing segment location makes it feasible to use pointers from one segment to other specific segments in other data bases or partitions and from secondary indexes. Pointer use in segments is also valuable within a data base to connect a parent segment to the first or first and last occurrence of each segment type. Pointers can also be used to establish secondary indexes through which an alternative organizational hierarchy perspective or an entry point for the record alternative to the root can be constructed.
Logical relationships can be established to logically link two segments which exist in separate physical databases, partitions or data sets. A logical child is used to construct the logical linkage between the two segments intended to be related. Multiple logical relationships can be constructed to create a hierarchical structure consisting of segments from multiple physical databases to create an alternative logical view of related data which can be seen by an application as a hierarchical database.
In the two segments to be related, the logical child has two parents; a physical parent and a logical parent. The leftmost field in the logical child contains the concatenated key of the logical parent that gives a symbolic address for the logical parent. An optional direct RBA pointer can be contained in the segment prefix. Thus, if an access request seeks the logical parent, but knows only the location of the physical parent, the path to the logical child (which is the child of the physical parent) is taken where, upon arrival at the logical child, the address of the logical parent is found through the key or pointer in the logical child.
Thus, many useful, logically-ruled organizational structures are dependent upon pointers amongst and between data elements to maintain logical interrelationships and indexes which, although they differ from the physical relationships of the data, depend for their continuance upon the awareness of the physical siting of any data into which pointers direct the process flow. Further, pointers allow entry to a data base at any level of the hierarchy or any instance of a segment type without traversal of the hierarchical path. If a data segment which had been pointed to by the relative byte pointer in another segment is physically moved, established secondary indexes and logical relationships are destroyed unless the new location of the moved target data can be determined. Consequently, two countervailing trends contend in IMS reorganization. The need for operational efficiency dictates periodic reestablishment of physical data clustering. But, because reorganization moves data to reestablish physical groupin

Affiliated with

Also associated with

Rating

Database utilities does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Database utilities, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Database utilities will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3039324