training idms

Post on 07-Apr-2015

361 Views

Category:

Documents

7 Downloads

Preview:

Click to see full reader

TRANSCRIPT

IDMS

Integrated Database Management System

2

Topics Covered Introduction to DBMS Logical and Physical Database Structure Record Characteristics Set Characteristics IDD and DDDL Data Description Language (DDL) Data Manipulation Language (DML) Recovery & Restart of Database Locking Facilities

3

Introduction to DBMS

4

TYPES of DBMS

Inverted List (e.g. Datacom DB)

Hierarchical (e.g. IMS)

Data records are typically connected with embedded pointers to form a tree structure. Each node (except root) can have one and only one parent

5

Continue …..

• Network (e.g. IDMS)The database forms a mesh structureEntity-Relationship is implemented using Record type and Set.

• Relational (e.g. Oracle, Sybase, etc)Entity-Relationship is implemented in the normalized form. Data represented in the form of rows and columns (two dimensional table)

6

Logical & Physical Database Structure

7

A database can be viewed as consisting of Entities. An entity is an object about which we are storing data. Every entity has attributes. e.g. Employee is an entity which has attributes like, emp-code, emp-name, emp-address, emp-sex, emp-dept etc. There exists relationship between different entities. We can represent the relationship between our entities using an entity-relationship (E-R) diagram as below.

LOGICAL DATABASE STRUCTURE

8

SALARY

BONUS

TEAM

PLAYER GAME

E-R Diagram

9

The example discussed here:

Taken from a typical American baseball league.

A team has many players but at a given time, a player is attached to only one team.

A team plays many games. Games can be further identified by Game-Home and Game-Away.

THE EXAMPLE

10

A player plays many games and same way a game is played by many players.

A player has many salary records since his salary (contracts) may have undergone changes during a season. Similarly, a player might have many bonuses awarded to him.

Team, Player, Game, Salary and Bonus represent Entities.

Continue …..

11

Attributes of Team could be Team-Name, Team-City, Team-Address etc.

The relationship between entities is shown by line diagram. We use cardinality symbols to represent type of relationship. The crow-foot end of the line diagram represents Many type and single line represents One type of the relationship.

Continue …..

12

For example, consider relationship between Team and Player. Team end has a single line whereas Player end has a crow-foot. This represents a One-to-Many relationship. It means a team has many players. But a player is attached to only one team.

Note that relationship between Player and Game is Many-to-Many. i.e a game has many players and a player plays many games.

Continue …..

13

In IDMS, entities are implemented using Record types and relationships are implemented using Set types.

IDMS does not allow direct implementation of many-to-many relationship.

Continue …..

14

JUNCTION RECORD

It is a way of implementing many-to-many relationship in IDMS. It is a record type having attributes that are common (intersection) to both record types participating in many-to-many relation.

Thus, many-to-many relation between player and game can be resolved as below.

15

NOTE:

Here, referring to our earlier example of American baseball league, we had many-to-many relationship between entities player and game.

PLAYER GAME

POSITION

Continue …..

16

That is to say, a player has played in

many games and conversely, a game

has many players in it. To resolve this

M:M relationship, we introduce a

junction entity, Position record.

Position is called Junction record

because it represents junction or

intersection between Player and Game

record.

Continue …..

17

A player has many positions assigned to him for different games. But for a game he has a fixed position. Main attribute of the record is “player-position” for a game. It thus resolves many-to-many relationship between player and game. In other words, we can easily move from a player record occurrence to a game record occurrence using the position record occurrence.

Continue …..

18

BACHMAN DIAGRAM

Most installations use a line with an arrow at one end to show each one-to-many relationship instead of using crow’s foot notation. Such a diagram is called Bachman Diagram. Here, the end of the relationship without an arrow always represents “One” side of the relationship and the end with arrow always represents “Many” side of the relationship.

Our earlier E-R diagram is represented using Bachman diagram as:

19

SALARYTEAM

PLAYER GAME

BONUS POSITION

BACHMAN DIAGRAM

20

IDMS database is divided into one or more areas. An area is defined as the major named subdivision of addressable storage in database.

An area is further subdivided into pages. Page is a smallest unit of data transfer between main memory and hard disk.

PHYSICAL DATABASE STRUCTURE

Database Area, Page and File

21

Each page stores control information and can store up to a maximum 255 database records.

Database areas can be of different page sizes. e.g AREA-A can be of page size 2048 bytes where as AREA-B of same database can be of page size 4096 bytes.

Continue …..

22

Page size for an area is decided based on the record types that will be residing within that area.

ALL record occurances of a

particular record type are stored

within the SAME area. But, an area may store occurances of

multiple record types. Hence this holds true for a database page as well.

Continue …..

23

Record types are assigned to areas by designer.

Areas are mapped to files.

Many or all areas can be mapped into one file if all areas have the same page size.

Each area can be mapped into a different file.

One area can be mapped into several files.

Continue …..

24

AREAS and FILES MAPPING

Page

Area

Logical Database

Physical Database

File

Block

25

Team Area

921 922 … 930

IDMS File 1

Position Area

1021 1022 … 1040

Salary Area

821 822 ... 830

IDMS File 2

Area and File Mapping…..

26

SALARYTEAM

PLAYER GAME

BONUSPOSITION

ADVANTAGES OF MULTIPLE AREAS

SALARY-AREAPOSITION-AREA

TEAM-AREA

27

Continue …..

As shown in the previous slide, our example of baseball database is divided into three physical areas namely Team-area, Position-area and Salary-area.

The advantages of dividing a database into separate physical areas are:

Processing efficiency

Records those are accessed together during most of the processing can be grouped together into same area for efficiency.

28

Continue …..

Security

One can restrict access to certain record types. e.g. salary-area can be restricted to only finance department.

Database recovery and backup

Database can be initialized, reorganized and backed up on area-by-area basis. Backup of most updated areas can be made more frequent than other areas.

29

Continue …..

Concurrent updating

A program can request exclusive use of an area and prevent other programs from accessing it concurrently.

30

CONCEPT of DB-KEY

Each Record occurrence stored in the database is assigned an unique numeric identifier, called Database Key (db-key). A record’s db-key consists of a 32-bit field that contains a 23-bit page number and an 8-bit line number. The page number identifies the page in which the record is stored and the line number identifies location of the record occurrence within the page.

The format of db-key is as below:

31

8-bits23-bits1-bit

(Not

Used)

Line NumberDatabase Page NumberSign Bit

32

RECORD TYPE v/s RECORD OCCURRENCE

It is important to make distinction between Record Type and Record Occurrences.

Record type is like a template. It describes the format of all occurrences of a given record type stored in database.

Record occurrence represents the smallest directly addressable unit of data. It consists of fixed or variable number of characters that are subdivided into units called Data Elements.

33

Example of Record Type versus Record Occurrences

Employee Record Type ……….

Emp-code

PIC 9(4)

Emp-name

PIC X(30)

Sex

PIC X

Emp-dept

PIC X(10)

Employee Record Occurrences …..…...

1000 ABC M TECH

1500 XYZ F SALES

34

Database consists of collection of record occurrences.

The physical records stored in a database consist of more than the data elements used by the application program.

IDMS also maintains information about relationships that exists between records.

Relationships are implemented by linking record occurrences together with pointers.

STRUCTURE of DATABASE RECORDS

35

Pointers contain addresses of related record occurrences and are stored along with the data portion of the record occurrences.

A record occurrence in database consists of two parts. Data portion and Prefix portion.

Continue …..

36

Data elements values are stored in data portion. Pointers to related record occurrences are stored in prefix portion.

Application programs deal only with data portion of the record occurrence. Whereas, IDMS system maintains the pointers in the prefix portion.

Continue …..

PREFIX Portion DATA Portion

Pointer 1

Pointer 2

Pointer 3

…..Data element 1

Data element 2

...

37

Schema is the logical definition of a database. A schema is a complete database description (all records types & record elements, set types, files and areas).

There is ONE and ONLY ONE schema for a given database.

SCHEMA and SUBSCHEMA

38

A subschema defines a subset of the schema. One or more subschemas can be associated with a schema. A load module is created for each subschema. At runtime, a program can access only one subschema. A program can not access IDMS database by referring to the schema.

A subschema is similar to a view in a relational database.

Continue …..

39

Even if a program requires access to ALL record types, data elements and ALL set types defined in the schema, the designer must still define a subschema that includes all the entities defined in the schema.

A subschema can restrict access to a program. For example, it may give only read-only access to an area, or it may not allow deletion of some record types.

Continue …..

40

Record Characteristics

41

The characteristics that apply to record types are:

Record Name

Record Identifier

Storage Mode

Record Length

Location Mode

Duplicates Option

Area Name

42

RECORD NAME

Each record type must be assigned a 1

to 16 character name that identifies the

record type. The name must begin with

an alphabetic character. The

application program must reference the

record’s name in DML (Data

Manipulation Language).

43

It is a number that serves as in internal

identifier for the record type. It is in the

range 100 through 9999. Each record

type must be assigned an unique record

identifier within the installation. DBAs

assign this number to each record type.

Application programs do not refer to

record type using this number.

RECORD IDENTIFIER

44

STORAGE MODE

It indicates whether record occurances of the record type are fixed or variable length and whether they are stored in compressed format.

Allowable codes are:

F (fixed length)

V (variable length)

C (compressed)

45

NOTE:

Compressed mode can be used along with fixed or variable length storage modes. For example, one can mention storage mode as ‘FC’ or ‘VC’. DBMS takes care of compression and decompression of data.

Continue …..

46

RECORD LENGTH

It is expressed in BYTES. It is the

actual data length for fixed-length

record or the maximum data length

for variable-length record.

47

LOCATION MODE

It defines the way record occurances are stored in the database.

Allowable location modes are:

CALC

VIA

DIRECT

48

In CALC mode, a particular data

element within the record is declared as

the CALC-key. At the time of storing

the record in the database, IDMS

system uses value of the data element

to calculate the page number for

storing a record. Records stored with a

CALC mode can be retrieved from disk

in a single access.

CALC Mode

49

IDMS system uses randomizing routine

to distribute records evenly over its area,

minimizing overflow conditions and

leaving space for adding new records.

For retrieving the record (stored with

CALC location mode), appropriate

value for the data element is moved to

the storage area in the application

program and then DML retrieval

function is executed.

Continue …..

50

VIA Mode

Records stored with VIA location mode

are stored near another database record. This mode is generally used for storing

member records on the same page

containing owner record or on a page

near their owner record. This mode tends to reduce disk accesses

needed to retrieve all the records of a set

occurrence.

51

Continue …..

CALC retrieval is not possible in

case of records stored with VIA

location mode. Generally the owner

record is assigned CALC location

mode for easy access of member

records.

52

DIRECT Mode

In DIRECT location mode, application

program explicitly specifies the page

into which the record should be stored.

To retrieve this record, programmer

must specify its database address, i.e

db-key of the record. This mode is less often used than

CALC and VIA mode.

53

DUPLICATES OPTION

It is specified only for record types that

are stored using CALC location mode. It

specifies whether records with duplicate

CALC-key values are allowed and if so,

how would they be stored in the

database.

Possible codes are as on next slides:

54

Continue …..

DN (Duplicates Not Allowed)

Record occurances with

duplicate CALC-key value will

not be accepted. IDMS will

give an error if application

program tries to store a record

with duplicate CALC-key.

55

DF (Duplicates First)

IDMS will store the record with

duplicate CALC-key value

Before any record in the database

that has matching CALC-key

value. When CALC retrieval is

made using CALC-key value,

newly stored record will be

retrieved first.

Continue …..

56

DL (Duplicates Last)

IDMS will store the record with

duplicate CALC-key value After

any record in the database that has

matching CALC-key value. When

CALC retrieval is made using

CALC-key value, newly stored

record will be retrieved last.

Continue …..

57

AREA NAME

It is the name of the area into which all record occurances of the record type are to be stored. e.g. CUST-AREA, SALARY-AREA, SALES-AREA etc.

58

Data Structure Diagram (DSD) for Documenting Record Characteristics

Record name

Record-id Storagemode

Recordlength

Location mode

Calc key or VIA set name Duplicates option

Area name

59

Example : DSD for Employee Record Type

EMPLOYEE

1000 F 150 CALC

EMP-CODE DN

HR-AREA

60

Set Characteristics

61

A Set Occurrence consists of one occurrence of owner record type and any number of member record occurrences.

WHAT IS A SET ?

A Set consists of an OWNER record type and one or more MEMBER record types.

SET OCCURRENCE ?

62

A set occurrence that consists only of an owner record occurrence and no member record occurrences is called Empty Set occurrence.

Walking a Set

Records in each set occurrence are physically linked together by pointers.

Accessing members by following the pointers from one record occurrence to the next is called Walking the Set.

Empty Set Occurrence

63

• Any record type can be member in one or more sets.

• Any record type can be owner of one or more sets.

• Any record type can be member in one set and owner in another set.

• An owner record can own the same member record in more than one set.

• A record cannot be both owner and member in the same set.

RULES & FEATURES FOR SET RELATIONSHIP

64

Office

Employee

Department

Office-EmployeeDept-Employee

Any record type can be member in one or more sets

65

Any record type can be owner of one or more sets

Order

Cust-Order

Customer

Parts

Cust-Parts

66

Any record type can be member in one set and owner in another set

Order

Cust-Order

Customer

Items

Ord-Items

67

An owner record can own the same member record in more than one set

Class

Teacher

Teach-Class Examiner-Class

68

DBA typically assigns characteristics to each set when defining the IDMS database in the schema DDL.

The characteristics are:

Set Name

Linkage Options

Membership Options

Order Option

Duplicates Option

Set Characteristics

69

SET NAME

A unique name must be given to each set

type in the database. The name can be

maximum of 16 characters. The set name

must be referenced whenever an application

program accesses records using that set

relationship. Usually it will be owner record

name followed by member record name e.g.

CUST-ORDER.

70

LINKAGE OPTIONS

It indicates the types of pointers that are used to implement the set. Pointers provide flexibility in accessing records in a set occurrence. Pointers are stored in the prefix part of the database record occurrences.

Available options are:

71

Continue …..

N (NEXT pointer)Each record in the set contains a pointer to the next record occurrence. This option allows to access member records only in the forward direction. Next pointer is mandatory for all sets. All other pointers are optional.

72

Continue …..

NP (NEXT and PRIOR pointer)

With this, in addition to Next

pointer, each record contains

pointer to prior record occurrence

in the set. This allows us to access

member records in both forward

and backward direction.

73

Continue …..

NO (NEXT and OWNER pointer)

With this, in addition to Next

pointer, each record contains pointer

to owner record occurrence in the

set. This allows us to access the

owner record directly from any

member record occurrence.

74

NPO (NEXT, PRIOR and OWNER pointer)

Each record in the set contains all three pointers. This option allows to access member records in both forward and backward direction and also allows to access the owner record directly from any member record occurrence.

Continue …..

75

Xansa Mumbai DATAPN

EMP-A DATA PN

EMP-B DATAPN

EMP-C DATAPN

Record Occurrences with Next, Prior & Owner Pointers

Office

Employee

Office-Employee

O

O

O

76

The Membership options specify how a member record may be connected to or disconnected from a set occurrence.

The option is defined in two parts.

MEMBERSHIP OPTIONS

First part is a Disconnect Option, indicating the way a record is disconnected from a set.

Second part is the Connect Option, indicating how a record is connected to a set.

77

It specifies whether a member record can

be later disconnected from a set once its

membership has been established.

Possible values are:

M (Mandatory)

The record cannot be disconnected

from a set unless that record is

deleted (erased) from the database

using the ERASE command.

Disconnect Option

78

O (Optional)

A record occurrence can be disconnected from a set. The record remains in the database. It can be connected to some other set occurrence. It is optional to use ERASE for such records.

Continue …..

79

It specifies whether or not a member record is automatically connected to a set occurrence when it is added to the database.Possible values are: A (Automatic)

Automatic means when a member record is inserted in database, IDMS will automatically connect it to all its owner records (provided currencies have been established for owners).

Connect Option

80

M (Manual)

Manual option specifies that after

inserting a record, programmer

must explicitly connect it to its

owner record by issuing

CONNECT statement.

Continue …..

81

MA - Mandatory Automatic

MM - Mandatory Manual

OA - Optional Automatic

OM - Optional Manual

Combinations for Membership Options

82

The order option specifies logical order in which member record occurrences are placed within a set occurrence. Options available are: FIRST

Each new member record occurrence is placed immediately after the owner record (in the next direction). This option achieves a member record in LIFO.

ORDER OPTION

83

LAST

Each new member record occurrence is placed immediately before the owner record (in the prior direction). This option achieves a member record in FIFO. Prior pointer is a must to specify this option.

NEXT

Each new member record occurrence is placed immediately after the member record occurrence that was last accessed (in the next direction).

Continue …..

84

PRIOR

Each new member record occurrence

is placed immediately before the

member record occurrence that was

last accessed within the set (in the

prior direction). Prior pointer is a

must to specify this option.

Continue …..

85

SORTED

Each new member record occurrence is placed in ascending or descending sequence, based on the value of designated sort-control data element (sort-key) in each record occurrence. When the record is placed into a set, DBMS examines the sort-key value in each member to find the logical position of new member record in the set.

Continue …..

86

This option is useful only in case where set is defined with order option as Sorted. It indicates the action to be taken when a duplicate sort-key value occurs.

DN (Duplicates Not Allowed)

Record occurances with duplicate sort-key value will not be stored in the set. IDMS returns an error code if program tries to store such a record.

DUPLICATES OPTION

87

DF (Duplicates First)

Record with duplicate sort-key

value is stored Before any

existing record in the set that has

matching sort-key value. Most

recently stored duplicate record

will be retrieved first.

Continue …..

88

DL (Duplicates Last)

The record occurrence with

duplicate sort-key value is stored

After existing record in the set

that has matching sort-key value.

Most recently stored duplicate

record will be retrieved last.

Continue …..

89

IDMS allows retrieval of records using an index.

In conventional sets, member records are chained together by pointers. In an indexed set, DB-key values of member record occurrences are stored in a specified order in one or more index records.

Adds flexibility to data retrieval and retrieval is made faster in some cases.

INDEXED SETS

90

Indexed sets are useful in cases where: Records need to be accessed using

alternate-key (apart from CALC key).

Walking very long sets and when only key values of a member record is required.

Member records need to be retrieved randomly using partial key value.

Continue …..

91

Indexed sets can be implemented using two database record types or a single record type and a system defined record type.

Continue …..

92

Player

Position

Player-Position-Index

(Position-Name)

Continue .....

User-defined Owner Record

93

Employee

System-defined owner record

(Employee-Name)

Continue .....

SR7

94

Index between Player and Position

records allows to access position

records using Position-name values.

System owned index on employee

implements alternate index on

employee-name.

Continue …..

95

SALARY

PLAYER

BONUS

TEAM

POSITION

GAME

SALARY-AREA

SALARY-AREA

TEAM-AREA

TEAM-AREA

TEAM-AREA

POSITION-AREA

DN

DN

DN

VIA

VIA VIA

CALC CALC

CALC40

70

25 50

80

701000

1100

1200

2000

25003000

F

F

F

F

F

F

PLAYER-BONUS GAME-POSITION

PLAYER-NAME

PLAYER-SALARY

GAME-PLAY-DATE

TEAM-NAME

PLAYER-SALARY

NPO

OA

NEXT

TEAM-PLAYER

NPO

OM

LAST

PLAYER-BONUS

NPO

OA

NEXT

PLAYER-POSITION

NPO

OA

NEXT

GAME-POSITION

NPO

MA

NEXT

TEAM-GAME

NPO

MM

NEXT

96

IDD & Data Dictionary Definition Language (DDDL)

97

IDD stores meta-data about all the data items in the database (The data values are not stored in IDD).

Stores information about users, application programs, files, record, data element, module (date routine, error handling routine) and application systems.

IDD is integrated with every software component provided in IDMS like ADSO, OLQ and Report Generator. Each software component accesses IDD to get information about data items and programs.

INTEGRATED DATA DICTIONARY (IDD)

98

The data dictionary definition language (DDDL) is used to create, update or delete entity occurrences in IDD.

DDDL provides five verbs to manipulate entities:

ADD to add a new entity.

MODIFY to change entity description.

DATA DICTIONARY DEFINITION LANGUAGE (DDDL)

99

DELETE to remove an entity.

DISPLAY to display an entity description.

PUNCH to copy entity information from IDD to a file.

Continue …..

100

ADD ELEMENT CITY-NAME ELEMENT DESCRIPTION ‘TEAM CITY’ PICTURE X(20) USAGE DISPLAY.

DISPLAY PROGRAM STATSEDIT.

DELETE PROGRAM NAME IS STATSEDIT.

Examples of DDDL

101

MODIFY ELEMENT CITY-NAME PICTURE X(25). MODIFY PROGRAM NAME IS STATSEDIT

VERSION IS 1LANGUAGE IS COBOLESTIMATED LINES 5000.

SIGNOFF. ( to exit from IDD )

Continue …..

102

Data Description Language (DDL)

103

Data Description Language (DDL) is used to define:

SCHEMAS

DBA describes the logical structure of the database by coding set of schema DDL statements. There is a single schema for a given database.

104

DEVICE-MEDIA CONTROL

LANGUAGE (DMCL) MODULES

Many aspects of the database’s

physical structure, such as disk

device assignments and page sizes

are described in DMCL module

definition.

105

SUBSCHEMAS

The views that individual

applications programs have of the

database are defined in

subschemas and described with set

of subschema DDL statements.

Any number of subschemas can be

defined for a given database.

106

The logical structure of a database is defined in a set of schema DDL statements.

Schema DDL statements define the following:

Schema name Database file names Database areas Data elements and records Sets

SCHEMA DDL STATEMENTS

107

The schema DDL is processed by the

schema compiler. The schema

compiler reads the schema DDL

statements and stores a description of

the schema in the data dictionary.

The data dictionary is the central

repository of information that IDMS

maintains about the databases under its

control.

Process

108

Schema

DDL Statements

Schema Compiler

Data Dictionary

The schema compiler can also copy from the data dictionary descriptions of records that have been previously defined and stored in the data dictionary.

Continue …..

109

ADD SCHEMA NAME IS DDSCHEMA VERSION 1

ADD FILE NAME IS CUST-FILE ASSIGN TO CUST.ADD FILE NAME IS ORDER-FILE ASSIGN TO ORDER. ADD AREA NAME IS CUST-AREA PAGE RANGE IS 505000 FOR 100 WITHIN FILE CUST-FILE.ADD AREA NAME IS ORDER-AREA PAGE RANGE IS 809000 FOR 100 WITHIN FILE ORDER-FILE.

Example Of Schema DDL

110

ADD RECORD NAME IS CUST-REC RECORD ID IS 3000 LOCATION MODE IS CALC USING (CUST-NO)

DUPLICATES ARE NOT ALLOWEDWITHIN AREA CUST-AREA.

02 CUST-NO PICTURE IS 9(6) USAGE IS DISPLAY.02 CUST-NAME PICTURE IS X(40) USAGE IS DISPLAY.

ADD RECORD NAME IS ORDER-REC RECORD ID IS 3100 LOCATION MODE IS CALC USING (ORDER-NO)

DUPLICATES ARE NOT ALLOWEDWITHIN AREA ORDER-AREA.

02 ORDER-NO PICTURE IS 9(7) USAGE IS DISPLAY.02 ORDER-STATUS PICTURE IS X(1) USAGE IS DISPLAY.02 ORDER-REQ-DT PICTURE IS 9(8) USAGE IS DISPLAY.

Continue …..

111

ADD SET NAME IS CUST-ORDER

ORDER IS NEXT

MODE IS CHAIN LINKED TO

PRIOR

OWNER IS CUST-REC

MEMBER IS ORDER-REC LINKED

TO OWNER

MANDATORY

AUTOMATIC.

Continue …..

112

It completes the mapping of logical areas into physical files.

It also identifies and describes the journal files that are used by IDMS for recording changes made to the database. Journal files are used to implement IDMS backup and recovery facilities.

DMCL DDL STATEMENTS

113

DMCL DDL statements define the following:

DMCL module name Schema name associated with this

DMCL module Sizes of the I/O buffers Information about database areas Information about journal files used

for backup and recovery

Continue …..

114

The DMCL definition is processed by the DMCL compiler. The DMCL compiler reads the DMCL source statements and stores a description of the DMCL module in the data dictionary.

As the DMCL compiler processes DMCL source, it must have access to the corresponding schema definition in the data dictionary. For this reason, schema must be compiled before its associated DMCL module.

Process

115

The DMCL compiler also generates a set of assembler language source statements. These are in turn processed by assembler and the linkage editor to form an executable DMCL load module. The linkage editor places the load module into a load module library. The DMCL load module is used by IDMS at run-time when an application program accesses the database files described by the DMCL module.

Continue …..

116

DMCL Compiler

Assembler

Linkage Editor

DMCL Source

StatementsData

Dictionary

Load Library

Continue …..

117

DMCL DDL statements contain following four sections: Device-Media Description section

This sections gives name to the DMCL statements. It further provides other documentary details about the DMCL module.

Buffer section

It assigns page size and defines buffers that IDMS will use to handle database accesses.

Continue …..

118

Area section

Information about each area is copied from the schema into this section of the DMCL. The section assigns each area to buffers that were defined in the buffer section.

Journal section

The section describes disk files to be used for journaling. Journal files are used to keep track of database activity.

Continue …..

119

DEVICE-MEDIA DESCRIPTION.DEVICE-MEDIA NAME IS DMCLBASE OF SCHEMA NAME SCHBASE VERSION 1.DATE. DD/MM/YY.INSTALLATION. XANSA MUMBAI

INDIA.BUFFER SECTION.

BUFFER NAME IS XXBUFFPAGE CONTAINS 4096 CHARACTERSBUFFER CONTAINS 5 PAGES.

JOURNAL BUFFER NAME IS JJBUFFPAGE CONTAINS 3188 CHARACTERSBUFFER CONTAINS 3 PAGES.

Example Of DMCL DDL

120

AREA SECTION.COPY SALARY-AREA AREA

BUFFER IS XXBUFFJOURNAL SECTION.

JOURNAL BUFFER IS JJBUFF.FILE CONTAINS 3000 BLOCKS.FILE NAME IS XANJRNL1

ASSIGN TO SYSJRNL1.FILE NAME IS XANJRNL2

ASSIGN TO SYSJRNL2.

Continue …..

121

Subschema DDL statements specify the following:

Subschema, Schema and DMCL module name

Areas accessible in this subschema

Records and data elements accessible and

Sets included in the subschema

SUBSCHEMA DDL STATEMENTS

122

The subschema compiler reads subschema source statements and stores a description of the subschema in the data dictionary.

The compiler also stores an executable load module in an area of the data dictionary, known as Load Area.

IDMS uses this load module whenever an application program using this subschema is under execution.

Process

123

Like DMCL compiler, subschema compiler must have access to the corresponding schema definition in the data dictionary. Thus, schema must have been compiled before any of its associated subschemas can be compiled.

After compiling subschema, data dictionary has all the required information about the logical and physical structures of the database.

Continue …..

124

Data Dictionary

SubschemaDDL

SubschemaCompiler

Load Area

Continue …..

125

ADD SUBSCHEMA NAME SUBSCHA OF SCHEMA NAME SCHDBA VERSION 1 DMCL NAME SUBADMCL OF SCHEMA NAME SCHDBA VERSION 1 COMMENTS ‘LIMITED VIEW OF SCHDBA’

ADD AREA NAME IS CUST-AREA.ADD AREA NAME IS ORDER-AREA.ADD RECORD NAME CUST-REC

ELEMENTS ARE CUST-ID CUST-NAME.ADD RECORD NAME ORDER-REC

ELEMENTS ARE ALL.ADD SET NAME ORDER-ACTION.ADD SET NAME CUST-ORDER.GENERATE.

Example Of Subschema DDL

126

Data Manipulation Language (DML)

127

In an IDMS application program, there

are no file description statements or Open

and Close statements found in a typical

batch Cobol file processing application

system.

To access data stored in the database,

DML verbs are used.

Two new terms Run Unit and Currency

are important to understand.

128

The BIND RUN-UNIT statement is used to specify the start of a run unit.

A run unit ends when program completes all database processing and issues a FINISH statement.

CONCEPT of RUN UNIT

The run unit is that portion of program’s

processing during which it has access to

one or more database areas and services.

129

The program may perform some

processing before it begins a run

unit.

The program may perform some

processing even after it finishes a run

unit.

A program may begin and end more

than one run unit.

Continue …..

130

IDMS internally stores pointers to

the most recently accessed records.

It uses these pointers to access

subsequent records. These pointers

are referred to as CURRENCY.

At the beginning of a program, all

currencies are null.

CONCEPT of CURRENCY

131

Currency is established by DML verbs FIND, OBTAIN, STORE, CONNECT, DISCONNECT and ERASE.

There are four currencies maintained for a database: current of run unit current of record type current of set current of area

Continue …..

132

Current of Run Unit is a record which was most recently accessed.

Current of Record type is the most recently retrieved or inserted record of that record type.

Current of Set is the most recently retrieved or stored record of that set. Record may be a member or owner of that set.

Continue …..

133

Current of Area is the most recently retrieved or stored record of that area.

When a record is erased, currency of its record type and its sets is set to null. Run unit and area currencies are unchanged.

Continue …..

134

DML Processor

Compiler

Linkage Editor

Source Statements Data

Dictionary

Load Library

IDMS APPLICATION PROGRAM PREPARATION

135

The previous slide shows how an IDMS application program written in Cobol host language is compiled and made ready for execution.

Application program with embedded DML statements is first read by DML processor. It validates DML statements and creates translated version of the source program where in DML statements are replaced by appropriate calls to IDMS.

136

DML processors are available for a number of commonly used programming languages. DML processor may also generate a listing of errors. DML processor interacts with data dictionary for copying descriptions of various database elements like schema, subschema, record types etc.

Output from the DML processor acts as input to the host programming language compiler like a Cobol compiler.

Continue …..

137

After the translated source program is

compiled by host language compiler, the

resulting object module is processed by

Linkage editor to create an executable

load module.

The Linkage editor also includes copy of

a routine called IDMS Interface Module

in final load module. IDMS service

requests from the program are converted

in to calls to IDMS Interface Module.

Continue …..

138

Following are typical changes one

notices in a n IDMS Cobol application

program.

IDENTIFICATION DIVISION

remains exactly the same as before.

ENVIRONMENT DIVISION has

IDMS-CONTROL section.

Continue …..

139

DATA DIVISION begins with

SCHEMA SECTION which gives

the subschema used by the

program.

PROCEDURE DIVISION has

DML verbs for data retrieval and

data updation.

Continue …..

140

.….

ENVIRONMENT DIVISION.

IDMS-CONTROL SECTION.

PROTOCOL.

MODE IS BATCH DEBUG.

IDMS-RECORDS WITHIN WORKING-STORAGE SECTION.

CONFIGURATION SECTION.

…..

INPUT-OUTPUT SECTION.

FILE CONTROL.

…..

IDMS-CONTROL SECTION

141

The PROTOCOL statement is a

compiler-directive statement that

specifies the manner in which DML

processor will generate CALL

statements.

Application program must specify one

of the following modes to access

database:

Continue …..

142

BATCH (a batch program with

database processing)

IDMS-DC (a non-batch program i.e.

an online program called by a dialog

process)

DC-BATCH (a batch program with

access to DC queues)

Continue …..

143

‘DEBUG’ instructs DML processor to

produce additional code in the resulting

source code which is helpful during

program debugging.

The IDMS-RECORDS clause specifies

that the data definitions for records and

data elements will be copied automatically

from the data dictionary and placed at the

end of working storage section.

Continue …..

144

.….

DATA DIVISION.

SCHEMA SECTION.

DB SUBEMP WITHIN SCHEMP.

FILE SECTION.

…..

WORKING-STORAGE SECTION.

…..

DATA DIVISION

145

As explained earlier, DATA DIVISION

begins with SCHEMA SECTION which

gives the subschema used by the

program.

As shown in previous slide, SUBEMP is

the name of the subschema which will be

used by the program.

Further, the SCHEMP is the schema

from which SUBEMP subschema is

derived.

Continue …..

146

BIND RUN UNIT.

BIND EMPLOYEE.

BIND MANAGER.

BIND PROJECT.

BIND SALARY.

BIND STATEMENT

Above example displays the typical

code that might be found in the

Procedure Division of an IDMS

Cobol program.

147

Continue …..

The first statement BIND RUN-UNIT

establishes addressability for the IDMS

communications block and causes

IDMS to load the subschema identified

in the Schema section. Next four BIND

statements establish addressability for

the working-storage areas that will be

used to contain record occurrences.

148

Continue …..

Generally, following one statement is found in the source code.

COPY IDMS SUBSCHEMA-BINDS.

When the source is pre-compiled, the above single statement is replaced by previous set of five Bind statements.

BIND statement identifies the record types that the program will access.

149

COPY IDMS SUBSCHEMA-BINDS.

COPY IDMS IDMS-STATUS.

COPY STATEMENT

Copy statement is a compiler directive

statement.

In first example above, the copy

statement generates appropriate Bind

statements at the time of pre-compile.

150

Continue …..

As in second example, the copy

statement copies module called

IDMS-STATUS from data

dictionary into the source code

when it is pre-compiled. This

particular module (routine) checks

result of DML statement.

This helps in reducing coding effort.

151

READY EMPLOYEE-AREA USAGE IS RETRIEVAL.

READY SALARY-AREA USAGE EXCLUSIVE UPDATE.

READY USAGE IS RETRIEVAL.

READY STATEMENT

Ready statement can be thought as an

equivalent to Open statement for a

conventional file. It gives the program

access to all or named areas defined in

the subschema and begins a run-unit.

152

Continue …..

The usage-mode clause in the Ready

statement indicates whether the program

will only retrieve data or update it as

well.

If no areas are mentioned in the Ready

statement, then all areas mentioned in

the subschema will be made available to

the program by default in the mode

specified in the Ready statement.

153

FINISH.

FINISH STATEMENT

Finish statement is used to end a run

unit by releasing database resources and

terminating database processing.

Non-database processing of the program

may continue after the Finish statement,

but database can not be accessed (unless

a fresh run unit is started).

154

MOVE INPUT-EMP-CODE TO EMPLOYEE-CODE.

FIND CALC EMPLOYEE.

FIND STATEMENT

Find statement is used to locate a record

occurrence in the database. The record

occurrence is placed in the system

buffers but the data element values are

still not available in the variable storage

of the program.

155

Continue …..

There are situations when we do not need

actual values of a record occurrence like:

We may need to verify that a particular record occurrence exists

We need to establish currencies for some subsequent record retrieval

In the example, the required employee information is moved to the CALC-key data element of Employee.

156

Continue …..

Next, Find statement will locate the record occurrence with matching employee-code.

If the Find is successful, currencies are set

appropriately but the individual data

element values for the located record are not

yet made available to the program.

In case, data element values are required

then Find can be followed by Get statement.

157

MOVE INPUT-EMP-CODE TO EMPLOYEE-CODE.

FIND CALC EMPLOYEE.

GET.

GET STATEMENT

Get statement is used to make available to

the program a record occurrence that was

previously located by the Find statement.

Get statement can qualify the record type

or it can be unqualified as in this case.

158

Continue …..

To continue with our earlier Find example,

we have followed it with unqualified Get.

This will transfer record occurrence values

to the variable storage.

If Get is qualified then IDMS will verify

whether the previously located record is

Employee record or not. If it does not

match then IDMS will return error code.

159

The Obtain statement is equivalent of a Find followed by a Get.

It is most widely used statement for

data retrieval in IDMS.

There are many syntax variations

available for Obtain statement, based

on location mode of records and mode

of retrieval.

OBTAIN STATEMENT

160

CALC-key Retrieval

To perform Calc-key retrieval, the record

must have location mode specified as

CALC and we must know the Calc-key

value for the record to be retrieved.

Continue …..

MOVE INPUT-EMP-CODE TO EMPLOYEE-CODE.

OBTAIN CALC EMPLOYEE.

161

As shown in previous example, we move

the Calc-key value for the record to be

retrieved to the data element of Employee

record type that is defined as Calc-key

(Employee-Code) and issue Obtain

statement to get the record contents.

IDMS uses the Calc-key value supplied to

calculate the target page for the data

retrieval.

Continue …..

162

Duplicate CALC-key Retrieval

Sometimes the record type is defined

with duplicate allowed option. We can

try and find out if duplicate record exists

for a given Calc-key value.

Continue …..

MOVE INPUT-EMP-NAME TO EMPLOYEE-NAME.OBTAIN CALC EMPLOYEE.OBTAIN DUPLICATE EMPLOYEE.

163

As shown in previous example, we

obtain the first record with matching

Calc-key value in normal way.

The second Obtain statement tries to

obtain duplicate record with same

Calc-key value. If it exists then IDMS

fetches the record, else an error code is

returned.

Continue …..

164

Member Record Retrieval

Many times, we need to retrieve owner

record for a set occurrence and then

proceed to retrieve one or more of its

member records.

Following are some of the examples of

such typical member retrieval criteria.

Continue …..

165

Continue …..

OBTAIN FIRST EMPLOYEE WITHIN DEPT-EMP.

OBTAIN LAST EMPLOYEE WITHIN DEPT-EMP.

OBTAIN NEXT EMPLOYEE WITHIN DEPT-EMP.

OBTAIN PRIOR EMPLOYEE WITHIN DEPT-EMP.

OBTAIN OWNER WITH DEPT-EMP.

OBTAIN 5 WITHIN DEPT-EMP.

166

By first retrieving the owner record, we

establish the currency for the subsequent

member retrievals. Owner becomes

current of run unit, record type and set

type. Now to access members within the

set becomes relatively simple task.

“Obtain First” gets the first member

record within the set.

Continue …..

167

“Obtain Last” gets the last member

record within the set. For this, the set

must have been defined with prior

linkage option.

Next option is known as Walking the Set.

“Obtain Next” gets each member in

sequence. This statement is executed in a

loop generally till end-of-set is reached.

Continue …..

168

“Obtain Prior” gets the prior member

record within the set. For this, the set

must have been defined with prior

linkage option. This works in same

manner as Next option, except that it

is in the reverse direction.

Continue …..

169

We can obtain owner of the set by

“Obtain Owner” statement.

If set is defined with Owner linkage

pointer then owner is obtained directly.

In case such pointer is absent, IDMS

walks through the entire set and

obtains the set owner.

Continue …..

170

We can obtain specific member record

within the set by its relative position with

respect to owner in the ‘next’ direction.

“Obtain 5” will fetch the fifth member

record in the set.

Continue …..

171

Current Record Retrieval

Many times, we need to retrieve records

based directly on various currency.

Continue …..

OBTAIN CURRENT EMPLOYEE.

OBTAIN CURRENT WITHIN DEPT-EMP.

OBTAIN CURRENT WITHIN EMPLOYEE-AREA.

172

In first case, IDMS obtains the record

that is current of the Employee record

type.

In second case, current of the Dept-Emp

set type is fetched. Depending on

currency established, either Dept record

or an employee record is obtained.

Continue …..

173

In last case, current of employee-area

is obtained. e.g. if employee-area

contains employee, project, salary

record types then depending on

currency, the statement will fetch

record occurrence of any one of these

record types.

Continue …..

174

Area Sweep Retrieval

Sometimes, we need to retrieve records by

doing an area sweep. IDMS scans through

the area in physical sequence. Area sweep

is used when we need to access all records

within an area, irrespective of the record

types and/or their sequence.

Continue …..

175

Continue …..

OBTAIN FIRST WITHIN EMPLOYEE-AREA.

OBTAIN NEXT WITHIN EMPLOYEE-AREA.

OBTAIN LAST WITHIN EMPLOYEE-AREA.

In first case, the record having lowest db-key value within the employee-area will be fetched. In second case, record having next highest db-key value will be obtained. The sweep in reverse direction as in last case can also be made.

176

Continue …..

OBTAIN FIRST EMPLOYEE WITHIN EMPLOYEE-AREA.

OBTAIN LAST EMPLOYEE WITHIN EMPLOYEE-AREA.

As shown above, the record type can

also be qualified during the area sweep.

Only record type mentioned in the

statement (employee) are obtained

during the sweep. Occurrences of other

record types are ignored.

177

Sort-Key Retrieval

We can retrieve records based on sort-

key values, provided the set is defined as

Sorted order option.

In our earlier example of American

baseball league database, suppose that

set Player-Position is defined as Sorted.

Continue …..

178

The sort-key data element is Position-

name of Position record.

Then we can retrieve Position record

directly based on sort-key value as below.

Continue …..

….

OBTAIN CALC PLAYER-REC.

MOVE INPUT-POSITION-NAME TO POSITION-NAME.

OBTAIN POSITION-REC WITHIN PLAYER-POSITION

USING POSITION-NAME.

179

In the example, a position-name value is

moved to sort-key data element and then

Obtain statement is issued using reference

to the sort-key element.

The retrieval appears similar to CALC, but

in this case the search is made only within

the current set occurrence and not within

entire record type.

Continue …..

180

DB-Key Value Retrieval

If we know the dB-key value of a record

then we can directly obtain the record.

This is useful in cases where we know

that a particular record would be

processed again later on during the

execution of the program.

Continue …..

181

When a record is successfully retrieved

or located, its dB-key value is available

in the IDMS Communication Block. We

can store this value in a variable storage

area for later use. Later on, we can use

the stored dB-key value to retrieve the

record and process it.

Continue …..

182

…..

MOVE INPUT-PLAYER-NAME TO PLAYER-NAME.

OBTAIN CALC PLAYER-REC.

…..

MOVE DBKEY TO SAVE-DBKEY.

…..

OBTAIN DB-KEY IS SAVE-DBKEY.

Continue …..

183

Continue …..

Note that the dB-key of a record may

undergo change whenever a database is

reorganized. Hence it is not always

advisable to save dB-key values for a

long time with an intention of using

them later on.

184

Index Set Retrieval

Using index sets, following types of

retrieval is made possible which is

otherwise very difficult.

Generic key retrieval.

Random and sequential retrieval based

on a key other than Calc-key.

Continue …..

185

Suppose in our baseball example, Player

record has a system indexed set IX-

PLAYER-NAME on Player-Name

element, then we can retrieve a player

record by supplying a partial key to

IDMS. Such retrieval is not possible

unless we have index set. In Calc, we

have to supply complete key value.

Continue …..

186

01 WS-PLAYER-NAME.

05 FIRST-INITIAL PIC X.

05 FILLER PIC X(29).

…..

MOVE LOW-VALUES TO WS-PLAYER-NAME.

MOVE WS-INITIAL-LETTER TO FIRST-INITIAL.

MOVE WS-PLAYER-NAME TO PLAYER-NAME.

OBTAIN PLAYER-REC WITHIN IX-PLAYER-NAME USING PLAYER-NAME.

…..

OBTAIN NEXT PLAYER-REC WITHIN IX-PLAYER-NAME.

Continue …..

187

Say we want to retrieve all players

starting with ‘C’.

We move ‘C’ to ws-initial-letter and rest

is filled with low values. This is called

Partial key value.

Then the program will obtain first player

starting with name ‘C’.

Continue …..

188

First obtain statement is random retrieval

using index set. Whereas the second

obtain statement is the example of the

sequential retrieval using index set.

Program has to check name field after

every retrieval to determine if it has

reached the last record with specific name.

Continue …..

189

The Modify statement is used to change data element values in an existing record occurrence.

Important considerations for a record to be successfully modified are:

Ready the concerned area(s)

with one of the update options.

MODIFY STATEMENT

190

Making the record to be modified as current of run unit. This is done by retrieving the record before modifying it, by issuing Obtain.

Continue …..

191

MOVE INPUT-EMP-CODE TO EMPLOYEE-CODE.

OBTAIN CALC EMPLOYEE.

IF DB-STATUS-OK

MOVE NEW-EMP-ADDRESS

TO EMPLOYEE-ADDRESS

MODIFY EMPLOYEE

PERFORM IDMS-STATUS.

Continue …..

Above example shows how an existing

Employee record is modified.

192

The Store statement is used to add

new record occurrence to the database.

Adding a new record involves:

Constructing new record

occurrence in the record’s area in

variable storage.

STORE STATEMENT

193

Establish currency with correct set

occurrences in which the new

record participates as automatic

member. FIND is used to set

currencies.

Issue Store statement specifying

the name of the record type we are

adding.

Continue …..

194

Syntax for Store is :

STORE EMPLOYEE.

While adding a new record, following

information is analyzed.

Continue …..

195

Location where new record is added.

If location mode is CALC, then based

on Calc-key value, storage page is

calculated. For VIA, optimum

location is found based on the

location of owner record. If its

DIRECT, then program specifies the

page number through db-key.

Continue …..

196

Where in the set occurrences the new

record should be inserted.

Before issuing Store, set order option

is also examined. Depending on

order option like first, last, next etc,

the new record is inserted in

appropriate position within the set.

Continue …..

197

The Erase statement is used to delete

a record occurrence from the

database.

Erase statement often affects not only

the record that is being deleted but, in

some cases, also members of sets

owned by that record.

ERASE STATEMENT

198

There are four options available for the

Erase statement.

ERASE

ERASE ALL MEMBERS

ERASE PERMANENT MEMBERS

ERASE SELECTIVE MEMBERS

Continue …..

199

ERASE

We first begin by retrieving the record

we want to delete and thus the record is

made current of run unit. Then Erase is

issued naming the record type.

The record is first disconnected from all

sets in which it participates as a

member.

Continue …..

200

Then the record occurrence is

removed (erased) from the database.

If the record is the owner of any sets,

then the record will be erased ONLY

IF those sets are empty. If we try to

erase an owner record of a non-empty

set, IDMS returns an error status code.

Continue …..

201

MOVE INPUT-EMP-CODE TO EMPLOYEE-CODE.

OBTAIN CALC EMPLOYEE.

IF DB-STATUS-OK

ERASE EMPLOYEE

PERFORM IDMS-STATUS.

Continue …..

Above example shows how an existing

Employee record is erased from the

database with unqualified Erase statement.

202

ERASE ALL MEMBERS

First the record occurrence is removed

from the database.

Then all its members, both mandatory

and optional, are also removed from

the database.

Continue …..

203

ERASE PERMANENT MEMBERS

First the record occurrence is erased from

the database.

Then all mandatory members in all sets

owned by the record are also removed.

Optional members are NOT removed but

they are disconnected from their respective

set occurrences.

Continue …..

204

Lastly, if the erased member is the

owner of any sets then it is treated as if

an Erase Permanent is issued for it. Its

mandatory members are removed and

optional members are disconnected.

This process continues until all

members have been processed.

Continue …..

205

ERASE SELECTIVE MEMBERS

First the record occurrence is erased from

the database.

Then all mandatory members in all sets

owned by the record are also removed.

Optional members are removed, provided

they do not participate in any other sets.

Continue …..

206

Lastly, if the erased member is the

owner of any sets then it is treated as

if an Erase Selective is issued for it.

Its mandatory members are removed

and optional members are removed if

they do not participate in any other

sets.

Continue …..

207

Connect statement is used to link a record

occurrence into a set.

Here, we name both the record type and

the set type.

The statement causes the current of

record type to be connected to the set

occurrence that is current of set type.

CONNECT STATEMENT

CONNECT EMPLOYEE TO DEPT-EMP.

208

In case of set defined with Automatic

member option, the member record gets

connected automatically at the time of

addition to the database.

With set defined with Manual member

option, Connect statement is used to

manually set up link with a set

occurrence.

Continue …..

209

Disconnect statement cancels the membership

of a record in a set occurrence in which it

currently participates as a member.

The record that is being disconnected, must be

defined as an Optional member of that set. If

not then IDMS returns an error code.

DISCONNECT STATEMENT

DISCONNECT EMPLOYEE FROM DEPT-EMP.

210

QUEUE RECORDS

We write temporary data to a queue record in data dictionary and make this data available to other IDMS COBOL programs (DC-BATCH Mode).

Queue records are relatively permanent records in that they are not deleted when the system is shutdown or when the system crashes. They are deleted either explicitly by programmer or after the retention period specified for the queue is over.

211

Continue …..

As the data remains permanent, it is global in a sense. Thus, queues are used to pass data information from one application to another. Specifically this feature is very widely used in online applications using ADSO.

212

QUEUE MANAGEMENT COMMANDS

Commands to write a record to a queue, retrieve a queue record and delete a queue record are:

PUT QUEUE ID queue-id GET QUEUE ID queue-id DELETE QUEUE ID queue-id

213

Continue …..

PUT QUEUE ID queue-id

This command is used to store a

queue record in the data dictionary in

the queue identified by the queue-id.

An added parameter of Last or First

can be mentioned to instruct whether

the new queue record will be placed

at the beginning or end of the queue.

Default is Last.

214

Continue …..

GET QUEUE ID queue-id

This command is used to transfer the

contents of a queue record to a

specified location in the variable

storage. Added parameter is Delete or

Keep. Delete will remove the record

from the queue after retrieval. Keep

will retain the record in the queue.

Default is Delete.

215

Continue …..

DELETE QUEUE ID queue-id

This command is used to delete one

or all queue records in a specified

queue. Added parameter is Current or

All. Current will delete only the

current record of the queue. All will

delete all queue records, including the

header record.

Default is Current.

216

MAJOR CODES OF ERROR-STATUS

01 : FINISH

02 : ERASE

03 : FIND/OBTAIN

05 : GET

06 : KEEP

07 : CONNECT

08 : MODIFY

09 : READY

217

Continue …..

11 : DISCONNECT

12 : STORE

14 : BIND

18 : COMMIT

19 : ROLLBACK

218

Recovery and Restart of Database

219

IDMS provides facilities for recovering from problem situations and for restarting after failures occur.

Journal files are used for the recovery and restart facilities.

These facilities are dependant on the two operating environments in which an IDMS run unit executes.

220

IDMS application program can be run in two operating environments: Local Mode and Central Version.

All online applications that use IDMS/DC or some other telecommunication facilities must run under central version.

An application program that operates in batch mode can be run either under the control of central version or in local mode.

IDMS OPERATING ENVIRONMENTS

221

In central version, a single copy of the IDMS DBMS controls the operation of a particular set of IDMS application program run units.

No two central version are allowed to update the same database area. All run units that execute under the control of a particular central version are chosen such that they all access the same group of database areas.

Central Version

222

Multiple copies of IDMS central version can execute in same region with each copy accessing different set of database areas.

Central version implements facilities for

recovering from system failures.

It is the preferred method of running an

IDMS application.

Continue …..

223

A batch program that operates in local mode has its own copy of IDMS DBMS loaded into its partition, region or address space.

IDMS provides no automatic restart and

recovery facilities for applications running

in local mode.

Batch retrieval programs are good

candidates for running in local mode.

An application runs faster in local mode.

Local Mode

224

IDMS maintains set of journal files that

are used to recover from failures and to

restart aborted run units.

IDMS provides automatic restart and

recovery facilities for applications

running under central version AND for

run units those use journal files.

Journal files can be put on disk or tapes.

JOURNAL FILES

225

CHECKPOINTING

When a run unit begins by issuing BIND run-unit statement, IDMS system writes Begin checkpoint record on the journal file. This contains information about current status of the run-unit.

As run-unit executes, it may update records in database. As each update request is processed, IDMS writes Before Image and After Image of the updated record to the journal file.

226

Continue …..

When applications ends the run-unit by issuing FINISH statement, IDMS writes End checkpoint record on to the journal file.

Thus, at the end of execution of a run-unit the journal files contains:

Begin checkpoint

Before and After images of all updated records and

End checkpoint

227

RECOVERY and RESTART

Following are typical scenarios when run unit is terminated abnormally.

Program error may cause the run unit to terminate abnormally.

IDMS system may encounter situations that require it to terminate one or more run units abnormally. e.g. Wait time limit has been exceeded to avoid deadlock.

228

Continue …..

In situations like these, when a run unit is

terminated abnormally after it has updated

one or more database records; IDMS

works backwards in the journal file until it

reaches the Begin checkpoint record and

uses the information in the Before images

to reverse effect of database updates

already made by the aborted run unit. The

run unit can then be restarted from the

beginning.

229

RECOVERY UNIT

All run units that terminate normally, create at least two checkpoint records on the journal file: BEGIN CHECKPOINT when Ready/Bind statement is executed and END CHECKPOINT when FINISH statement is executed. Before and After images that are recorded between these two checkpoints constitute a record of all updating the run unit performed.

230

Continue …..

The processing that a run unit performs between checkpoints is called Recovery Unit.

In the simplest form, the entire run unit constitutes a single recovery unit.

231

INTERMEDIATE CHECKPOINTS

In real life scenario (for online applications), a separate run unit is started for each incoming transaction. Very few database updates are made during each run unit. Each run unit is a single recovery unit.

At times, it is required to divide the run unit into many smaller run units to reduce the time for recovery in case of run unit failure.

232

Continue …..

This leads us to the concept of

Intermediate checkpoints.

To create intermediate checkpoints, run

unit periodically issues a COMMIT

statement to make permanent all

database updates that has already been

made.

233

THE COMMIT VERB

Commit verb is used to write Intermediate checkpoints on the journal file.

There are two forms of Commit verb:

COMMIT and

COMMIT ALL

234

Continue …..

In COMMIT, IDMS releases all locks except those placed on current records.

In COMMIT ALL, IDMS releases all locks including those on current records.

In case of abnormal termination of run unit, IDMS works backwards in the journal file and reads till the most recent COMMIT checkpoint is encountered.

235

Continue …..

Now the journal file is not read till

Begin checkpoint. The run unit can

now be restarted from the point at

which it wrote the most recent Commit

checkpoint.

A simple thumb rule of 500 or 1000

database updates is used to issue a

Commit statement.

236

THE ROLLBACK VERB

Sometimes an application program may

itself decide to back out all the updates it

made to the database since it wrote the last

intermediate (commit) checkpoint.

To back out the changes, we can issue a ROLLBACK statement.

It has two forms:

ROLLBACK and

ROLLBACK CONTINUE

237

ROLLBACK

It writes a checkpoint record to the

journal file, nullifies all currencies,

automatically restore the database

using the before images from the

journal and terminate the run unit.

After issuing the unqualified Rollback,

to access the database again, we must

reissue the Bind.

238

ROLLBACK CONTINUE

This form causes IDMS to restore the

database but does not terminate the run

unit.

One can access the database immediately

without reissuing the Bind.

239

Locking Facilities

240

Locking facilities are provided for

controlling access to areas and to individual

records when two or more run units require

access to the same database areas.

IDMS provides three levels of locking

facilities.

Area Locks

Area Usage Modes

Record Locks

241

AREA LOCKS

Area locks are used to prevent areas from being updated by more than one IDMS central version or by more than one application running in local mode.

If an area is locked that is required by a

local mode application, that local mode

application ABENDS.

242

Continue …..

If required area is NOT locked, area

lock is set by the local mode

application and no other local mode

application is given access to that area.

An area locked by a local mode

application is released when that

application executes its FINISH

statement (end of run-unit).

243

Continue …..

If required area is locked that is required by an IDMS central version, execution of that central version continues, but no run-unit executing under the control of that central version will be given access to the already locked area. The area is said to be Varied Offline to that central version. Run-units that attempt to access an area that is varied offline are abnormally terminated.

244

Continue …..

If required areas are NOT locked, then those areas are locked by the central version. Run-units executing under the control of that central version can access those areas. No other central version will be given access to locked areas.

An area that is locked by a central version is released only when the central version is shut down.

245

AREA USAGE MODES

Area usage modes control how database areas under the control of a central version can be accessed.

The six possible usage modes for a database area are as follows:

246

Continue …..

(Shared) Retrieval

Protected Retrieval

Exclusive Retrieval

(Shared) Update

Protected Update

Exclusive Update

The ‘Shared’ word is optional.

247

Retrieval or Shared Retrieval

We are requesting only retrieval access

to the area. All other run units under

the same central version will be

allowed to RETRIEVE or UPDATE

records in the same area, provided they

do no specify usage mode Exclusive.

248

Protected Retrieval

We are requesting retrieval access to the area and it also specifies that no other run unit should update records in the same area for the duration of our run unit. All other run units under the same central version will be allowed to retrieve records in the same area provided they did not specify usage mode Exclusive.

If another run unit is already updating records

in the area, our run unit will wait till that run

unit finishes.

249

Exclusive Retrieval

We are requesting retrieval access to the area

and it also specifies that no other run unit

should retrieve or update records in the same

area till our run unit finishes execution. If

another run unit is already accessing the

area, our run unit will wait till that run unit

finishes.

This mode is not often used.

250

Update or Shared Update

We are requesting updating access to the

area. All other run units under the same

central version will be allowed to retrieve

or update records in the same area as long

as they do not specify Protected or

Exclusive usage mode options in Ready

statements.

251

Protected Update

We are requesting update access to the area and

specifying that other run units should not be able

to update records in the same area till our run unit

finishes. All other run units will be allowed to

retrieve records in the same area as long as they

do not specify Protected or Exclusive as usage

modes.

If another run unit is already accessing the area,

our run unit will wait till that run unit finishes.

252

Exclusive Update

We are requesting update access to the area

and further specifying that no other run unit

should either retrieve or update records in the

same area while our run unit is executing.

This option gives our run unit exclusive use

of the area.

253

Sha

red

Upd

ate

Sha

red

Ret

rieva

l

Pro

tect

ed U

pdat

e

Pro

tect

ed R

etrie

val

Exc

lusi

ve U

pdat

e

Exc

lusi

ve R

etrie

val

Shared Update Y Y N N N N

Shared Retrieval Y Y Y Y N N

Protected Update N Y N N N N

Protected Retrieval N Y N Y N N

Exclusive Update N N N N N N

Exclusive Retrieval N N N N N N

Run

Uni

t 1

Run Unit 2

254

IMPLICIT RECORD LOCKS

It is maintained only for run-units operating under Central Version.

When we use usage mode options that allow more than one run-unit to access the same area while updating is taking place, IDMS sets Implicit record locks to prevent same record from being updated simultaneously by two or more run-units.

255

Continue …..

Record locks are generally maintained for shared and protected update mode and shared retrieval mode.

Implicit locks can be either Shared or Exclusive.

Implicit Shared lock guarantees that only one run-unit is allowed to update a record while others can retrieve the same record.

Implicit Exclusive lock ensures that no other run-unit can either update or retrieve the record.

256

Implicit shared locks is placed on a record when it is retrieved and Implicit Exclusive lock is placed on a record when it is accessed through a DML update verb.

Shared Lock remains in effect till currency changes.

Exclusive Lock remains in effect till run-unit ends (Finish) or until Commit is issued.

Continue …..

257

Explicit locking is the most efficient way of placing record locks. It can be set by coding the KEEP statement or by the KEEP clause of the FIND/OBTAIN statement.

Explicit lock remain is effect until Finish is executed or Commit is issued.

e.g. Obtain Keep Calc Employee.

Obtain Keep Exclusive Calc Employee.

First is explicit shared lock and second is explicit exclusive lock.

EXPLICIT RECORD LOCKS

258

Another way of setting explicit locks is to code separate Keep statement after a record is retrieved.

e.g.

Keep exclusive current.

Keep current within Dept-Employee.

Keep exclusive current within employee-area.

Continue …..

top related