| <<< PRSOIS::SYS$SYSDEVICE:[NOTES$LIBRARY]CLUB_GESTION.NOTE;1 >>>
-< SIG Club Gestion >-
================================================================================
Note 8.0 ADABAS (Merci P.Grogniet!) 1 reply
PRSSOS::LECORE "Guy Lecore, RDB/VAXclusters, Count" 488 lines 20-FEB-1987 12:15
--------------------------------------------------------------------------------
ADAPTABLE DATA BASE
(ADABAS V.4.1; ADABAS (VMS) V.1.1)
Initial compilation: September, 1985.
Revised: September, 1986.
1.1 ******** COMPETITIVE ANALYSIS ********
1.1.1 PRODUCT OVERVIEW (ADABAS V.4.1; ADABAS (VMS) V.1.1)
ADABAS is an inverted index data base management system with
relational-like features. ADABAS uses a number of utility programs
under DOS or OS with EXCP for data base generation and access. The
ADABAS nucleus supports concurrent batch and online processing.
Interfaces are provided for TP monitors such as ComPlete, CICS, TSO
and Taskmaster as well as IMS DB/DC and FASTER. All ADABAS (VMS)
communications are dependent on the operating system facilities. A
data compression algorithm to load into the data base is an integral
function of the system. Also featured is the separation of physical
data storage from the representation of logical relationships in the
data base. ADAMINT is used to generate high-level interface routines
for applications programs. ADABAS also includes a comprehensive
security system and automatic restart/recovery capabilities. [4,5]
ADABAS (VMS) is a version of the system for use with the DEC
VAX-11 series of minicomputers. ADABAS (VMS) is a full implementation
of mainframe ADABAS for VAX processors, and is currently bundled with
NATURAL (VMS). The ADABAS/VSAM Bridge is a conversion aid that allows
users with VSAM data bases to migrate to ADABAS with no new
applications programming. This facility is available to both OS/VS
and DOS/VSE users. [5]
1.1.2 GENERAL TRENDS, POSITIONING, COMPETITIVE INFORMATION
o We will see a significant challenge in the future from this
product, as it is a good performer and has a good reputation
among the user base.
o ADABAS does support clusters as a network, but routes all
requests through one node.
1.1.3 SELLING STRATEGIES
KNOCKOFFS (Rdb/VMS vs. ADABAS)
o Unstructured data support (any digitized data can be stored
in Rdb/VMS databases).
o The price of ADABAS ($50,000 on VMS) is steep compared to
$15,600 (for Rdb/VMS and DATATRIEVE); this will make Rdb/VMS
more suited to the smaller applications and tight budgets.
o Rdb/VMS supports several more datatypes, including varying
text, date, and floating point.
o VIA integration allows us to sell forms (TDMS), graphics and
a report writer (DATATRIEVE), spreadsheet (DECalc), an
Application Control and Management System (ACMS) and
dictionary (CDD) for well under the total cost of ADABAS with
options. There is no spreadsheet available with ADABAS.
o Users will likely find ADABAS more difficult to use, as
physical and logical database design must be taken into
consideration. Rdb/VMS users must only consider logical
database design or have the ability to easily change
the design to get good performance with no further tuning.
There are no tradeoffs of flexibility vs. performance.
o Digital offers strength in distributed database management
through the remote access and update capabilities of Rdb/VMS
and remote transaction processing capabilities of VAX ACMS
V2. This enables us to put together a TP solution based on
the Digital style of computing (many machines cooperating to
do the job) rather than selling the centralized approach
typical with an ADABAS installation.
o Rdb/VMS provides ease of use and flexibility, especially at
the design phase. We make it much easier to get the database
up and running with good performance for smaller
organizations and applications that do not have the benefit
of a DBA, Rdb/VMS should have the edge.
STRENGTHS (ADABAS)
o ADABAS excels in high volume data base management
environments where there is a centrally administered file, as
compared to VAX DBMS.
o ADABAS is easy to use. It is a field oriented system with no
embedded structure. This structure facilitates changes.
o Security is implemented at the field, not just the file
level.
o Automatic START RESTART capability provides control of the
timing of updates.
o Users can dump very large data files onto disk without having
to first massage it with a data definition language. Those
files are then converted to ADABAS format in a fraction of
the time normally required. [1]
o A variety of user languages can accomodate many different
types of users. [1]
o The product has extensive network support facilities
including ADABAS/VTAM (provides network capabilities in
remote processors), Channel-To-Channel Software system (CTCS)
(provides network capabilities to local processors) and
ADABAS/VM (provides communications in virtual machine
environments. [1]
o The ADABAS/VSAM Bridge can migrate VSAM programs.
o Its "loose" file structure allows the database to support
unlimited modifications without having to be reloaded. All
of the changes are automatically captured by indexes. [1]
o The flexible structure of ADABAS allows a DBA to choose the
data base architecture that best fits an application's
requirements. [2]
o ADABAS does not need to design chain pointers and
predetermined paths, so reloading is relatively easy. [2]
o ADABAS performs best in a highly centralized environment. It
is built to handle the continuous operations of a highly
interactive environment. The ADABAS DBMS automatically
reclaims space when it becomes available and the storage
space in ADABAS does not have to be reorganized. As records
are updated, deleted, or added, the random access efficiency
capabilities of ADABAS are not degraded. All integrated
ADABAS components contain multithreading and multiuser
subtasks. The latter capability allows all system users to
share work areas. [2]
o All ADABAS's TP programs are user friendly and do not take up
large amounts of overhead. [2]
o ADABAS is a good DBMS that is pretty simple to learn and
master for all users. ADABAS has established a good
benchmark record that is getting better with each new
enhancement. [2]
o Software AG has done a good job in implementing distributed
data management. [3] Three optional products, ADABAS/VTAM,
the Channel-to-Channel Software System (CTCS), and ADABAS/VM,
permit the use of ADABAS in all types of distributed
environments. [5]
o ADABAS has a good data dictionary which offers an effective
approach to distributed data. [3] PREDICT, the ADABAS data
dictionary, allows users to access data from other DBMS's or
files and perform updates and changes. PREDICT is monolithic
and contains multithreading capabilities. [2]
WEAKNESSES (ADABAS)
o The query language, NATURAL, is an option which adds $40,000
- $60,000 to the base price of ADABAS. [1]
o ADAMINT is not available on VAX systems. ADAMINT supports a
MACRO interface between ADABAS and applications programs
written in any high-level language with a CALL facility. [1]
o Faces more competition from relational tools such as Ingres,
ORACLE, and Britton-Lee's IDM systems. ADABAS' future
limitation may be an inability to provide more of the same
query-type capabilities as it moves into more direct
competition with relational databases. [1]
o The lack of tightly linked data chains puts ADABAS at a
disadvantage with some competitors when operating in highly
interactive production environments. [2]
o ADABAS does not allow a DBA to create a physical table for a
user. [2]
o SUPERNATURAL, an end-user language, is not very spectacular.
[3]
1.2 ******** PRODUCT PROFILE ********
1.2.1 PRODUCT NAME:
Adaptable Data Base (ADABAS) (V.4.1), (ADABAS (VMS) V.1.1)
1.2.2 VENDOR:
Software AG of North America, Inc.
Reston International Center
11800 Sunrise Valley Drive
Reston, VA 22091
703-860-5050
1.2.3 TYPE OF SYSTEM
Inverted list with relational-like features. [5]
1.2.4 MODE OF OPERATION
Interactive
1.2.5 TYPES OF USERS
Large government agencies and industrial establishments.
1.2.6 NUMBER OF INSTALLATIONS
About 1,800 users worldwide. [5]
1.2.7 SOME INSTALLATIONS
1.2.8 COMPUTERS/OPERATING SYSTEMS SUPPORTED
IBM 370/125 and larger, 4300 series, 30 series; SSX/VSE, DOS/VS,
DOS/VSE, OS/VS1, OS/VS2(MVS) and CMS component of VM/370.
All equivalent plug-compatible systems produced by non-IBM
vendors.
Digital Equipment Corporation's VAX-11/750 and VAX-11/780; VMS.
Siemens 4004/45 and larger;; PBS, BS100, BS2000.
1.2.9 MINIMUM CONFIGURATION
ADABAS kernal requires a minimum of 320K bytes of virtual
storage. ADABAS (VMS) can be installed on DEC's VAX 11/725 - 11/785
and 8600 under the VMS operating system. VT100 and VT200 terminals
are supported. Dedicated working sets of .5M to 1M bytes are
required.
1.2.10 PRODUCT COMPONENTS
The components of the ADABAS DBMS are PREDICT, NATURAL,
ADASCRIPT, ADACOM, and ADAMINT. PREDICT, the data dictionary, tightly
integrates the four components.
Only ADABAS (VMS) and NATURAL (VMS) are available on VAXs.
1.2.11 HOST LANGUAGES
Any high-level language with CALL facility.
1.2.12 COMMUNICATIONS INTERFACES SUPPORTED
COM-PLETE; IBM CICS, TSO, DC component of IMS DB/DC and FASTER;
Polygon INTERCOMM; Westinghouse WESTI; or ADABAS (VMS), depending on
the operating system used.
NATURAL CONNECTION is an option that enales uploading and
downloading of data between a micro and a mainframe.
1.2.13 TERMS OF AVAILABILITY
Package licenses (except the VMS version, which is not available
for lease) can be purchased or leased. Purchases can be financed
through Software AG.
1.2.14 COST OF OWNERSHIP
License purchase ranges from $106,000 (the ADABAS kernal for any
DOS environment) to $242,000 (the ADABAS kernel version for IBM
OS/VS2(MVS) systems with NATURAL and PREDICT options).
ADABAS (VMS), which is bundled with NATURAL (VMS) costs $50,000.
Discounts of 25% to 60% can be negotiated for educational usrs and
quantity purchasers.
First year service is included in the license price. The cost of
service for five years ranges from $42,000 to $96,800.
1.2.15 SUPPORT
One year of free technical services and three days of free data
base administrator (DBA) on-site orientation in conjunction with
installation are provided for all leased or purchased licenses.
Continuing technical support is proveded for an annual fee based on
10% of the current license purchase price for each component. Ongoing
service maintenance costs 15% of current purchase price.
1.2.16 SECURITY AND INTEGRITY
Password file access; multilevel field access/update control;
field-value-dependent record access control; cipher facilities for
user-provided code.
1.3 ******** TECHNICAL SUMMARY OF PRODUCT CAPABILITIES ********
1.3.1 DATABASE DEFINITION
1.3.2 DATABASE STRUCTURE
o LOGICAL
The major index component is an inverted-file
associator. The inverted file scheme is an index by
descriptor (domain) rather than by record (tuple) identity.
An inverted file approach structurally supports generalized
queries such as "List all the sales personel who are female
and who make more $40,000." rather than direct queries such
as "How much does Jane Doe earn?"
A logical-address-to-physical-address conversion
technique permits the indexes to be decoupled from the data
and allows substantial data movement on disk with minimum
effect on the indexes.
Search criteria can be fully modulated through Boolean
and mathematical operators.
o PHYSICAL
All data is stored in tabular form and there are no
built in address pointers to superimpose a hierarchical or
complex overlay on the data. The tables or relations,
however, are not normalized and can contain repeating groups.
Data compression is standard. This in combination with
ADABAS' buffer management techniques, enhances processing
efficiency because disk input/output is reduced.
Data access is handled through EXCP; certain utility
functions use QSAM. In addition, an ADABAS direct access
method called ADAM can be used to optimize random access
operations on a uniquely defined description field.
In ADABAS, rows (tuples) are called records and columns
(domains) are descriptors. A record/descriptor intersection
produces a field (value of a data item). A record can
contain up to 500 fields, which can be defined by up to 200
descriptors.
Every record is identified by an ADABAS-assigned
Internal Sequence Number (ISN).
A file can contain up to 16.7 million records, and an
ADABAS database can consist of up to 255 files. ADABAS
supports multiple databases. Multithreading is supported for
2 to 20 concurrent threads. [5]
o DATATYPES
Alphanumeric
Binary
Fixed-point
Packed or unpacked decimal format
o PREDEFINED FUNCTIONS
ADASCRIPT includes the following capabilities:
Addition
Subtraction
Multiplicaton
Division
Counts
Floating point options
o INTEGRITY CHECKING
During program creation, NATURAL performs a syntax check
of each line as it is entered. Error messages are displayed.
The ADABAS Reflective Data Base Feature, introduced in
beta sites in 1984, is designed to eliminate the possibility
of interruptions to database processing caused by I/O errors.
With this feature, some of the blocks in the primary database
and/or all of the blocks in the primary protection log can be
maintained in mirror-image copies called "Reflective Data
Base." Then, in the case of errors, the reflective data base
can be used to restore data.
o SECURITY
Password security can be used by the DBA to control other
users' access to particular files, and, for each file, to
specify access or update privileges. Access to fields within
files can also be controlled.
Encryption can be used to make it difficult to read the
contents of a data base obtained from a physical dump.
1.3.3 DATA DEFINITION INTERFACES
o LANGUAGE
File entry/conversion/creation is handled by a nucleus
utility with an exit to user-written programs to accommodate
unusual data occurrences.
o PROCESSING OF LANGUAGE
In the interactive mode, a user is prompted for each
statement line. In the batch mode, all statement lines
contained in the input stream are read and interpreted by
NATURAL.
NATURAL operates as a standard transaction processing
(TP) program and follows the rules that apply to programs
executing under the control of the specific TP monitor in
use.
o DATA DICTIONARY
PREDICT, the ADABAS Data Dictionary, is an active system
that contains information that describes a data base's data
and its use. The description of data includes files, fields,
and the relationship between the two.
The Data Dictionary file is comprised of the Standard
Definition File, which contains commonly used fields and
their basic attributes; the Physical Definition File; and the
Userview File, which contains a data base's Userview
definitions. A "pull forward" facility permits entries in
the Standard Definiton file to be automatically inserted into
the other two.
1.3.4 DATABASE MANIPULATION
FUNCTIONS
o RETRIEVAL
ADASCRIPT enables users to retrieve and display
information from an ADABAS data base at the terminal. It
contains the following features: arithmetic, report
formatting, totals and count, calculator mode, code
translation, condition statements, control statements,
dynamic overrides, record selectoin, record sorting,
statistical option, etc.
================================================================================
Note 8.1 ADABAS (Merci P.Grogniet!) 1 of 1
LYO01::LACOMBE_D 99 lines 1-APR-1987 09:01
-< IS ADABAS RELATIONAL >-
--------------------------------------------------------------------------------
< adabas discussion on relational aspect >
my name is DIDIER LACOMBE
I am DATABASE consultant in LYON FRANCE
I used ADABAS 1 years ago on IBM system
ADABAS IS NOT RELATIONAL
------------------------------------------------------
1/ tables can have repeating group , so they are not under first normal form
, this option leads to much better performance then for a full
relational system, but... what happens with redundancy, and when repeating
group increase in size : rewrite your program .
repeating group ar limited to 99 , and the total memory size can' t grow up to
4k bytes for one record ,ADABAS can be seen as a hierchical dbms with
a limited level of hierarchies
this point violate CODD's first rule
2/ you can' t use ADABAS if you don't understand what an ISN is . an ISN
( internal system number ) is like a dbkey . You MUST use isn if you plan
to use COBOL . If you READ by KEY , then ADABAS returns you an ISN , this ISN
is used to read the record.
Access by ISN is very performant, and generally that the way they implement
bench_mark, first read by key , then store isn list in table and
continue precessing with isn only. But what happens if the table must grow:
they change the program. We can also use DBKEY with same performances.
this point violates another of the CODD'S rules : acces by data , without
link , vector or pointer.
3/ DDL is not dynamic, you cannot mixed DDL and DML. You can't do
CREATE TABLE
CREATE VIEW
this point violates another of the CODD's rules : DDL and DML must be
at the same level.
4/ Metatada are not under relational form : metata data can't be accessed
by DML.
this point violates another of the CODD'S rule: DATA and MATADATA must be
in the same organisation ( RELATIONAL IS SELF CONTAIN )
5/ CREATE VIEW is impossible, you must specify a KEY to access records.
this point violates another of the CODD'S rule : program and data must be
LOGICALY independent, any field can be a key at any time, a program is
never concerned by the logical organization, VIEW is necessary to keep
logical independance.
6/ Under release 4 they were no INTEGRITY CONSTRAINT, check that under the
current release.
7/ They offer only a level 2 consistency, because they have only the PROTECTED
WRITE LOCK , not the PROTECTED READ.No SNAPSHOT is available.
in record creation they offer only the level 1: you can always read a
record stored by a not yet committed transaction. remember that RDB offers
a level 3 consistency.
THEY PRETEND TO BE RELATIONAL BECAUSE:
they just manage files ( and a file is a table )
they don't have links or rings or sets
they can do boolean operation on ISN LIST ( LIKE with KEEPLIST)
they can do find customer gt 'balbla' and lt 'wwww' or eq ' '
with some kind of optimization to find the best strategy.
they can create new fields in batch mode only.
MAIN STRENGTH
--------------
BENCH MARK IN FRANCE ORGANIZED BY DECUS , FOUND MUCH , MUCH BETTER PERFORMANCES
COMPARED TO RDB AND OTHER RELATIONAL ( ORACLE INGRES) DATABASES PROBABLY BECAUSE
THEY USE REPEATED GROUPS IN TUPLES INSTEAD OF USING JOIN, AND BECAUSE THEY USE
ISN.
IN FACT , ADABAS MUST BE COMPARED TO RMS WITH RECOVERY , FOR IT'S
ONLY A MULTI KEY TOOL.
OTHER WEAKNESSES:
-----------------
VERY COMPLEX TO USE UNDER COBOL , MORE THAN 10 PARAMETERS MUST BE SENT
OTHER REPLY EXPLAIN THEY DON' T SUPPORT CLUSTER
NATURAL IS LIMITED TO DO SIMPLE THING SIMPLY
DIDIER LACOMBE
LYO01::LACOMBE_D
DATABASE CONSULTANT LYON FRANCE
|