Note: Descriptions are shown in the official language in which they were submitted.
21 86777
Field of the Invention
The invention concerns a method of linking data processing units which
each use an independent opel~lhlg system. It is also directed to a method of
controlling an exchange, wherein two or more data processing units use differentOpel~ g ~y~lenls that cooperate to perform ~wilchh~g functions. It is also
directed to a data processing unit conlplising at least one coll~uler system as a
har.lwale platform, an independent opelaling system, and a plurality of control
programs, wllelèill at least one of the control programs coopel~tes during its
execution with at least one second data processing unit that is provided with anindependent operation system, a controller comprising at least one first data
processillg unit and a second data processing unit, the first and second data
processing units being each provided with an independent opelalillg system and
- with a plurality of control programs, at least one control program of the first data
processing unit being designed to cooperate with the second data processing unitduring its execution. It is still further directed to a controller having at least first
and second data processing units each with an independent opel~lh-g system, and
with control programs, with at least one control program of the first data
processing unit coopel~lillg with the second data processing unit by means of data
bases. It is also directed to an exchange colllplising a ~wilehillg nclwolh and such
a controller.
Descliplion of the Prior Art
To link the sequences or processes of a colll~,uler system it is necessary
for them to c(JIlllllullicdlè with each other. Sequence in this case means the
operation of a program, and process means the operation of a program if this
sequence is managed by an opel~lh~g system (also see chapter 1.2 Processes, page15 of the book 'lo~elalillg System: An Introduction" by H.-J. Siegert, OldenburgPublishers, 1991).
Mech~ni~m.~ to enable the linking of processes managed by the same
c,pelaling system are described for example in the chapter "Process
21 86777
Co~ llunication" of the book "Ol~e~ g System: An Introduction" by H.-J.
Siegert, Oldenburg Publishers, 1991 on pages 81-96).
The con~lllullication belweell the processes, which is also called
h~lel~locess co...~ ic~lion, takes place on the one hand through the exchange ofmessages by means of services of the opel~lillg system. On the other hand, it isalso possible to realize the conl,llullicdlion without support and knowledge of the
oper~lillg system by exch~ngi~-g messages via common memory areas provided by
the opel~ g system. The processes that take part in the collnnllllication can also
operate in dirrelelll COIll~u~ , if the opelaling system is distributed to several
colllpul~l~. However, these mech~nicm~ can only be used for coll~ln~ icalion
b~lw~en plvcesses managed by the same operatillg system, and cannot be used for
coll~lllunicalion belweell processes managed by different operating systems.
It is furthellllol~ known to manage data for a number of processes
centrally by means of a data base. The data base enables the processes to jointly
access the data in accordance with the data model implemented in the data base.
In this case it is also possible to access such a data base through another conl~ulel.
Summary of the Invention
The invention starts with the controller for an ISDN (Integrated Services
Digital Network) co--.-----nicalion system, which is described in the German Patent
Application DE 41 04 365 C1, wll~rein processes managed by different opel~lh~g
sy~Lelns comml-nic~te with each other.
This controller has two IJpelaling sy~l~lns which work with the same
col~ t~ r. On the one hand, this is a subscliber ope.~ti,lg system for subscriber-
individual data processing, and on the other it is a real time opel~tillg system for
teleco.. ~.inic-lion tasks. These opelali.lg systems are linked with each other in a
way so that the data or data applications managed by the real time operating
system can be ~upelilllposed on the access to the processes of the subscriber
opel~ling system, and on the other hand the data applications of processes of th
~1 86~7
.
real time operating system con-llJcted under the subscriber operating system can be
taken over. In that case the communication belween processes managed by
dirrelcnl opel~lhlg systems takes place through the exchange of messages via a
common memory segmelll.
This common memory seg,.,c"l, which is described as an
intercomm--nications data seglllelll, is established by two equal data segment
scriplols, which make this memory segment useable by the sequences and
processes of both upcl~lillg ~y~lcllls.
This type of linkage has the disadvantage that at least one coulpuler must
operate in parallel with both operating systems in order to enable the access byboth operating systems to the common memory segment.
It is furthermore disadvantageous that any llan~rcr of data belwcen
processes of different ope.~lillg systems requires precisely tuned, e~pellsive
ll~n~rcr procedures for all particil~alh~g processes. This is colu~ecled to a high
imple,llenlation cost and complex dependencies bclwcen control programs which
operate in dirrerclll operating systems.
The invention now has the task of enabling the construction of a
controller for data processing units which operate with independent opelating
S,~ lllS.
In this h~l~llce data processing units mean a logic unit which is formed
by an opclalillg system and the control programs lu~uung in this operating system,
and by the underlying haldwdl~ platform. As a rule, a data processing unit is a
colll~uler with pellhlelll peripheral components, which uses an independent
opelalhlg system. However, it may be a colll~ulcr system with a distributed
opelalillg system, or also a logical or functional part of a coll~ ler, if the
colll~ulcr uses several independent opel~lhlg ~y~lcllls.
21 86777
` ~ This task is fulfilled by a method of linking data processing units that
each use an independent operating system, wheleill for each of the data processing
units, an independent data base is provided which manages at least part of the data
of the l~s~eclive data processing unit in accordance with a respective data model,
that a ffrst data base of a first data processing unit and a second data base of a
second data processing unit are so linked together that the first data base has
access to a record of data of the second data processing unit which are managed
- by the second data base, and makes the data processable for processes of the first
data processing unit in accoldance with their data model, and that for the
cooperation ~lween the first and second data processing units, processes of the
f1rst data pl~cessil~ unit access data of the record of the second data processing
unit by accessing the first data base.
The task is also fulfilled by a method of controlling an exchange whelei
two or more data processillg units using different opelating systems cooperate in
performing ~wi~chin~5 functions, that in each of the data processing units, an
indepel~delll data base is provided which manages at least part of the data of the
respe~ e data processin~ unit in accordance with a resl)e~;live data model, that a
first data base of a first data processing unit and a second data base of a second
data processing unit are so linked together that the ffrst data base has access to a
record of data of the second data processing unit which are managed by the second
data base, and makes said data processable for processes of the first data
processing unit in accoldallce with their data model, and that for the cooperation
b~Lweel- the first and second data processing units, processes of the first dataprocçscing unit access data of the record of the second data processing unit by
accessing the first data base.
This task is further fulfilled by a data processhlg unit comprising at least
one co~ ulel system as a haldwale platform, an independent opel~ g system,
and a plurality of control programs, at least one of the control programs being
desi~nPcl to cooperate, during its execution, with at least one second data
proces.sing unit which is also provided with an independent operating system,
!
21 86777
. ~
characterized in that the first data processing unit is provided with a first data base
for m~n~ging data in accordance with a first data model, that the first data base is
linked with a second data base of the second data processing unit, the second data
base being designed to manage at least part of the data of the second data
processing unit in accordance with a second data model, that the first data base is
designed to access a record of data of the second data processing unit which aremanaged by the second data base, and to make said data processab1e for processesof the first data processing unit in accordance with the first data model, and that
for the cooperation with the second data processing unit, the control program isdesiglled to access, during its execution, data of the record of the second dataprocessing unit by accessing the first data base.
It is also fulfilled by a controller comprising at least one co~ ,ulcr system
as a hal-lw~rc platform and at least a first data processing unit and a second data
processing unit, the first and second data processing units being each provided
with an independent opel~ g system and with a plurality of control programs, at
least one control program of the first data processing unit being designed to
cooperate with the second data proces~ing unit during its execution, characterized
in that the first and second data processing units are provided with a first data base
and a second data base, lespec~ivcly, which are designed to manage at least part of
the data of the lcspe~;live data processing unit in accordallce with a respective data
model, that the first data base is linked with the second data base, that the first
data base is designed to access a record of data of the second data processing unit
which are managed by the second data base, and to make the data processable for
processes of the first data processing unit in accordance with their data model, and
that for the cooperation with the second data processing unit, the control program
is designed to access, during its execution, data of the record of the second data
ces.~ g unit by accessing the first data base.
It is also fulfilled by an exchange COlllpliSillg a swilching network and
such a controller.
2 1 8~777
The basic idea of the invention is to manage at least part of the data
of each data processing unit in a lcspeclivc data base, and to link these data bases
in such a way that the data of the res~e~;live other data processing unit are made
visible (processable) to processes of the re~eclive data processing unit in
accordance with the data model of the respective data base. To the processes this
a~a~ as if these data were managed by the data base of their own data
processing unit. This linking of the data bases provides access to the data of one
of the other data processing units. The manipulation or hllel~lc~lion of these data
enables a deep access possibility to the processes managed by another operating
system using these data, and thereby a deep access possibility to the other dataprocessing unit.
It is advantageous to use this basic idea for the construction of controllers
in exchallges. The tasks to be fulfilled by such controllers make very differentdemands on the opel~tillg system that must be used. Thus, there are advantages if
such a controller is built of several data processing units, which use differentopelalillg sy~lc-.ls and are linked for cooperation in accordance with the abovebasic idea.
In this case the resulting advantages are particularly the reduction of the
software implementation and testing costs, since operating ~y~Lcllls, data bases and
programming languages, which are adapted to the lcspecli~e type of task, can be
used.
The invention is advantageous in that it provides multiple viewing and
ncces~ g possibilities for the sequences or plocesses of a different opel~ting
system, where said viewing and accessing possibilities need not to be precisely
defined in advance by interchange lvulilles that are ~yncl~lu~ ed with each other.
This reduces the dependencies which must be observed when developing the
c~,llc~ollding control programs. Even a subsequent need to expand the viewing
and accesshlg possibilities to another data processing unit only involves a low cost.
2 1 86777
`
In this way the cost of developing, testing and caring for controllers, which
colllplise several coope~ g data processing units, is
allGgetller reduced.
Another advantage of the invention is that it makes it possible to separate
the development cycles of different coopel~ g data processing units. This
advantage is the result of that, based on the linkage of the invention, a further
development of the control programs of a data processing unit only has minor
effects on the control pl~gl~lllS of other data processing units which cooperatewith the former.
Descliplion of the Drawin~s
The invention is explained in greater detail in the following by means of
three configuration examples with the aid of the attached drawings.
Figure 1 is a block diagram of an exchange with a controller according to the
invention.
Figure 2 is a functional illustration of the controller according to the
invention in figure 1 for a first configuration example.
Figures 3a
and 3b are functional illustrations of a linkage of two data processing units
of the controller according to the invention in figure 2.
Figure 4 is a functional illustration of the controller according to the
invention in figure 1 for a second configuration example.
Figure 5 is a functional illustration of the controller according to the
invention in figure 1 for a third configuration example.
Detailed Des~ ion of the Preferred Embodiment
The first configuration example describes the use of the method of the
invention for controlling an exchange in an exchange according to the invention,whose controller is built of two data processing units according to the invention,
~1 ~b~
-
which are linked to each other in accordance with the method of the invention for
linhing data processing units.
Figure 1 illllsl.ales an exchange EX with a controller CONTR and a
switching ~ WO1k DSN. The controller CONTR and the swilehillg network DSN
interchange data.
The controller CONTR is used to control all exchange functions and
service ~lllibules that are provided by the exchange EX.
The switching I~Lwclh DSN is a usual digital ~wilchillg network for
exchanges. It is controlled by the controller CONTR and is used to both switch
colll-ections for subscribers of the exchange EX and as a co~ .nic~tions medium
for the internal co~ mication of the controller CONTR.
It is also possible to use an independent cc,.~ ullications nelw(JIh for this
internal collmlul~ication. Such a communications network can also be a network
with a distributed exchange process for example a LAN (Local Area Network)
with an ethernet or a token access protocol.
The controller CONTR contains four independent c~ uler ~y~lellls CS1
to CS4 which coll~l~ullicate with each other via the switching network DSN.
Each of the cc,lll~uler systems CS1 to CS4 conlplises a com~uLel and the
co~ c~ed peripheral components which are required to carry out the l~speclive
functions and to coll~llu.licate with the other compuler systems CS1 to CS4.
The col~.l,uler system CSl co--~plises a co...~uler which contains one or
more processors an internal coml,uler bus a central memory at least one mass
memory such as for example a hard disk and modules for the input and output of
data. In addition the co...~uler system CS1 has pe-iphel~l components to enable
- 21 86777
the man-machine co~ ,ullication and the exchange of data via the switching
~Lwolk DSN.
The co~ ,ul~r systems CS1 to CS4 are computer systems designed to
provide telecolmllullicalion functions, the ~llu~;lule of which can be found forexample in the article "Haldwdlc Structure", pages 135 to 147 of the m~g~7.ine
Electrical Coll..ll~ulications, volume 56 no.2/3,1981 and in other articles of this
volume.
Each of the colll~uler systems CS1 to CS4 forms a hal-lw~lc platform for
a number of control programs and for an operating system that particularly
~u~poll~ and manages the sequence of the control programs. The precise definition
of an ope~ g system can be found for example in the norm DIN 44300. The
control programs are stored in the respeclive colll~uler system and their sequence
determines the function of the l~peclive Coll~; uler system.
The colll~ er system CS1 works with an opel~thlg system OS1, and the
colllp~t~r~y~lems CS2 to CS4 with a disLlib~l~d opelalillg system OS2, where oneopel~lhlg system part OS2' works with each of the colll~uler systems CS2 to CS4.
Thus, the sequence of the control programs in the colll~uler system CS1
and the sequence of the control programs in conll)uler systems CS2 to CS4 are
jointly managed by the opelaling system OS2. In this way the con~puler system
CSl with the pellhlelll control ploglallls and opel~ lg system OS1 form a data
processing unit PU1, and the conll,ul~l systems CS2 to CS4 with the pellhle~l
control programs and the distributed operating system OS2 form a further data
processillg unit PU2. The operating systems OS1 and OS2 are independent
operatiilg sy~ ls, i.e. they autonomously manage the sequence of the control
programs and are not the operalillg system parts of a higher distributed upelalhlg
system.
1_ 21 86777
It is also possible for the data processing unit PU2 to colllp~ise only one
collll,ul~r system. It is furthermore possible that the data processing units PU1
and PU2 are logic data plvcessillg units, and although each of their processes is
managed by an independent opelalhlg system, they have however a common
hardware platform. In that way the colll~ler system would operate in parallel
with the two opel~ g systems OSl and OS2.
The functional structure and the interaction of the data processing units
PU1 and PU2 will now be explained by means of figure 2.
Figure 2 illustrates the controller CONTR with the data processing units
PUl and PU2. The data processing unit PU1 has three processes P1 to P3, a data
base DBMS1 and the opcl~lhlg system OS1. The data processing unit PU2 has
three processes P4 to P6, a data base DBMS2 and the ~pelatillg system OS2. The
processes Pl to P3 conllllullicate with each other and with the data base DBMSl.The processes P4 to P6 CO~ llulliCal~ with each other and with the data base
DBMS2. The data bases DBMSl and DBMS2 are interlinked.
Each of the processes Pl to P3 r~leselll~ the execution of a control
program by the con~ ller system CSl, e.g. the execution of a certain task by theC(JIII~UI~I system CSl. The plucesses Pl to P3 are the processes which provide
functions within the framework of the switching tasks of controller CONTR. In
this i~ e the number of three prucesses Pl to P3 is only chosen as an example
and usually always changes as a function of the status of the data processing unit
PUl. The same applies to the processes P4 to P6.
The ope-~thlg system OS1 manages the processes P1 to P3 and
particularly determines the sequence of their pelrollllallce, it starts them,
delellllilles their status and provides them with ~l~lldald services which particularly
carry out the llal~l~oll~lion of data between processors and between processors
and devices of the conlyuler system CS1. Ful~ llore, during the ~im~ n~ous
sequence of several processes the operating system OS1 coordinates and
!
21 86777
~ynclll~ cs the access to common operating means, such as for example the
~ .
cGIll~uler core, memories or devices. The same applies to processes P4 to P6.
In particular, the opela~ g system OS2 is a real time opelathlg system
which is especially designed to provide the functions needed to control an
exchange. By contrast the opelalillg system OSl is a non-real time operating
system, such as is normally used in general purpose co",pulc,~. An example of
such an opc,~ting system is the opcl~ting system sold under the designation Unix.
It is also possible for the operating ~y~lc,,,s OSl and OS2 to be two
opcl~ting ~y~lcllls of the same type.
The data base DBMSl manages part of the data of data processing unit
PUl with the knowledge of the ~cs~ecli~/ely implemented data model. This is an
independent data base, i.e. it is not part of a higher distributed data base. These
data are for example semipcllnanelll variables or data files which are used by the
control l"c,gl~"ls of data processing unit PUl. Access to these data is obtained in
that the processes Pl to P3 request or manipulate the data from the data base
DBMSl by means of special messages.
From the logic point of view, in this way the data base DBMSl
rcpl~csellls one or more special processes which manage data and make them
available to processes Pl to P3 upon request, or manipulate them upon their
request. Co~ llunicalion bclwccn the processes Pl to P3 and the data base
DBMSl is supported in this il~nce by services of the OPC1~ g system OSl,
which particularly enables the di~llibulion of the processes Pl to P3 and the data
base DBMSl to dirrerclll colll~lcr syslellls which work with a common
distributed opelalillg system. The same applies to the data base DBMS2.
The data bases DBMSl and DBMS2 are dirrcrclll data bases. These data
bases are chosen in accordance with the lcspeclive system platform, i.e. of the
coll,l,uler platform and that of the opelalillg system, and in accordance with the
21 ~36~77
functions to be provided by the data processing unit PUl or PU2. Thus the data
base DBMSl is a conventional relational data base, for example according to the
SQL ~lalldald. Such a data base is for example the data base sold under the
~lesign~tion ORAKEL. In co~ as~, the data base DBMS2 for example is a special
distributed data base which is optimized to provide ~wilchillg functions. However,
it is also possible for the two data bases DBMS1 and DBMS2 to be data bases of
the same type.
In particular, each of the data bases DBMSl and DBMS2 has an access
i,llclrace INTl or INT2, a data base manager DBM1 or DBM2 and a data base
DBl or DB2.
The processes Pl to P3 access the data base DBMSl via access interface
INTl. The processes P4 to P6 access the data base DBMS2 via access interface
INT2. The data base managers DBMl and DBM2 have access to the data bases
DB1 or DB2 and conlll~ullicate with each other.
- The access interfaces INT1 and INT2 make available the interfaces
belweell the data bases DBMS1 and DBMS2 and the processes Pl to P3 or P4 to
P4, which are also called application processes in this connection. In this way
they support for example a data base access language, whereby access to the
pe~ elll data base is made possible. For example such data base access languagesare the data base access 1anguages SQL (Structured Query Language) or NDL
(Network Database Language). These can also be special ploplielal~ data base
access languages or access protocols.
The data base DBl is lcl,~senlt;d by a storage medium wheleill the data
managed by the data base DBMS1 are physically stored. This storage medium is
one or more mass memories of the collll~uler CSl, for example hard disk drives.
The same applies to the data base DB2 which manages the data of data base
DBMS2.
21 86777
-
It is therefore possible that the data base is distributed in the co~ uler
systems CS2 to CS4, i.e. that in each case part of the data are stored in the
storage media of coll~uler systems CS2, CS3 or CS4.
The data base m~n~ers DBMl and DBM2 perform the actual m~n~ging
functions for the data of data bases DBMS1 or DBMS2 which are stored in data
bases DB1 or DB2. They d~ te~ e the possible access mech~ni.cm~ for the data
and manage the data model in accordance with which the data can be accessed.
With the knowledge of this data model they perform the actual accessing
operations to the data stored in the respective data base.
In addition the data bases DBMS1 and DBMS2 are interlinked via data
base managers DBMl and DBM2. This linkage makes a selected record of data
from data processing unit PU2 visible for the processes P1 to P3 accordillg to the
data model managed by the data base manager DBM1, i.e. the record becomes
processable. This means that the processes P1 to P3 can access this record of data
by means of the access mech~ni~m~ of data base DBMS1 and in accordance with
its data model. In this way the access to this record of data does not differ from
an access to the data of its own data processing unit PUl. The same applies in
reverse for processes P4 to P6.
The ~7wilchillg functions are provided as follows by the controller
CONTR:
The data pl~cessillg unit PU2 is responsible for p~lrolllling ~7~lldard swilchhlg
tasks or basic ~7wilching tasks in real time. In this case the individual tasks are
provided by the processes P4 to P6.
The data processing unit PU1 is responsible for providing more complex
~7wil~hing functions, for which there are no specif1c real time requirements. The
task of such functions conlplises for example providing ISDN features7 s.lbscliber
~alures, centrex selvices, the detailed call registration or applications or services
2~ 8-6~
for mobile radio or intelligent nelwolk~. These tasks are provided by the
plocessols Pl to P3.
The data which plocesses P1 to P6 need to fulfill their tasks are managed
by the data base DBMSl or DBMS2. Such data are for example semipermanent
variables of the control programs whose sequences are l~pleselll~d by the
processes P1 to P6, subscliber data, data about the status of the switching network
DSN, data about telecolmlllmications services to be provided, or data about the
traffic load of exchange EX. The linkage of the data bases DBMS1 and DBMS2
makes these data available to all plocesses P1 to P6 and they are able to accessthese data by means of their usual access mech~nicm~ in accoldallce with the data
model of their data base, regardless of which of the two data bases manages saiddata. In addition, such data are also managed by the data bases DBMS1 and
DBMS2 and are made visible via the linkage to the processes of the other data
processing unit, which delellllille the status of the processes, or whose changealters the sequence or the result of the processes. The extent to which data aremade visible via the linkage to processes of the respective other data processing
- unit is del~llnined by the depth of the mutual access possibilities.
It is also possible for the data bases DBMS1 and DBMS2 to manage other
data of data processing units PU1 or PU2. Such data could for example be data
which are used by system processes of the opelaling ~y~l~llls US1 or US2, and
which influence their sequence. They could also be parameters of the operating
system core, m~king it possible to directly influence the sequence control of the
re~l,eclive other opelaling system.
It is rullh~"llore possible that data of the other operating system, made
visible via the linkage, are copied to an object-oriented information model, and the
cooperation of the two data processing units PU1 is achieved with an object-
oriented application.
2 1 ~6777
-
The linkage of the two data bases DBMS1 and DBMS2 by means of a
replication mechanism will now be explained in more detail with the aid of figures
3a and 3b.
Figure 3a illustrates the data base manager DBM1 and the data processing
unit PU2, which interchange messages. The data base manager DBM1 has several
data base processes PDB, a collvelLer MAP, a data set DATMOD and a
co~ lunications facility KOM. The collllllullications facility KOM exchanges
messages with the data processing unit PU2, and receives messages from the data
base processes PDB via the converter MAP. The converter MAP has access to
the data set DATMOD.
The co~ ic~lions facility KOM provides the n~cessA.~ conllllullication
services for comlllullica~ion with the data processing unit PU2 and with the data
base manager DBM2.
The data set DATMOD contains meta-data which describe the data
models of the two data bases DBMS1 and DBMS2.
With knowledge of the data models of both data bases DBMS1 and
DBMS2, the collv~ MAP con~/~lls messages which concern the change of data
in a data model, into messages which concern the change of these data in the other
data model.
The data base plocesses PDB repleselll processes of the data base
manager DBM1.
The data base manager DBMl has knowledge of the a~ignment of
syncl l~ni~lion units to a predel~llllilled record of data managed by it. In this
re said record of data de~llllhles the data that are made visible to the data
proces~ing unit PU2. If orle of the processes P1 to P3 now executes a data base
access INSERT which causes data to change, the data base processes PDB, which
!
2~ 8677~
process this data base access, check whether these data have a ~ynchlo~ ion unitassigned to them.
If this is the case, they send a message TRIGGER to the converter MAP
and wait with storing the change in data base DB1 until they receive an
acknowledgement message ACK from the colllnlullications facility KOM. If this isnot the case, they do not store this change of data in the data base DB1.
The message TRIGGER describes this change of data in accordance with
the data model of the data base DBMS1 and is converted by the collvel~r MAP
into a Illessage STORE, which describes the corresponding change of data in
acculdallce with the data model of data base DBMS2. From the message STORE,
the co~ ications facility KOM produces collesl,oll~ g messages which cause
this change in the second data base, and then sends these messages to the data base
manager DBM2. If this change of data has been entered in the data base DB2, the
data base manager DBM2 sends corresponding acknowledgment messages, which
are received by the co-------l.~ir-~tions facility KOM and passed on to the data base
processes PDB as an acknowledgement message ACK. The acknowledgment
message ACK can also be omitted.
It is also possible for the data base DBMS2 to be a distributed data base
with several copies of the data to be changed, and in each case only a
predelel.~ led group of processes has access to one of the copies. It would be
advantageous for such a data base if the co~ llunications facility KOM knew about
the di~libu~ion of the replicated data. In that way it could send targeted data to
those parts of the data base manager DBM2 which are responsible for m~n~ging
the respeclive data base.
Figure 3b illustrates the data base manager DBM1 and the data processing
unit PU2, which hl~e,cllallge messages. The data base manager DBM1 complises
the data base p,ucesses PDB, the co"v~"er MAP, the data set DATMOD and the
co... nications facility KOM. The collllllunications facility KOM exchanges
16
21 ~36777
-
messages with the data processing unit PU2, and sends messages via converter
MAP to the data base processes PDB. The collvelLer MAP has access to the data
set DATMOD.
What was said with respect to figure 3a also applies accordingly to the
data base processes PDB, the conveller MAP, the data set DATMOD and the
conllllullications facility KOM.
The collllllullications facility KOM receives a message from the data
processing unit PU2, which informs it of the change of data in the data base
DBMS2. This message directs it to the converter MAP, which converts this
message into a nlessagc INSERT describing the respective change of data in
acco~lance with the data model of the data base DBMS2. This message INSERT
is sent to the data base processes PDB and causes them to store this change of data
in data base DB1.
The functions of the data base manager DBM2 are carried out in
accoldallce with figure 3a and figure 3b, with the dirr~cl~e that the converter
MAP is omitted there.
It is also possible that changes in the data of the data base DBMS2 are not
announced by data base manager DBM2 to data base manager DBMl. An internal
message from the data base DBMS2 would then be evaluated by the
co---.---~,~ic~lions facility KOM and passed on to the data base processes, if they
concern replicated data changes.
This replication method achieves that only the changes of special data are
informed to the other of the data base managers DBMl and DBM2, thus the
amount of data that must be exchanged for ~ynchloniG~lion remains small.
Furthermore, a high degree of synchrollis.n is achieved.
17
~ 8677~1
-
~ It is also possible that such a replication method does not ensure the
`A
syncluollisl,l of the data bases, but that all the data of data bases DBMS1 and
DBMS2 needed for the cooperation are exchanged at regular intervals, for
example. However, the use of a replication mechanism has the advantage that the
data processing units PU1 and PU2 are linked by a linkage in such a way, that itcan support high speed requirements.
It is also possible not to use the controller CONTR to control the
exchange EX, but for any other control tasks. The system platforms and the task
distributions of the data processing units PU1 and PU2 would then also change
accordingly.
Another possibility of the linkage of data bases DBMS1 and DBMS2 will
now be explained with a second configuration example by means of figure 4.
Figure 4 illustrates the data processing units PU1 and PU2 with the
processes P1 to P3, the data base DBMS1 and the opel~ling system OS1, or with
the processes P4 to P6, the data base DBMS2 and the operating system OS2. The
data bases DBMS1 and DBMS2 colll~lise the access interface INT1, the data base
manager DBM1 and the data base DB1, or the access interface INT2, the data
base manager DBM2 and the data base DB2. The access interfaces INT1 and
INT2 hltelcllange data.
The data processing units PU1 and PU2 are constructed as shown in
figure 2. Only the linkage of the data bases DBMS1 and DBMS2 is achieved with
a dirr~lellt nlecl-~ m.
The data needed to influence the other data processing unit are not stored
in its own data base as with the first configuration example, and they are
~yllchro~ ed by means of a replication mech~nism In this configuration example,
access to these data takes place through the exchange of messages betweell the
access interfaces INT1 and INT2. If the access interface INT1 receives a request
18
21 ~6~7~
'
` . ~ for access to data from the data processing unit PU2 which corresponds to the
access language SQL of the data base, the access interface INTl recognizes this,then pelrolms a conversion of the data model and passes the request on to the
access interface INT2. The latter requests the corresponding data from the data
base manager DBM2 and passes the data on to the access interface INT1 as soon
as they have been received.
With such a linkage of data bases DBMSl and DBMS2 it is advantageous
if the data bases DBMSl and DBMS2 are of the same kind and the access
interfaces INTl and INT2 support the same data base access language. If this is
not the case, a full conversion may not be possible under certain circumstances
between such data base access languages.
A further possibility of linking the data bases DBMSl and DBMS2 will
now be explained with a third configuration example by means of figure 5.
Figure 5 illustrates the data processing units PUl and PU2 with the
plocesses Pl to P3, the data base DBMSl and the opel~ g system OSl, or the
plucesses P4 to P6, the data base DBMS2 and the opel~lhlg system OS2. The
data bases DBMSl and DBMS2 have an access interface INTl, a data base
manager DBMl and a data base DBl or an access interface INT2, a data base
manager DBM2 and a data base DB2. In addition both data bases have a common
data base DB3.
The data processing units PUl and PU2 are con~ d as shown in
figure 2. Contrary to the first configuMtion example however, the data processing
units PUl and PU2 use a common hardware platform, i.e. both data processing
units PUl and PU2 work with the same colll~uler system.
In this case the linkage of data bases DBMSl and DBMS2 is achieved
through the common data base TB3: all the data required to influence the
e~;live other data processing unit are stored in the data base TB3. Since both
19
21 86777
i~ ~
data base managers DBMl and DBM2 are able to access this common data
plOCeS!3il1g unit, data of the other data processing unit can be accessed in this way
in accoldallce with its own data model. Syllcl~Jni~illg the access requires a
~yllchlo~ tion mech~ni~m, such as is used for example by ope,~ting systems for
access by several processes to an opelalhlg means used in common.
