This section covers the following topics:
The Natural RPC facility enables a client Natural program to issue a
CALLNAT statement to
invoke a subprogram in a server Natural. The Natural client and server sessions
may run on the same or on a different computer. For example, a Natural client
program on a Windows computer can issue a CALLNAT statement
against a mainframe server in order to retrieve data from a mainframe database.
The same Windows computer can act as a server if a Natural client program
running under, for example, UNIX issues a CALLNAT statement
requesting data from this server Natural.
Natural RPC exploits the advantages of client server computing. In a typical scenario, Natural on a Windows client computer accesses server data (using a middleware layer) from a Natural on a mainframe computer. The following advantages arise from that:
The end user on the client side can use a Natural application with a graphical user interface.
A large database can be accessed on a mainframe server.
Network traffic can be minimized when only relevant data are sent from client to server and back.
The Natural RPC (Remote Procedure) Call offers the following modes of operation:
non-conversational mode (in the following texts this mode is meant unless otherwise specified)
These modes are described in detail in the following sections. For a comparison of the advantages and disadvantages of these modes, refer to Conversational versus Non-Conversational Mode.
You can use the Natural RPC on various platforms under the following operating systems:
z/OS
z/VSE
BS2000/OSD
Natural RPC on mainframes is supported under the following TP monitors:
Com-plete
CICS
IMS TM
TSO
TIAM
openUTM
Also, it is available in batch mode.
Windows
UNIX
OpenVMS
On all of these platforms, Natural can act as both client and server.
Non-Natural environments (3GL and other programming languages) are supported on the client and the server side. Thus, a non-Natural client can communicate with a Natural RPC server, and a Natural client can communicate with a non-Natural RPC server. This is enabled by the use of the EntireX RPC.
The non-conversational mode should be used only to accomplish a single exchange of data with a partner. See also Conversational versus Non-Conversational Mode.
The Natural RPC technique uses the Natural statement
CALLNAT, so that both
local and remote subprogram calls can be issued in parallel. Remote program
calls work synchronously. As a remote procedure call, a CALLNAT
would, simply speaking, take the following route:
The CALLNAT issued from the Natural client is routed via a
middleware layer to the Natural server which passes data back to the
client.
Usually, the middleware layer consists of the Software AG product EntireX Broker which uses the ACI protocol. EntireX Broker uses either Entire Net-Work or TCP/IP as communication layer.
A detailed example of the RPC control flow is described below.
CALLNAT
control flow details in a remote procedure are illustrated below. For greater
clarity, the return path is not shown, but it is analogous; the numbers refer
to the description:
From the Natural client, the program PGM1 issues a
CALLNAT to the
subprogram SUB1. PGM1 does not know if its
CALLNAT will result in a local or in a remote
CALLNAT.
As the target SUB1 resides on a server, the
CALLNAT accesses an interface object SUB1
instead. This client interface object has been created automatically or
manually (by using the SYSRPC utility's interface object
generation (IG) function).
The interface object has the same name as the target subprogram and
contains parameters identical with those used in program PGM1 and
in the target subprogram SUB1 on the server. It also contains
control information used internally by the RPC.
If the AUTORPC
profile parameter is set to
ON and Natural cannot find the subprogram in the local
environment, Natural interprets this as a remote procedure call and generates
the parameter data
area (PDA) dynamically during runtime.
Natural also tries to find this subprogram in the
service
directory NATCLTGS.
For further information on the SYSRPC interface object
generation function, see Creating Interface
Objects.
If you want to work without interface objects, see Working with Automatic Natural RPC Execution.
The interface object then sets up a CALLNAT to an RPC
client service routine.
The client RPC runtime checks in the service directory
NATCLTGS on which node and server the CALLNAT is to
be performed and whether a logon is required.
The CALLNAT data including the parameter list and, if
required, the logon data are passed to a middleware layer.
In this example, this middleware layer consists of the Software AG
product EntireX Broker. Therefore, the CALLNAT data is first
passed to an EntireX Broker stub on the client.
From the EntireX Broker stub, the CALLNAT data is passed
to the EntireX Broker. The EntireX Broker is a product that can reside:
on the client computer
on the server computer or
on a third platform.
For the data to be passed on successfully, the server
SRV1 must be defined in the EntireX Broker attribute file and
SRV1 must be already up, thus having registered with EntireX
Broker.
For information on how to define servers in the EntireX Broker attribute file, see the EntireX Broker documentation.
From the middleware layer, the CALLNAT data is passed on
to the EntireX Broker Stub on the Natural Server platform and from there to the
RPC server service routine.
The RPC server service routine validates the logon data (if present) and performs a logon (if requested).
The RPC server service routine invokes the target subprogram
SUB1 and passes the data, if requested.
At this point, the target subprogram SUB1 has all the
required data to execute just as if it had been invoked by a local program
PGM1.
Then, for example, the subprogram SUB1 can issue a
FIND statement to the
server's Adabas database. SUB1 does not know whether it has been
started by a local or by a remote CALLNAT.
Adabas FINDs the data and passes them to
SUB1.
Then, SUB1 returns the Adabas data to the calling
server service routine. From there, it is passed it back to PGM1
via the middleware layer. It takes the same route as described in Steps 1 to 8,
but in reverse order.
A conversational RPC is a static connection of limited duration between
a client and a server. It provides a number of services (subprograms) defined
by the client, which are all executed within one server task that is
exclusively available to the client for the duration of the conversation. It is
implemented in a program using an OPEN
CONVERSATION statement and a
CLOSE CONVERSATION
statement.
Multiple connections (conversations) can exist at the same time. They are maintained by the client by means of conversation IDs, and each of them is performed on a different server. Remote procedure calls which do not belong to a given conversation are executed on a different server, within a different server task.
During a conversation, you can define and share a data area called context area between the remote subprograms on the server side. For further information, see Defining Context Variables for Natural RPC in the Natural Statements documentation.
A conversation may be local or remote.
OPEN CONVERSATION USING SUBPROGRAM 'S1''S2'
CALLNAT 'S1' PARMS1
CALLNAT 'S2' PARMS2
CLOSE CONVERSATION ALL
Both subprograms (S1 and S2) must be accessed
at the same location, that is, either locally or remotely. It is not admissible
to mix up local and remote CALLNATs within a conversation. If the
subprograms are executed remotely, both subprograms will be executed by the
same server task.
Analogously to non-conversational RPC CALLNATs,
conversations may first be written and tested locally and can then be
transferred to the servers.
If you execute subprograms locally, the following rule applies:
A subprogram may not call another subprogram which is a member of the conversation.
Other subprograms not listed in the OPEN CONVERSATION statement may
be called. They are however executed in
non-conversational
mode.
If you execute subprograms remotely, the following rule applies:
A subprogram S1 may call another subprogram
S2 which is a member of the conversation.
This CALLNAT will be executed in non-conversational mode
because it was invoked indirectly. Thus, the subprogram S2 does not have access
to the context area.
In a client-server environment where several clients access several
servers in non-conversational mode, there may be the problem that identical
CALLNAT requests from
different clients are executed on the same server.
This means, for example, that a CALLNAT 'S1' from Client 1
executes Subprogram S1 on Server 1 (S1 is writing a
record to the database). The transaction for Client 1 is not yet complete (no
END TRANSACTION)
when Client 2 also sends a CALLNAT 'S1' to Server 1, thus
overwriting the data from Client 1. If Client 1 then sends a CALLNAT
'S2' (meaning END TRANSACTION), Client 1 supposes its data
have been saved correctly, although in fact the data from Client 2's identical
CALLNAT were saved.
The diagram below illustrates this with two clients and two servers. In
such a scenario, you cannot control whether two identical CALLNATs
from two different clients access the same subprogram on the same server:
In the above example, CALLNAT 'S2' from Client 1 can access
subprogram S2 on Server 1 and on Server 2. CALLNAT
'S2' from Client 2 has the same choice.
Similarly, CALLNAT 'S1' from Client 1 could access
Subprogram S1 on Server 1 and on Server 2, while CALLNAT
'S1' from Client 2 has the same choice.
It is obvious that interference can be a problem here if the subprograms are designed to be executed within one server task context.
You can avoid the potential problems of a non-conversational RPC by defining a more complex RPC transaction in conversational mode:
You do this by opening a conversation. This involves the use of the
OPEN CONVERSATION statement on the client side, referring to
CALLNAT 'S1' and CALLNAT 'S2'. Opening such a
conversation reserves one entire server task (for example, Server 1) and no
other remote CALLNATs may interrupt this conversation on this
server before this conversation has been closed. In addition, you can define a
common context area for the two subprograms on the server side by using the
DEFINE DATA CONTEXT statement.
As a general rule, the following applies:
Use the conversational RPC to ensure that a defined list of subprograms is executed exclusively within one context.
Use the non-conversational RPC if each of your subprograms can be used within a different server task or if the transaction does not extend over more than one server call. The advantage of this is that no server blocks over a significant amount of time and you only need a relatively small number of server tasks.
A possible disadvantage of conversational RPCs is that you reserve an
entire server task, thus blocking all other subprograms on this server. As a
consequence, other CALLNATs might have to wait or more server
tasks must be started.
The database transactions on the client and server sides run independent
of each other. That is, an END
TRANSACTION or BACKOUT TRANSACTION executed on
the server side does not influence the database transaction on the client side
and vice-versa.
At the end of each non-conversational CALLNAT and at the
end of each conversation, an implicit BACKOUT TRANSACTION is executed
on the server side. To commit the changes made by the remote
CALLNAT(s), you have the following options:
Non-conversational CALLNAT
Execute an explicit END
TRANSACTION before leaving the CALLNAT.
Set the Natural profile parameter ETEOP to ON. This
results in an implicit END
TRANSACTION at the end of each non-conversational
CALLNAT.
Depending on the setting of the parameter
SRVCMIT,
the END TRANSACTION is executed either before the reply is sent to
the client (SRVCMIT=B) or after the reply has been successfully
sent to the client (SRVCMIT=A). SRVCMIT=B is the
default and is compatible with earlier versions of the RPC.
Conversational CALLNAT
Execute an explicit END
TRANSACTION on the server before the conversation is
terminated by the client
Set the Natural profile parameter ETEOP to ON. This
results in an implicit END TRANSACTION at the end of each
conversation.
Depending on the setting of the parameter
SRVCMIT,
the END TRANSACTION is executed either before the reply is sent to
the client (SRVCMIT=B) or after the reply has been successfully
sent to the client (SRVCMIT=A). SRVCMIT=B is the
default and is compatible with earlier versions of the RPC.
Before executing the CLOSE
CONVERSATION statement, call the application programming
interface USR2032N on the
client side. This will cause an implicit END TRANSACTION at the
end of the individual conversation.
The following Natural limits apply to each individual remote
CALLNAT execution on the Natural RPC server:
| Natural Profile Parameter | Type of Limit |
|---|---|
LT |
Limit for Processing Loops |
MADIO |
Maximum DBMS Calls between Screen I/O Operations |
MAXCL |
Maximum Number of Program Calls |
These limits are reset after each remote CALLNAT execution,
and you can enforce certain limits on each remote CALLNAT
execution.
This measure helps to avoid unpredictable error situations when the execution of a client request reaches one of the limits due to the processing of a previous client request.
Both subprograms S1 and S2 (shown in the
figure above) must be accessed at the same location, i.e. either locally or
remotely. You may not mix up local and remote CALLNATs within a
conversation. If the subprograms are executed remotely, both subprograms will
be executed by the same server task.
The following table provides an overview of important key terms used in
the SYSRPC Utility and the Natural RPC documentation:
| Term | Explanation |
| Client Stub | Obsolete term, see Client Interface Object below. |
| EntireX Broker Stub | Interface between the Natural RPC runtime and the EntireX Broker transport layer which exchanges marshalled data between client and server. |
| Impersonation | Impersonation assumes that access to the operating system on which a Natural RPC server is running is controlled by an SAF-compliant external security system. User authentication is performed by this external security system. Impersonation means that after the authentication has been successful and the user's identity is established, any subsequent authorization checks will be performed based on this identity. This includes authorization checks for access to external resources (for example, databases or work files). After successful authentication the user cannot "change his/her identity", that is, he/she cannot use a different user ID. See Impersonation in Using Security. |
| Interface Object | In earlier versions of EntireX, the term "stub" was also used to refer to application-dependent, Workbench-generated pieces of code for issuing and receiving remote procedure calls. These objects are now referred to as interface objects. |
| Client Interface Object |
Accepts the The interface object stub is the local subprogram via which the server subprogram is called. The client interface object has the same name and contains the same parameters as the corresponding server subprogram. |
NATCLTGS |
The name of the Natural subprogram generated with the
SYSRPC utility to implement the
service directory
(see below).
|
| Node Name | The name of the node to which the remote CALLNAT is
sent.
In case of communication via the EntireX Broker, the node name is
the name of the EntireX Broker for example, as defined in the EntireX Broker
attribute file, in the field |
| RPC Parameters |
All parameters available to control a Natural RPC are described in detail in the Natural Parameter Reference documentation. See the section Profile Parameters. |
SERVDIRX |
The name of the XML-formatted file (Natural text member)
generated with the SYSRPC utility to implement the service
directory (see below).
|
| Service Directory | The service directory contains information on the services
(subprograms) that a server provides. It can be locally available on each
client node, or it can be located on a remote directory server, which is
referenced on Windows, UNIX, OpenVMS by profile parameter
RDS, and on Mainframes by the
corresponding keyword subparameter RDS of profile parameter
RPC or parameter macro
NTRPC.
|
| Server Name |
The name of the server on which the In case of communication via EntireX Broker, the server name is the
name as defined in the EntireX Broker attribute file in the field
|
| Server Task | A Natural task which offers services (subprograms). This is typically a batch task or asynchronous task. It is identified by a server name. |