 |
Description  |
|
|
A portion of the disclosure of this patent document contains material which
is subject to copyright protection. The copyright owner has no objection
to the facsimile reproduction by anyone of the patent document or the
patent disclosure, as it appears in the Patent and Trademark Office patent
file or records, but otherwise reserves all copyright rights whatsoever.
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a network of data processing systems, and more
specifically to the interconnection of a plurality of data processing
systems between different network protocol domains, such as the different
network protocol domains of SNA and TCP/IP.
2. Description of the Related Art
A system having multiple domains has at least one data processing system
that is interconnected to at least two different data processing systems
through at least two different network domains, i.e. network protocol
architectures. A problem with multiple domains is the difficulty in
allowing communication between machines which are connected to another
type of network. For example, a data processing system utilizing SNA LU
6.2 as its network protocol can not automatically communicate with another
data processing system utilizing TCP/IP as its network protocol. Both SNA
LU 6.2 and TCP/IP are examples of stream protocols where data flows as a
stream of indeterminate lengths, and the bytes are delivered in the
correct order. The problem is routing a stream of bytes from a data
processing system that utilizes a reasonably equivalent protocol, such as
a stream protocol, to another data processing system that also utilizes a
reasonable equivalent protocol, such as the stream protocol of this
example, but wherein the two protocols are not
the exact same protocol, such as SNA LU 6.2 and TCP/IP.
It is known to solve the above problem at the application program level. An
application program which is running on a data processing system at one
end of the connection may be designed to utilize a specific network
protocol. In this case, it is known to modify the application in order to
reimplement the application to work over another protocol. This requires
changing the source program code of the original application by some
amount. Depending upon how the application program was originally
designed, this may require a substantial amount of changes to the program
code.
It is also known to solve the above problem by implementing the same
protocol on both machines. For example, in order to use an SNA transaction
application running in an SNA network, to apply transactions against data
processing systems utilizing a TCP network, one could reimplement that
transaction application against TCP by then putting TCP on the client data
processing system, put IP over SNA, and gateway between the two. The
client data processing system can then be implemented utilizing TCP/IP.
The problem with this approach is having to reimplement the application to
utilize the different protocol at one end of the network or the other.
This is especially burdensome if the application is large and complex.
There are some application level protocols that handshake back and forth
over SNA, e.g. 3270 SNA. These have their own data format with meta-data
in the data stream. There are other application level protocols, such as
Telnet over TCP, that talk back and forth that have meta-data and data in
the data stream. However, one can not get these two to talk together since
these two have different data and meta-data in their data streams.
If an application utilized one protocol, and that application were to run
on a data processing having a different protocol, knowing the data stream
format, one could write the client half of the application on the data
processing system utilizing the other protocol.
Therefore, in order to extend network connectivity, it is known to
reimplement the application to utilize the different protocol, put one
protocol on top of the other, and gateway between the two. It is also
known to build a larger network utilizing each type of protocol through
replication and duplication.
The term "sockets" is an application program interface (API) that was
developed for the Berkeley version of AT&T's UNIX.sup.1 operating system
for interconnecting applications running on data processing systems in a
network. The term socket is used to define an object that identifies a
communication end point in a network. A socket can be connected to other
sockets. Data can go into a socket via the underlying protocol of the
socket, and be directed to appear at another socket. A socket hides the
protocol of the network architecture beneath a lower layer. This lower
layer may be a stream connection model (virtual circuit), or a datagram
model (packet), or another model.
.sup.1 UNIX is licensed and developed by AT&T. UNIX is a registered
trademark of AT&T in the U.S.A. and other countries.
A stream connection model refers to a data transmission in which the bytes
of data are not separated by any record or marker. A virtual circuit
implies that there appears to be one communications end point connected to
one other communications endpoint. When the connection is established,
only those two end points can communicate with each other.
Sockets are typed by domain (address family or network type), and model
type (stream, datagram, etc.). If needed, the socket can be further
specified by protocol type or subtype. The domain specifies the addressing
concept utilized. For example, there is an internet IP domain, and also a
SNA domain for networks utilizing TCP and SNA, respectively. As used
herein, the word "domain" is used to refer to the address family of a
socket, and not to a domain-naming domain. A domain-naming domain is a
concept of a related group of hierarchical addresses, wherein each part of
the address is separated by a delimiter, such as a period.
Since a socket is specified by the domain, sockets do not allow cross
domain connections. This means that if an application program creates a
socket in the Internet (Darpa) domain, it can only connect to sockets in
that same domain. Note: "Darpa" is used to specify that Internet, short
for internetworking, is not only used herein both to generically specify
the internet layer of a particular protocol family which contains means
for forwarding, routing control, and congestion control, etc., but also as
a name for a particular implementation of an internet called the Internet
or the Darpa Internet, or the Arpa Internet. Another name for this
internet layer is the Internet Protocol (IP). TCP/IP is also commonly used
to refer to this protocol.
Originally, the requirement that a socket can only connect to sockets in
the same domain was a reasonable restriction. This simplified the program
code when there was only one really useful domain anyway. With the advent
of the usage of other domains (specifically SNA), cross domain connections
have become desirable. For example, cross domain connections would allow
mailers to transport mail among domains. Also, cross domain connections
would allow programs to communicate using the existing communication
networks.
SUMMARY OF THE INVENTION
It is therefore an object of this invention to automatically route
connections between data processing systems that utilize different
protocols, independently of said applications running on said data
processing systems.
It is a further object of this invention to route, at the socket level,
between two networks when a cross-domain connection attempt is detected.
It is a further object of this invention to facilitate the interconnection
between data processing systems by allowing socket based applications to
easily span across different networks.
It is a further object of this invention to communicate between data
processing systems in which one of the data processing systems utilizes
TCP/IP and the other data processing system utilizes SNA.
It is a further object of this invention to communicate between two data
processing systems via a third data processing system utilized as a TCP to
SNA gateway.
It is a further object of this invention to communicate through a
connection between two data processing systems both utilizing TCP on each
of their local Internets, by bridging the network connection with a long
haul SNA connection.
The system and method of this invention automatically routes a connection
between data processing systems, independently of an application running
on the data processing systems, having different network domains. The
preferred embodiment describes the cross domain interconnections with
reference to the different network domains of TCP (transmission control
protocol) and SNA (systems network architecture).
The routing is automatically performed at a layer which contains the
communication end point objects. In the AIX.sup.2 operating system, and
other operating systems based upon the Berkeley version of the UNIX
operating system, this layer is called the socket layer.
.sup.2 Trademark of IBM Corporation
An intermediate processing system is utilized to gateway between a
processing system utilizing a network domain such as TCP, and another
processing system utilizing a different network domain such as SNA.
Alternatively, the client data processing system can be implemented
utilizing TCP/IP which can then be gatewayed through socket routing on the
same machine into an SNA data stream without an intermediate processing
system performing the socket routing.
In any event, the socket layer which performs the socket routing contains
facilities to automatically route a connection across different domains.
In the client processing system which is attempting to create a connection,
a socket is created in a particular domain. If the socket is in a
different domain, the socket does not fail if the socket routing facility
of this invention is implemented. The connect function is modified to
catch the attempts at a cross domain connection. If a connect function is
attempted on a socket in a different domain, then the socket routing
facility of this invention is invoked.
Alternatively, a connectto function can be implemented which takes the
place of and combines the functions of the socket function and the connect
function. With the connectto function, a socket is not created until the
route is known. This alleviates the unnecessary work of creating a socket
which may fail, and then performing actions as a result of the failed
socket. The connectto function determines how a connection can be made,
and then creates a socket in the domain that is needed to establish the
determined connection.
Through either of the above approaches, a connection to a socket in a
different domain can be made through an intermediate socket. When data
arrives from one end of the connection to the intermediate socket, the
intermediate socket immediately sends the data to the other end of the
connection instead of queuing the data for process intervention at the
intermediate processing system.
In addition, if the intermediate socket is queried for the address of the
other end of the connection, the intermediate socket identifies the
connecting host as opposed to the intermediate host. In this way, the
socket routing facility of the intermediate host is transparent to the
hosts at each end of the connection.
BRIEF DESCRIPTION OF THE DRAWING
FIG. 1 is a diagram showing a connection from a process AA on host A to a
process CC on host C. Socket routing is utilized to cross the boundary
between the networks of type A and type C at host B.
FIG. 2 is a flow diagram showing the operational scenario of FIG. 1 using
explicit and implicit routing.
FIG. 3 is a flow diagram showing the modified steps in performing a connect
( ) function to a destination.
FIG. 4 is a flow diagram showing the steps of creating a socket if the host
does not have a socket in the specified domain.
FIG. 5 is a flow diagram showing the steps performed at host B.
FIG. 6 is a flow diagram showing the steps of a connectto ( ) function.
FIG. 7 is a more detailed diagram of the socket routing facility of this
invention.
DESCRIPTION OF THE PREFERRED EMBODIMENT
The following description describes an architecture for routing virtual
circuits based on sockets. Although this implies stream sockets, the
invention is not limited to stream protocols or to sockets. The concepts
of this invention could be applied to similar communication end points
that are utilized within other operating systems.
Referring to FIG. 1, a process AA, 10, in a data processing system 11, host
A, desires to connect its socket facilities 12 to the process CC, 40, in a
data processing system 41, host C. The data processing system 11 is shown
as only supporting a particular domain of sockets AF.sub.-- A, 13, such as
TCP, and data processing system 41 is shown as only supporting sockets
that exist in the domain having address family C, 43. Since the naming
conventions and the underlying transport mechanisms are different between
address family A, 13, and address family C, 43, no interconnection can
take place without an intermediate facility. The intermediate facility is
the socket routing facility 70 in socket layer 32, which exists in data
processing system 31, shown as host B.
To describe the initiation of a connection, the process AA, 10, in the data
processing system 11, will activate a connection through the sockets
programming interface to the general socket code, 12, which in turn goes
through the address family specific socket code for AF.sub.-- A, 13. The
necessary data and control information will be handled by the interface
and physical access layers, 14. The data will then go out on the network
50 and end up going into data processing system 31, shown as host B, via
the interface layer 34, and then through the code for address family 30 A,
shown as AF.sub.-- A, 36.
For comparison, data processing system 21, shown as host D, shows existing
internet routing within a single address family, the address family A,
AF.sub.-- A, 23. It should be noted that the cross connection occurs
within the address family A, 23. Almost any TCP/IP implementation can
route within its own address family. Likewise, SNA has similar gateway and
forwarding capabilities. The cross over as shown in data processing system
21 is independent of the model type of either stream or datagram. It is
only dependent upon being within the same network domain.
In data processing system 31, the connection request packets will go
through the interface layer code 34 to the address family A code,
AF.sub.-- A, 36, through the general socket layer 32, and into the socket
routing code 70. The socket routing code facility 70, is where the address
mapping and cross connection takes place. The cross connection arrows 37
are shown drawn in the socket routing layer 70 of data processing system
31, as opposed to the cross connection arrows 27 which are shown in the
address family code 23 of data processing system 21.
A connection request generated in the socket routing code 70 of data
processing system 31 will then go down through the address family C code,
AF.sub.-- C, 33, and through the interface layer code 35 for the other
network 60, such as SNA. The connection request packets go across the
network 60 to the interface layer code 44, up to the address family C
code, AF.sub.-- C, 43, continuing through the general socket interface
layer code 42 where the connection is registered. Then the process CC, 40,
can respond to the connection request in order to establish the connection
between cross domain networks.
FIG. 7 shows item 70 of FIG. 1 in greater detail. Item 701 is the programs
and data for controlling the socket routing facility. A connection request
to establish socket routing will come in on the sockets for this service,
items 704, and 705. The routing agent software, item 703, will accept the
connection, which creates a data socket, items 709-714. The route request
message will come in on that data socket, and the routing agent, 703, will
consult its route database, 702, to see if a route is possible. If a route
is possible, the routing agent, 703, will consult its route database, 702,
on how to establish the route. Then, the routing agent creates a matching
data socket (item 710 for item 709, etc.), and connects to the next hop.
When the routing agent software receives any replies for further route
hops, it forwards them back to the socket routing requestor via the
accepted data socket. When all hops are made, the socket routing agent
will create a data transfer agent, items 706-708, that joins the pairs of
data sockets, and forwards data from one to the other and vice versa.
The above scenario is further described in the following programming design
language code. The following includes | | |
