TECHNICAL FIELD

The invention relates to the fields of networking and route calculation, and specifically to the problem of allowing support of dependent logical units (DLU) in a SNA (System Network Architecture) Advanced-Peer-to-Peer Architecture (APPN) environment to be located downstream of a branch extender and still have the network function properly as to locating of the DLUs, route calculation to and communication with the DLUs.

BACKGROUND OF THE INVENTION

IBM's Advanced Peer-to-Peer Architecture (APPN) provides networking between nodes of a network on a peer-to-peer basis. That is, all nodes are able to communicate amongst themselves on an equal basis, without regard to any hierarchical network structure. An APPN network contains network nodes (NNs) and end nodes (Ens). A network node contains all of the functionality to initiate control point sessions with other network nodes, to locate resources within the network and to calculate routes to other nodes. Control point (CP) sessions are sessions over which control messages are passed between so-called network control points. Control points (CP) characterize the functionality in network nodes that provide network services, such as route calculation. Among other functions, network nodes serve end nodes for locating resources and calculating routes, among other things. End nodes are nodes of lesser capability and typically include such things as user workstations, printers, and the like.

Users of IBM's SNA architecture and APPN architecture have a large investment in a type of end node called a dependent logical unit (DLU). A DLU is an end node device that, unlike an independent logical unit, does not have the ability to initiate sessions and to manage log on requests to server nodes. Instead, DLUs rely on other nodes to provide these services on its behalf via the CP capability. Because of their limited capability, DLUs are cheaper than end nodes that have full capability to initiate and manage sessions to other nodes. DLUs are supported in APPN by a mechanism called a dependent LU requester (DLUR) and dependent LU server (DLUS). Essentially, a session pipe is established between a CP DLUS node (a network node) and a DLUR node that owns (serves) the DLUs. Once a communication pipe called a CP-SVR pipe has been activated between the DLUR and DLUS, DLU message flows are encapsulated over sessions contained within the CP-SVR pipe between the DLUR and the DLUS. Numerous IBM publications describe APPn and DLUS/DLUR operation in detail. Many of these publications are available at IBM's Redbook web site http://www.redbooks.ibm.com. At this site, see for example the Redbook SG24-4656-01, May 29, 1998, Subarea to APPN Migration: VTAM and APPN Implementation.

The APPN architecture also allows users to configure an APPN network as an upstream component connected by a wide area network (WAN) to downstream components. The downstream components are typically local area networks (LANs). The mechanism that allows this configuration is called a branch extender (BEX) node. The nodes downstream of the BEX are usually end nodes and are considered to be within the domain, or branch, of the BEX. To these domain nodes, the BEX node looks exactly like an APPN network node that serves the branch nodes. To upstream nodes, the BEX node looks exactly like an APPN end node. Thus, a BEX node has an upstream node that acts to it as a serving network node.

More detailed information regarding APPN branch extenders is also available at web site http://www.redbooks.ibm.com. See, for example, the Redpiece article BRAN-CHXX, Jun. 1, 1997, APPN Branch Extender (BX), and the Redbook article SG24-5232-00, Oct. 10, 1998, IBM eNetwork Communications Server for Windows NT Version 6.0 Enhancements, both available at this web site.

While a BEX node knows the topology of its branch, it knows nothing of the upstream network topology. When a BEX node can't resolve a request for network services locally, it refers the request to its network node. This can be thought of as default routing. As far as upstream nodes are concerned, everything in the branch of the BEX node resides in the BEX node. The BEX node, when necessary, provides pass-through to its branch nodes for locate requests, session binds and unbinds, and the like. A BEX node registers all branch resources with its serving network node, including those that reside on the BEX node itself as well as those that reside on served end nodes. When registering the branch resources, the BEX node indicates that it is the EN control point (CP) for all resources it registers, even when those resources actually reside on a different end node. When sending or responding to directory searches, the BEX node modifies the directory search so that it includes the BEX nodes trunk vectors prior to sending the directory search into the WAN.

A BEX node provides information to the upstream network consistent with its end node appearance in that network. For directory services purposes, a BEX appears to own all of its branch resources. For route selection purposes, the BEX appears to be the origin or destination of all sessions involving logical units (LUs) in the branch of the BEX. For session routing, the BEX modifies bind messages destined for an end node of the BEX branch before the branch node ever sees it, by inserting the missing hop between the BEX and the branch node.

Branch extenders are very useful, in ways not important for this discussion, to provide network users with a flexible and inexpensive way to connect LANs and WANs in an APPN environment. DLU support is also very important to network users. It is also important to users to combine the advantages of BEXs and DLUS/DLUR in a single network. However, this combination introduces a major obstacle to network users. In the prior art, the DLUR support for DLUs must reside in a BEX node. Because a BEX appears as an end node to the upstream network, it is required to provide the DLUR capability for all of its branch nodes. The result is that only the BEX that is closest to the upstream WAN may support DLUR. On the other hand, the advantages of DLUS/DLUR are optimized when the DLUR function is located as close to a supported DLU as possible. In many cases, this is downstream of a BEX. Customers that presently have the DLUR located near the DLUs and who wish to install an upstream BEX to obtain other advantages are required to change their network configuration to relocate the DLUR support to the BEX.

A number of other problems also exist in locating DLUR support downstream of a BEX. A DLUS ordinarily calculates routes to a DLUR during operations relating to DLU support. A DLUR ordinarily registers with a DLUS its trunk group (TG) vector information necessary to calculate the routes to its DLUs. TG vector information contains the link addresses for all DLUs connected to a DLUR. However, a BEX provides isolation between an upstream DLUS node and the branch nodes of the BEX. Because of this, the DLUS node cannot calculate routes to the branch end nodes of the BEX. This means that a DLUS cannot calculate routes to a DLUR that resides downstream of a BEX. Another problem occurs because all branch nodes of a BEX are considered by a BEX to reside on the BEX. However, in a DLUS/DLUR environment, a DLUR registers with the DLUS as the owner of DLUs. In response to a resource Locate search request in a network containing both a BEX and DLUS/DLUR support, then both the BEX and the DLUR may appear in different ways to the Locate request as the owner of the resource that is being located. This, of course, can cause major malfunction of the network.

SUMMARY OF THE INVENTION

The invention resides in a method and apparatus that allows the DLUR function to reside in an end node served by a branch extender node in an APPN network. Two problems are solved by the invention that allows the DLUR support to be located downstream of a BEX. First, in an APPN network, when the DLUS and the DLUR reside in the same APPN network (a network in which all network nodes share topology information) and are separated by a BEX, the BEX provides isolation between the DLUS and the DLUR such that the DLUS cannot calculate routes to the DLUR. Second, a way must be found to avoid the problem of having both the DLUR and the BEX report ownership of a DLU.

The first problem is overcome by the invention by forcing the DLUS to view the DLUR as residing in a different network, even though this is not the reality. This forces the DLUS to initiate a resource Locate search request to determine routes to the DLUR. The DLUS is made to view the DLUR as being in a different network by having the BEX set a fictional network boundary crossed indication in Locate request messages passed between the DLUS and DLUR when control communication paths are first established between the DLUS and the DLUR. Then for all subsequent session setup requests to a DLU served by the DLUR, the DLUS treats the DLUR as being in a different network for which it has no network topology information, and causes it to initiate a Locate request to find a path to the DLUR.

The second problem is solved by having a BEX examine every resource Locate request or resource Locate reply to determine if the resource is a DLU. If it is, then the BEX is instructed not to substitute itself as the owner of the resource in Locate requests and responses. Two conventional fields in resource Locate and resource reply messages allow this determination. If both an Owning Control Point Respond (OCR) indicator and a DLUS served LU (DLS) indicator are set in the message, then the resource is a DLU served by a DLUR and the BEX does not substitute itself as the resource owner in this instance.

BRIEF DESCRIPTION OF THE DRAWING

FIG. 1

illustrates a prior art APPN network containing DLUS/DLUR nodes for supporting dependent logical units (DLUs);

FIG. 2

illustrates a prior art APPN network containing branch extender (BEX) nodes;

FIG. 3

illustrates an APPN network in which BEX nodes are combined with DLUR support located downstream of a BEX. This Fig. is used to describe both the problem and the solution of the invention;

FIGS. 4 and 5

show illustrative flowcharts of the operations performed by the invention in establishing a CP-SVR pipe between DLUS and DLUR nodes in the network of

FIG. 3

; and

FIGS. 6 through 8

show illustrative flowcharts of the operations performed by the invention in establishing communication sessions with DLUs supported by the DLUR in the network of FIG.

3

.

DETAILED DESCRIPTION

Dependent logical units (DLUs) are supported in an APPN network by a dependent logical unit server (DLUS) network node and a dependent logical unit requester (DLUR) node to which the DLUs are connected. The DLUR may be a network node or an end node. A special communications path, referred to as a control point-server (CP-SVR) pipe, is established between the DLUS and the DLUR. Control messages are transmitted over this pipe to locate a path to a DLU in response to a request from a DLU to establish a session with an application at another node or a request by an application at another node to establish a session with the DLU. The node containing the application is eventually provided information to allow it to calculate a route to the DLUR serving the DLU. The CP-SVR pipe consists of two LU 6.2 sessions between the DLUS and DLUR nodes. These sessions are similar to the control point to control point (CP-CP) sessions between adjacent APPN control points. The CP-SVR pipe sessions use a special class of service name, CPSVRMGR to identify themselves as DLUS/DLUR sessions.

FIGS. 1 through 3

explain the support of DLUs and BEXs in an APPN network and illustrate the problem addressed by the invention.

FIG. 1

illustrates a prior art APPN network, including basic DLUS/DLUR support for DLUs.

FIG. 1

shows a wide area network (WAN)

100

. Upstream of WAN

100

is a network node (NN)

102

, which, by way of example, is indicated to contain the DLUS function

104

. Downstream of WAN

100

are a number of NNs, such as NN

106

,

108

and

112

, and ENs

110

and

114

. EN

110

is served by NN

106

. EN

114

is served by NN

112

. EN

114

contains DLUR support

116

. DLUR

116

, along with DLUS

104

, supports one or more DLUs connected to EN

114

. These DLUs are not shown to simplify the drawing.

Before the DLUs can be supported, the CP-SVR pipe between NN

102

and EN

114

must be activated. To do this, the DLUS

104

sends a resource Locate search request to find the DLUR

116

. The operation of the Locate function in an APPN network is explained in detail in U.S. Pat. No. 4,914,571. The Locate search request includes a special class of service name CPSVRMGR, indicating the search request is being sent as part of CP-SVR pipe activation. The Locate search request is received by NN

112

, which is the NN server for the DLUR EN

114

. NN

112

forwards the Locate search request to DLUR EN

114

. EN

114

responds to NN

112

with a positive response to the Locate search request. NN

112

then sends a positive Locate reply to the DLUS NN

102

. Upon receiving the Locate reply, DLUS NN

102

sends a BIND for a CPSVRMGR session to the DLUR EN

114

. The DLUR EN

114

responds with a positive BIND response. At this point, one half of the CP-SVR pipe is active. The DLUS NN

102

can now send a conventional request over this half of the pipe to EN

114

to activate a physical unit (PU) which manages the DLUs.

After the PU on EN

114

is activated, EN

114

needs to return a positive activation response to NN

102

. However, to do this, EN

114

must first activate the second half of the CP-SVR pipe. To do this, EN

114

sends a Locate search request to NN

112

to find its DLUS. The Locate request includes the CPSVRMGR class of service name, indicating that the search request is sent as part of CP-SVR pipe activation. NN

112

forwards the request to the DLUS NN

102

in the conventional way as described in U.S. Pat. No. 4,914,571. NN

102

responds to the search request with a positive reply, which is returned to NN

112

. Upon receipt of the reply, NN

112

calculates a session route from the DLUR EN

114

to the DLUS NN

102

. NN

112

returns a positive Locate reply to DLUR EN

114

and includes the session route just calculated in the reply. The DLUR EN

114

uses this session route to send a BIND request for a CPSVRMGR session to DLUS NN

102

. The CP-SVR pipe is now fully active and the PU at EN

114

that manages the DLUs is also fully active.

Once the CP-SVR pipe is active, the DLUS can activate dependent LUs (DLUs) which are managed by the DLUR. To activate a DLU, the DLUS sends an activate LU message to the DLUR over the CP-SVR pipe. The DLUR responds to an activate message with a positive response, at which time the DLU is active. Upon receipt of the response, the DLUS registers the fact that the DLU is active and that the owning node for the DLU is the DLUR. This information is used during later LU-LU session activation between other applications in the network to the DLU, as discussed later in this disclosure.

After sending a positive response to an activate DLU message, the DLUR must also register the DLU with its NN server. The NN server for DLUR EN

114

is NN

112

in FIG.

1

. By registering the DLU with its NN server, resources connected to the WAN

100

are able to locate the DLU and establish LU-LU sessions with it. DLUR

116

uses an APPN Register general data stream (GDS) variable to register the DLU with its NN

112

. The Register GDS variable also identifies the DLUR

116

as the owning node for the DLU.

In the case of

FIG. 1

, it is important to note that both the activate LU response and the register GDS variable cause the same information to be registered in WAN

100

; that is, a registration is made in DLUS

104

that the DLU resides on the DLUR EN

114

. APPN requires that a given resource reside only on one node; APPN cannot tolerate a LU being registered as owned by two or more nodes. This is a source of one problem solved by the invention when a DLUR is located downstream of a branch extender node, as discussed in more detail below.

To calculate a route to DLUR EN

114

for communications with a DLU, it is sometimes necessary for the DLUR EN

114

to register its trunk group (TG) vectors with the DLUS NN

102

. This is true, for example, when a DLUR and DLUS reside in the same APPN network and the DLUR is an EN, as in the case of FIG.

1

. For purposes of this discussion, an APPN network is defined as a set of APPN NNs which share a common network topology database, and the ENs which are served by these NNs. The TG vectors are used for LU-LU session activation, which is discussed below. The important point is this registration at the DLUS of TG vectors for a DLUR only occurs when the DLUR and DLUS are in the same network. Both the DLUR and DLUS learn whether they are in the same network during the activation of the CP-SVR pipe. A resource Locate search, on both the request and reply, contains a special field, which identifies the number of APPN network boundaries that have been crossed as the search or reply message propagates through a network and between networks. An APPN network boundary is defined as a TG between a pair of APPN networks that do not share a common network topology database. As a resource Locate search traverses this boundary, the special field in the locate search message is incremented. At the destination of the requests or replies, if the field is zero, no network boundary has been crossed and the source and destination nodes are considered to be in the same network. If the field is non-zero, at least one network boundary has been crossed and the nodes are considered to be in different networks.

The manner in which a LU-LU session is established from an application at some node in the network to a DLU is now discussed for the prior art arrangement of FIG.

1

. Once a DLU is active it can connect to other applications in the network. This is accomplished by having the DLUR send a logon request to the DLUS over the CP-SVR pipe. This request indicates that a DLU wishes to establish a session (i.e., logon to) an identified application. In

FIG. 1

, assume that the other application resides on NN

108

. The DLUS sends a resource Locate search request into the network, which eventually finds it way to NN

108

, to find the application. Included in the resource Locate search request is information about the DLU that is attempting to logon to the application. The information registered about the DLU by NN

112

when the activate LU response was received over the CP-SVR pipe is used when sending the Locate search request, and includes the DLUR

116

as the owning node. The Locate search also includes the DLUR TG vectors registered with the DLUS by the DLUR

116

over the CP-SVR pipe.

When NN

108

receives the Locate request, it calculates a session route to the owning node, which in this case is the DLUR

116

. NN

108

uses both the network topology database and the DLUR TG vectors included in the Locate search to calculate the session route. The network topology database includes all connections between NNs in the APPN network, but does not have any information about TGs between two ENs or between a NN and an EN. The DLUR TG vectors contained in the Locate search contains all TGs from the DLUR

116

to other ENs and from the DLUR to other NNs. By using both the DLUR TGs and the network topology database, NN

108

is able to calculate a route from itself to the DLUR

116

. NN

108

send a BIND to the DLUR

116

using the calculated session route. The DLUR responds with a positive BIND response, and the LU-LU session between the application and the DLU is established.

FIG. 2

shows a prior art APPN network that is augmented with branch extender (BEX) nodes. BEX nodes allow WANs and LANs to be included in the topology database of an APPN network, while at the same time providing isolation between the WAN and the LANs. Isolation is accomplished because the BEX acts as the owner (the residence) of all resources contained within its domain (its branch of the network). This allows great flexibility of network configuration for users of APPN networks. APPN supports a form of data transmission called high performance routing, or HPR. Most HPR network control messages are restricted to network nodes, rather than end nodes, such as front end processors, routers, or communications servers that direct traffic in the network. Prior to BEX nodes, any node that routed HPR data between other nodes was required to be a network node, and to send and receive such network control messages. Since a BEX node is a hybrid network node and end node, the BEX node is able to route HPR data between other nodes like a network node. Since the WAN views the BEX node as an end node the BEX node does not exchange topology information with the WAN and only processes directory searches for resources in its branch like an end node. This reduces the processing overhead for a BEX node while allowing data to be routed through the BEX node.

FIG. 2

shows a NN

202

upstream of a WAN

200

. Various nodes are located downstream of the WAN, such as NN

208

and other nodes. In particular, a BEX node

206

is connected to the WAN

200

and is shown as also containing DLUR

210

. A cascaded BEX is shown at

214

, which serves EN

216

. A cascaded BEX is one whose NN server in the WAN is also a BEX. The cascaded BEX poses as an EN to its NN BEX. The NN BEX poses as a NN to the cascaded BEX. The only difference between a non-cascaded BEX and a cascaded BEX is that a non-cascaded BEX supports a DLUR, whereas a cascaded BEX does not. In the example arrangement of

FIG. 2

, DLUR

210

supports any DLUs in the branch of BEX

206

, which includes DLUs connected to EN

212

and EN

216

.

A BEX node filters out network control messages that are unnecessary to the network branch below the BEX node. Technically, to nodes within its branch, a BEX node poses as a network node (NN) and it may contain DLUR support. At the same time it poses as an end node (EN) to computers upstream of the WAN

100

.

The following describes how resources in a branch of a BEX are registered with the BEX. After activating CP-CP sessions, an EN, whose NN server is a BEX, performs normal APPN functions, such as registering resources with the BEX that are owned by the EN. In

FIG. 2

, EN

216

registers all local resources with its NN server, BEX

214

. These resources, sent to the BEX in a Register GDS variable, consist of EN

216

(the CP name of EN

216

) and any local LUs contained in EN

216

. If it is assumed that EN

216

contains one LU identified as LUX (not shown), the Register GDS variable contains two registration resources, EN

216

and LUX. The owning CP in the Register GDS variable is EN

216

. The Register GDS variable looks like:

Owning node=EN

216

, Resources=EN

216

, LUX

BEX

214

then activates a CP-CP session with its NN server BEX

206

. BEX

214

registers its CP name with BEX

206

, as well as all its local resources with BEX

206

. In

FIG. 2

, BEX

206

has no local resources, so only its CP name, BEX

206

, is registered. BEX

206

also registers all resources located on its branch ENs. In

FIG. 2

, EN

216

has registered two resources, EN

216

and LUX, with BEX

214

. BEX

214

includes these resources when it registers with BEX

206

. However, BEX

206

alters the resource hierarchy so that it appears as the owning node for all of these resources. This requires that the node containing the DLUR function be located adjacent to the WAN

200

. This is another important point in that it becomes another problem solved by the invention when a network is configured having DLUR support located near the actual DLUs being supported. In the present case, the register GDS variable looks like:

Owning node=BEX

214

, Resources=BEX

214

, EN

216

, LUX

When BEX

206

activates its CP-CP session with its NN server

202

, it also performs resource registration with NN

202

. BEX

206

also registers all resources that have been registered by its branch ENs as well. The Register GDS variable looks like:

Owning node=BEX

206

, Resources=BEX

206

, BEX

214

, EN

216

, LUX

Once this Register GDS variable is sent, all nodes in the WAN are able to locate resources in the branch network downstream of BEX

206

.

One key point in Branch Extender resource registration is that the owning node is always the node which is sending the Register GDS variable. The APPN architecture requires an EN to register only resources that reside on that EN. As a result, APPN does not allow an EN to register a resource which has an owning node other than the EN which sends the Register GDS variable.

In

FIG. 2

, assume that LUX wishes to establish a session to an application that resides on NN

208

. To do this, LUX must first obtain a session route. This is accomplished by sending a Locate search into the APPN network. On behalf of LUX, EN

216

sends a Locate search to its NN server, BEX

214

, requesting a session route from EN

216

to the node on which the application resides. Because EN TG vectors are not part of the APPN network, EN

216

must include all TG vectors for which it is an endpoint in the Locate search so that they may be used in calculating the session route.

When BEX

214

receives the search it does not know the location of the application. Since BEX

214

is a Branch Extender, it sends a Locate search to its NN server BEX

206

requesting a session route from BEX

214

to the node that owns the application. When sending the Locate search, BEX

214

alters the origin LU resource hierarchy for LUX so that BEX

214

appears as the owning CP for LUX. This causes one of the problems solved by the invention, namely of creating the possibility of two nodes reporting ownership of a DLU when the DLUR is located downstream of a BEX. BEX

214

also removes all TG vectors included by EN

216

from the Locate search and replaces them with TG vectors for which BEX

214

is an endpoint. These two changes are necessary so that BEX

206

can calculate a session route from BEX

214

to the node on which the application resides.

When BEX

206

receives the search it also does not know the location of the application. Like BEX

214

, BEX

206

is a Branch Extender and must modify the Locate search before sending it into the WAN. BEX

206

changes the origin LU resource hierarchy for LUX so that BEX

206

appears as the owning CP for LUX. BEX

206

also removes all TG vectors included by BEX

214

from the Locate search and replaces them with TG vectors for which BEX

206

is an endpoint.

NN

202

eventually receives the Locate request and performs existing APPN search logic and successfully finds the application on NN

208

. NN

202

calculates a session route from BEX

206

to NN

208

using the network topology database and the TG vectors provided by BEX

206

on the Locate search request. NN

202

returns the session route in the Locate reply sent to BEX

206

. The session route returned to BEX

206

looks like:

BEX

206

—NN

202

—NN

208

Upon receipt of the positive Locate reply, BEX

206

calculates a session route from BEX

214

to BEX

206

using the TG vectors provided by BEX

214

on the Locate search request. BEX

214

then prepends this session route to the one it received and returns the modified session route to BEX

214

in the Locate reply. The session route now looks like:

BEX

214

—BEX

206

—NN

202

—NN

208

BEX

214

performs logic similar to BEX

206

. BEX

214

calculates a session route from EN

216

to BEX

214

using the TG vectors provided by EN

216

on the Locate search request. BEX

214

then prepends this session route to the one received from BEX

206

on the Locate reply and sends this session route to EN

216

on the Locate search reply. The session route now looks like:

EN

216

—BEX

214

—BEX

206

—NN

202

—NN

208

EN

216

uses this session route to send a BIND request to the application that resides on NN

208

.

The above discussion focused on prior art DLU support and the use of branch extenders in the APPN architecture. In the prior art described above, only the BEX node “closest” to the WAN (BEX

206

) can provide DLUR support. That is, only a BEX whose NN server is not another BEX can be a DLUR. A BEX enforces this restriction by not allowing the CP-SVR pipe to be established between a DLUR in the branch and a DLUS in the WAN. This is accomplished by sending a negative reply to any Locate search request for the CPSVRMGR mode, which is used exclusively for CP-SVR pipes.

When an attempt is made to provide DLUR support in a BEX that is not the closest BEX to the WAN, the problems mentioned above arise.

FIG. 3

shows an attempt to combine FIG.

1

and

FIG. 2

, and illustrates these problems addressed by this invention.

FIG. 3

shows a NN

302

upstream of the WAN

300

and containing the DLUS function

304

. Downstream of the WAN

300

is a NN

308

and a BEX

306

. Connected to BEX

306

are EN

310

and a cascaded BEX

312

. BEX

312

is connected to EN

314

, which contains the DLUR function

316

. Contrary to the example of

FIG. 2

, the DLUR

316

is located downstream from the Branch Extender

306

that is closest to the WAN

300

.

The first problem that arises is that a BEX explicitly disables activation of a CP-SVR pipe between the downstream DLUR

316

and the DLUS

304

. To understand why, assume that BEX

306

allows the CP-SVR pipe to be activated between DLUR

316

and DLUS

304

. Since the DLUR is in EN

314

and is in the same APPN network as the DLUS

304

, the DLUR

316

registers its TG vectors with the DLUS once the CP-SVR pipe is activated. Thereafter, during session activations, nodes upstream of the WAN

300

will attempt to use the network topology database and these TG vectors to calculate routes to the DLUR

316

. However, since the DLUR

316

is not attached directly to WAN

300

, none of the TG vectors point to any node which can be reached directly from the WAN

300

. Therefore, no LU-LU session can be established between a DLU supported by the DLUR

316

to an application upstream of WAN

300

.

A second problem with downstream DLUR support for BEX nodes concerns resource registration in the BEX NN server and in the DLUS function. As mentioned previously, an activate LU response causes the DLU to be registered at the DLUS

304

with the DLUR

316

as the owning node. The DLUR

316

also registers the DLU with its NN server BEX

312

and BEX

312

with its NN server BEX

306

. As the Register GDS variable is processed by the BEX

306

, BEX

306

modifies the resource hierarchy so that it appears as the owner for the DLU resource being registered. Thus, the DLU appears to reside on the BEX closest to the WAN, which in this case is BEX

306

. However, the DLU has also been registered with the WAN during activate LU processing, as residing on DLUR

316

. Thus, we have a resource defined in different ways as residing at different nodes of the network. This causes APPN nodes upstream of WAN

300

to be unable to locate the DLU using APPN search mechanisms.

One part of the solution to allow a successful Locate search for DLUR

316

is to force DLUS

304

to view DLUR

316

as residing in a different APPN network. When the DLUR and DLUS are in different APPN networks, the DLUR does not register its TG vectors with the DLUS. Instead, the DLUS sends a Locate search to obtain the necessary routing information to allow a session initiation request to be sent to the DLUR. This avoids one part of the problem.

To cause the DLUR and DLUS to view each other as residing in different APPN networks, in accordance with one aspect of the invention, a BEX modifies Locate search requests and replies for the CP-SVR pipe sent between the DLUR and DLUS. Locate searches contain a special field, known as the ISTG (inter-subnet trunk group) crossed-count, which specifies the number of APPN network boundaries which have been crossed as a Locate or Locate reply propagates. If the ISTG crossed-count is zero when a Locate request or reply arrives at its destination, then no network boundaries have been crossed and both nodes are in the same network. If the ISTG crossed count is non-zero, then at least one boundary has been crossed and the two nodes are in different APPN networks. When processing a Locate search that is part of CP-SVR pipe activation (the class of service name in the Locate search is CPSVRMGR), the invention causes a BEX to increment the ISTG crossed count. For all other Locate search requests the ISTG crossed count remains unaltered. As a result, both the DLUR and DLUS view each other as being in different APPN networks, but only for the purpose of CP-SVR pipe activation. For all other functions, both the DLUR and DLUS appear to be in the same APPN network. This, in turn, forces the DLUS to initiate Locate search requests for a DLUR downstream of a BEX, rather than erroneously using registered DLUR TG vectors.

FIG. 3

, viewed in conjunction with the flowcharts in

FIGS. 4 through 8

, illustrate the invention in more detail. Assume that DLUS

304

wants to activate a physical unit (PU) on DLUR

316

so that it might communicate with a DLU served by DLUR

316

. DLUS

304

must first activate the CP-SVR pipe between itself and the DLUR

316

. In step

402

of

FIG. 4

, DLUS

304

first determines if a CP-SVR pipe already is active to DLUR

316

. Assuming that a CP-SVR pipe is not active, at step

404

DLUS

304

sends a Locate search request to find DLUR

316

. The Locate request includes the CPSVRMGR class of service name, indicating that the search is being sent as part of CP-SVR pipe activation. The Locate request is received by BEX

306

at step

406

. At step

408

, BEX

306

determines from the CPSVRMGR class of

10

service that the request is for activating a CP-SVR pipe. As a result, at step

410

, BEX

306

increments the ISTG crossed-count from 0 to 1 and forwards the Locate search to BEX

312

at step

412

. BEX

312

also determines at step

414

that the search concerns activation of a CP-SVR pipe. As a result, step

416

increments the ISTG crossed count from 1 to 2. Step

418

sends the Locate search to EN

314

and DLUR

316

. At step

420

, DLUR

316

responds with a positive Locate search reply that is sent to BEX

312

. For the same reason already discussed, at step

422

, BEX

312

increments the ISTG crossed count in the Locate reply from 0 to 1 and forwards the reply to BEX

306

. At step

424

, BEX

306

also increments the ISTG crossed-connect from 1 to 2 and sends the Locate reply to NN

302

and DLUS

304

. Upon receiving the Locate reply, at step

426

, DLUS

304

sends a session BIND for a CPSVRMGR session to DLUR

316

. DLUR

316

replies with a positive BIND response. At this point, one half of the CP-SVR pipe is active.

DLUS

304

is now able to send an activate PU command (ACTPU) over the half CP-SVR CPSVRMGR session to DLUR

316

, which it does at step

502

of FIG.

5

. DLUR

316

receives the ACTPU at step

504

and activates the PU. To respond to DLUS

304

, DLUR

316

must first activate the second half of the CP-SVR pipe. As part of step

504

, DLUR

316

sends a Locate search to BEX

312

to find DLUS

304

. The Locate search includes the CPSVRMGR class of service name, indicating that the search request is being sent as part of CP-SVR pipe activation. BEX

312

increments the ISTG crossed-connect from 0 to 1 for reasons previously discussed and sends the Locate search to BEX

306

. BEX

306

also increments the ISTG crossed-connect from 1 to 2 and forwards the request to NN

302

and DLUS

304

. DLUS

304

responds with a positive Locate search reply. Upon receipt of the reply, BEX

306

increments the ISTG crossed-connect from 0 to 1 and forwards the Locate search reply to BEX

312

. BEX

312

increments the ISTG crossed-connect from 1 to 2 and forwards the Locate reply to DLUR

316

. DLUR

316

sends a BIND request for a CPSVRMGR session to DLUS

316

. DLUS

316

responds with a positive BIND response, completing the activation of the second half of the CP-SVR pipe. The CP-SVR pipe is now fully active. Because the ISTG crossed-count on the earlier Locate for DLUR

316

was non-zero, indicating that DLUS

304

and DLUR

316

are in different networks, DLUR

316

does register its TG vectors following the activation of the CP-SVR pipe. DLUR

316

, at step

506

now builds a positive response to the earlier ACTPU command. At step

508

, DLUR

316

sends the ACTPU response to DLUS

304

and the DLUR

316

PU supporting DLUs served by DLUR

316

is fully active.

Now that the CP-SVR pipe between DLUS

304

and DLUR

316

is active, and the PU is active at DLUS

316

, dependent LUs (DLUs) served by DLUR

316

can be activated. To activate a DLU, at step

510

, DLUS

304

sends an ACTLU command to DLUR

316

over the CP-SVR pipe. The DLUR responds to the ACTLU with a positive ACTLU response, at which time the DLU is active. Upon receipt of the ACTLU response, DLUS

304

registers the fact that the DLU is active and that the owning node for the DLU is DLUR

316

. This information is used during LU-LU session activation to the DLU, which is discussed later in this disclosure.

After sending the positive ACTLU response, DLUR

316

must register the DLU with its NN BEX

312

. By registering, resources connected to WAN

300

are able to locate the DLU and establish LU-LU sessions with it. At step

512

, DLUR

316

uses the APPN Register GDS variable to register the DLU with BEX

312

. The Register contains an indication that the LU is a DLU and it also contains the name of the DLU. The Register identifies DLUR

316

as the owning node for the DLU. At step

514

, BEX

312

determines if the resource is a DLU. Since in this example, it is, step

516

that would otherwise mark BEX

312

as the DLU owner is omitted. Step

518

forwards the Register to BEX

306

with DLUR

316

indicated as the owning node. At step

520

, BEX

306

also determines if the resource is a DLU and since it is step

522

, which would otherwise mark BEX

306

as the owner, is also omitted. Step

524

registers the DLU at DLUS

304

with DLUR

316

appearing as the owner.

The flowcharts of

FIGS. 6 through 8

illustrate operation of the invention when a DLU served by DLUR

316

wishes to establish a session with another application in the network, after being activated and registered as described above. Step

602

describes the situation that the DLU desires a session with an application that is assumed to reside at NN

308

. To setup the session, at step

604

the DLU sends a log on request to DLUS

304

over the CP-SVR pipe, which is received by DLUS at step

606

. Recall that when the CP-SVR pipe was established, the ISTG crossed-count was non-zero. Thus, step

608

causes DLUS

304

to treat DLUR

316

as being in a different APPN network. As a result, step

610

sends a Locate search request into the network to obtain routing information for DLUR

316

that can be used to calculate a session route from WAN

300

to DLUR

316

. Both an Owning Control Point Respond (OCR) field and a DLUS Served LU (DLS) field are set in the Locate search request. The OCR and DLS fields are defined fields in the APPN architecture; when both are set, downstream BEX nodes receiving the Locate search command do not mark themselves as owner of the DLU as is normally done by BEXs. The Locate request is received by BEX

306

at step

612

. Step

614

determines that the Locate request is not from a downstream node. In this case, there is no need for the BEXs to concern themselves with changing ownership of the resource searched for, so steps

616

,

618

and

622

are omitted. Step

624

forwards the Locate request to BEX

312

; BEX

312

merely forwards the request to DLUR

316

at step

632

. The steps

628

,

630

,and

631

are not executed in this instance because step

626

determines that the Locate request is from an unstream node.

At step

702

of

FIG. 7

, in response to receiving the Locate request, DLUR

316

returns a positive Locate reply to BEX

312

. The OCR and DLS fields are set in this reply to prevent the upstream BEX nodes from changing the ownership hierarchy as described above. The reply also includes the TG vectors for DLUR

316

so that DLUS

304

can eventually calculate a route to DLUR

316

. At step

704

, BEX

312

determines if the Locate reply request is from a downstream node. Since it is in this case, step

706

determines that both the OCR and DLS fields are set. This tells BEX

312

that this is a Locate reply for a DLU. As a result, step

708

is skipped so that BEX

312

does not modify the resource ownership hierarchy in the Locate reply as it would for an independent LU. This leaves DLUR

316

as the owning node in the resource hierarchy. At step

710

, BEX

312

does replace the DLUR

316

TG vectors in the Locate reply with its own TG vectors, as it does for independent LUs. This is necessary so that DLUS

304

can calculate a route to DLUR

316

. At step

714

, BEX

312

forwards the Locate reply to BEX

306

. When BEX

306

receives the Locate reply, step

716

determines that the message is from a downstream node. Thus, step

718

is executed and BEX

306

determines that both the OCR and DLS indicators are set in the reply. Since both indicators are set, BEX

306

realizes that this is a reply for a DLU. Just as discussed for BEX

312

, step

720

is omitted so that BEX

306

does not modify the resource ownership hierarchy in the Locate reply. However, at step

722

, BEX

306

does replace the BEX

312

TG vectors in the Locate reply with its own TG vectors. BEX

306

forwards the reply to NN

302

and DLUS

304

at step

726

. At this point NN

302

sends a Locate search request at step

728

into the network to find the application that the DLU has requested. This Locate request contains both the TG vectors that NN

302

received from BEX

306

and an indication that this Locate request is initiated by a request from a DLU. The Locate request eventually finds its way to NN

308

, which owns the application sought. Node

308

also determines from the DLU indication in the Locate request that it must send a BIND command to the DLU to establish communications. Therefore, at step

802

of

FIG. 8

, NN

308

uses the BEX

306

TG vectors contained in the Locate request to calculate a route to BEX

306

. At step

804

, NN

308

sends the BIND to BEX

308

. The BIND identifies the DLU. BEX

306

is able to calculate a route to BEX

312

to the DLU and it forwards the BIND accordingly to BEX

312

, which in turn calculates the remaining route to EN

314

and DLUR

316

and forwards the BIND. DLUR

316

contains the conventional functionality to now establish a session with the DLU, which DLUR

316

supports.

It is to be understood that the above described arrangements are merely illustrative of the application of principles of the invention and that other arrangements may be devised by workers skilled in the art without departing from the spirit and scope of the invention.