« PreviousContinue »
ENHANCED DIRECTORY SERVICES IN
COMPOUND WIDE/LOCAL AREA
TECHNICAL FIELD 5
This invention relates to enhanced central directory services in Wide Area Networks (WANs) and, more particularly, to the registration of user resources residing on nodes, possibly themselves interconnected by Local Area Networks (LANs), in such WANs, across boundaries between logical partitions or topology domains of the WAN.
BACKGROUND OF THE INVENTION
Distributed directory services for computer networks are 15 disclosed in A. E. Baratz et al. U.S. Pat. No. 4,914,571, granted Apr. 3, 1990. Resource identifications in the Baratz patent, however, are stored only by the network nodes (a network node being the node in a domain to which all that domain's end nodes are connected). As a result, the directory 20 data base is generally limited to a very small proportion of all of the resources in the network. This limitation greatly increases the number of broadcast searches needed to create routes in the network, i.e., all cross-domain searches must be initially broadcast. Thereafter remote resource data is 25 cached in the network node initiating the search.
The directory services of the Baratz patent can be further enhanced by the central directory server disclosed in "Advances in APPN Architecture," by R. Bird et al., IBM Systems Journal, Vol. 4, No. 3, pages 430-451 (1995). 30 Before conducting broadcast searches, Bird's network nodes consult a central directory server, which may have a larger storage and hold entries for more resources. The total number of directory broadcasts in the logical partition of a network served by one such central directory server is 35 thereby reduced from nxm, where n is the number of network nodes and m is the number of resources, to m. That is, only the central directory need conduct broadcast searches, and resource information obtained on behalf of one network node may later be supplied to satisfy a different 40 network node's query. While the Bird technique does reduce the number of broadcast searches, it does not eliminate such broadcast searches.
It has become increasingly common to use Wide Area Networks to interconnect large numbers of Local Area 45 Networks (LANs). It is difficult, however, to provide resource directory services in such compound networks because the LAN resources often pre-exist the interconnecting WAN and hence are not constrained to utilize LAN resource identifications which are necessarily different from 50 the previously assigned WAN resource identifications. In addition, the size of topology data bases used to control interconnections in the WAN are normally chosen to reflect the number of nodes and links in the WAN network. Typically, these topology data bases are therefore often too 55 small to contain all of the possibly non-unique identification information about all of the LAN resources which are connected to that WAN. This forces a very serious limitation on the maximum size to which such compound WAN/LAN networks can grow. Indeed, network users would prefer 60 internal interconnection services with no limit on ultimate size. To overcome these perceived limitations, users tend to partition WANs into separate topology domains, each of which may have its own central directory server(s), that can provide directory services only within that topological 65 domain. In such a scheme, a unique high-level qualifier, called the "NETID," precedes the resource identifier for all
network nodes and most end user resources within the same topological domain Providing unique resource identifications across the set of interconnected topological domains despite these pre-existing resource naming conventions is difficult. For example, for flexibility such naming schemes may permit end nodes to attach to topological domains with NETIDs different than their own, since an end node might well access multiple networks through dial-up transmission facilities.
This searching problem is exacerbated when the WAN takes the form of an Advanced Peer-to-Peer Networking (APPN®) network. APPN networks offer a powerful, flexible and easy-to-use networking solution for client/server and distributed applications. Such applications create many additional directory service problems, however, when attempting to use APPN for very large networks interconnecting large pluralities of LANs. The large number of APPN network nodes, and corresponding network node identifications, in such large APPN networks, creates even more serious searching problems across the APPN network.
Finally, existing limitations on the size of the available WAN topology data bases often make it difficult to select routes on the basis of class of service (requiring quality of service information in the data base). Customers would, of course, like to interconnect very large numbers of small LANs as gateways to the customer's branch offices. Each branch site, however, is usually itself configured as a LAN, thereby effectively doubling the possible number of topology data base entries. All of these constraints combine to render the provision of efficient directory services across such a large compound WAN/LAN network, whether or not an APPN network, difficult if not impossible.
SUMMARY OF THE INVENTION
In accordance with the illustrative embodiment of the present invention, enhanced directory services are provided in large, unconstrained, compound Wide Area/Local Area Networks by a hierarchical, automatically initiated directory registration process for all new resources added to a network at the time those new resources are initially added to the network. Such a central directory is designed to contain information about all resources in and connected to the WAN/LAN network. To support this registration activity, a central directory services server, together with a large central directory data base, is made available in one of the WAN network nodes. The size of the central directory data base used to store resource information can be made responsive to user requests, and hence be of any size necessary to support routing between any two resources in the compound network and still include Quality of Service parameters.
More particularly, each Local Area Network connected to the Wide Area Network is connected to the WAN through a WAN network node and, together with the connected WAN network node, forms a "resource domain" within which resources can be uniquely identified by identifying the network node domain in which they appear (as well as the resource end node and the resource itself). In order to increase the efficiency in the processing of such identifications, the domain network node identification is added to the end node user resource identifications to form a triplet. This triplet includes the identification of the resource itself, the identification of the domain in which the resource resides, and the identification of the domain network node in which the triplet resource itself resides. This triplet is always necessarily unique in the compound Wide Area Network.