Data processing: artificial intelligence – Neural network – Learning task
Reexamination Certificate
1999-10-05
2003-06-10
Follansbee, John (Department: 2121)
Data processing: artificial intelligence
Neural network
Learning task
Reexamination Certificate
active
06578021
ABSTRACT:
TECHNICAL FIELD
This invention relates to network administration applications and more particularly to applications allowing administration of switched LANs (Local Area Networks) organized in Virtual LANs.
BACKGROUND ART
Bridged and router based LAN networks are migrating to switched LAN networks. With the increasing size of networks and the type of traffic in the LANs which is rather broadcast than unicast, it appears that the traditional LAN is not sufficient. The LAN protocol such as Ethernet and Token Ring is responsible for regulating the traffic within the communication channel so that only one device can make use of the communication channel to send data at one time. From one other point, the operations a router has to perform for each packet are the following: look up the (OSI) Layer
3
address in its tables and determine the outbound port, update the layer
3
header (for example, to decrement the hop count) and strip and replace the Layer
2
header. Furthermore in the routers these operations are implemented in software. These operations have to be done by all the routers in the path between two end stations that want to communicate; the more routers, the more work. On the contrary, switching technology is faster, easier and cheaper; it rather relies on hardware implementations which improve costs and performance. A LAN switch, like a multiple port bridge, learns all the MAC addresses in the segments to which its ports attach and bridges the traffic between its ports.
A second advantage of the switched LAN network is the possibility of building VLANs or logical subsets of a physical network or networks. A VLAN is a LAN which maps workstations on some other basis than the geographic location: the workstations may be grouped for example, by department, type of user or primary application. The grouping into broadcast domains, was previously done in hubs and routers at layer
3
; unfortunately, if the address of the device changed (for example, due to a move of the physical workstation), an administrator had to go to the device and assign a new network address. With VLANs, the grouping is independent of the physical location. A VLAN could be based on physical address location, MAC address, network address or some other defining characteristics such as the protocol. A standardization of the VLAN for legacy LANs is in process under the reference IEEE 802.1Q. As with the VLANs the traffic is directed to only those users that need it, the bandwidth usage is improved and thus, the performance of the network is improved. One other advantage of the VLANs is to be able to build scalable switched networks: the hierarchical deployment of switched networks will help in deployment of large switched networks. Finally, one major advantage of the VLANs is to avoid manual reconfiguration of the hubs when endstations are moved/added: whatever will be the new physical address of the user, it will still keep its grouping in LAN. For instance with IP networks, the IP addresses are conserved when the stations are moved.
The network administrator in charge of the VLAN configuration defines the initial configuration and can change or add workstations and manage load balancing and bandwidth allocation more adequately than with the LANs. Network management software keeps track of relating the virtual picture of the local area network with the actual physical picture. This operation of defining the VLAN configurations is performed from the network management station by entering through the user interface, commands for the switching devices connecting the workstations over the network. VLAN manager programs such as ClearVISN VLAN Manager of COMPAQ is a graphical SNMP (Simple Network Management Protocol) application that enables network managers to configure and manage VLANs graphically. This program particularly helps in configuring port-defined VLANs and manage moves, adds and changes through graphical software. Whatever the tools used to configure the VLANs in the switching devices, the VLAN administrator always has to classify the network workstations by VLANs.
A first problem with VLAN classification is the time consumed by the network administrator to group network devices by VLAN on the basis of the criterion chosen for classification. Without any method this is time consuming and subject to errors. While vendors have promoted the simplicity of Virtual LANs, a reason they are not popular today, given the perceived benefits, is the manual operations usually required for the initial setup, especially if the network to be configured contains several thousand nodes.
A second problem to classify workstation by VLANs is the choice of the criterion. Classification can be port based: it is the simplest form of VLANs which is just a collection of different ports in a LAN switch or number of switches. But it cannot address the problem of adding devices or changing physical location of device in the network. There are also MAC address based VLANs: in each VLAN there is a list of MAC addresses. This model can track network devices automatically when they change their location. However it may not be so easy to manage a large number of MAC addresses. It is also possible to build a VLAN based on workstation IP subnet address, the IPX subnet address and so on . . . . It gives much flexibility to the network administrator and is easier to manage than the MAC addresses based VLANs. However, the most flexible VLAN implementation is the policy based VLAN which can include all the ways mentioned above of defining a VLAN and it is possible to select a method suitable for a particular network. These policy based methods are mostly used today to obtain the best optimization of the bandwidth. The difficulty to establish a good criterion to classify VLANs is to extract useful patterns from voluminous amount of data.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide an automatic method for VLAN classification.
A further object of the present invention is to provide an automatic VLAN classification which optimizes the bandwidth use while taking into account the evolution of the traffic changes in the network.
These objects are achieved by a method for classifying network devices in VLANs in a switched network having a network management application managing said network devices, said method being characterized in that it comprises the steps of:
collecting network management variables stored on said network devices using said network management application;
feeding a Neural Network operating on a computer with said network management variables reflecting the activity of said network devices, and
running said neural network, said neural network providing an output list of network devices grouped in VLANs according to the value of the network management variables.
By using the network elements (physical devices or applications running on such devices) addresses communicating together and forming VLANs on the basis of this criterion, the network administrator will be sure that the broadcast domain formed per VLAN, will optimize the use of network bandwidth; but the major interest of the solution is the use of a neural network able to automatically classify and also the fact of obtaining this information, which is stored in the ARP tables of the switches, in a network management variable stored by the network devices: this classification can be done automatically and periodically refreshed because the network devices maintain the information up to date. The application automatically provides a VLAN list to the administrator who can then update the VLAN definitions in the switches of the network; furthermore, using the classifier of the invention, one can easily add to this application after classification steps an automatic update of switches in the network with the VLAN definition via the network management protocol facilities.
REFERENCES:
patent: 5373486 (1994-12-01), Dowla et al.
patent: 6085238 (2000-07-01), Yuasa et al.
Kishan Mehrotra et al; Artificial Neural Networks; 1997; M
Flynn John D.
Follansbee John
Grosser George E.
Hirl Joseph P.
LandOfFree
Method and system for classifying network devices in virtual... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for classifying network devices in virtual..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for classifying network devices in virtual... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3160353