Multiplex communications – Data flow congestion prevention or control – Flow control of data transmission through a network
Reexamination Certificate
1996-10-24
2002-02-05
Hsu, Alpus H. (Department: 2662)
Multiplex communications
Data flow congestion prevention or control
Flow control of data transmission through a network
C370S252000, C370S401000, C370S465000, C709S234000, C709S250000
Reexamination Certificate
active
06345041
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field
The invention relates to electronic communications networks. More particularly, the invention relates to the automatic configuration of hardware and software networking products for network segmentation to balance the flow of communications and data between nodes on such networks.
2. Description of the Prior Art
A network segment consists of repeaters, end stations, and connectivity between the stations (for example, see standard ISO/IEC 8802-3 for information on Ethernet networks). A set of predefined rules for communicating, called “protocols,” determines how and when end-nodes on the network segment are allowed to communicate with one another, and what form such communication takes.
In a data communications network, data are transferred in the form of packets which contain the address of the sending station (the source address) and the address of the intended recipient (the destination address). The exact packet format, and the predefined rules for interpreting the contents of such packets, are again specific to the network type and are made up of several levels of protocols. As described herein, a network segment has the capacity to transfer a limited amount of data per second. When more than the maximum media speed of network data transfer capacity is required, multiple segments can be used. These segments can be interconnected through the use of bridges or other packet-forwarding devices, which allow communication between different segments (bridges are defined in standard ISO/IEC 10038).
The term switch has recently been applied to bridges. Switches are fast bridges which generally do not use a CPU to process the packets. Bridges contain two or more ports, each of which connects to a segment. When two stations on different segments communicate with each other, the bridge forwards the packets between the two segments. When the stations are on the same segment, the bridge does not forward the packets to any other segment. The bridge may buffer data received on its ports to allow forwarding of the data later onto a different segment that was busy when the transmission first occurred.
Over time, a network may grow as more users (end-nodes) are added. Each of these new end-nodes must then compete with existing end-nodes for access to the network. As the number of end-nodes which might potentially access the network medium grows, it becomes statistically more likely that the medium is in use when a given end-node wants to transmit, thereby requiring the end-node to wait. In addition, available network resources may be constrained when just a few end-nodes have heavy resource impact. Nodes which transmit a lot of packets, which transmit a lot of broadcast or multicast packets (received by all or many stations), or which create a lot of network disruptions (e.g., collisions, late collisions, etc.), consume a lot of the available network resources, leaving less for other end-nodes. Thus, schemes which increase the amount of network resources available (i.e. the amount of potential accessibility to the network medium), without changing the type of physical network used (preserving the users' investment), are of great use and interest.
When the network medium is shared (as opposed to switched), one way of increasing the amount of available network resources is to split the shared medium into multiple segments (i.e., segment the network). This limits the number of users who have access to any one of these segments. If it is desired to have end-nodes in a segment communicate with end-nodes in other segments, bridging or switching or routing across the segments may be employed, so that network packets for devices that are not in the immediate segment are forwarded until they reach the segment of the destination end-node.
While segmenting the network medium can improve resource availability, the choice of where to segment, e.g. which nodes should be assigned to which segment, affects the success of the results. If all of the nodes which impact network resources the most remain on the same segment, access for any one of those nodes is not much improved by this scheme, while the less resource-impacting nodes on other segments have greatly increased opportunity for access but far less need of it. In this situation, the increased resources are wasted on the wrong end-nodes.
Additionally, network systems are inherently complex. A significant amount of technical education is required to understand, configure, and maintain any particular type of physical network, e.g. 10Base-T, Ethernet, Token Ring, or FDDI networks. As the use of networks becomes widespread, the number of users responsible for such networks grows, while those among them who have sufficient knowledge to accomplish these tasks declines. Thus, schemes which can configure, maintain, or improve network conditions without requiring much specific knowledge on the part of the user are also of great use and interest. Such ease-of-use, i.e. user-friendly, schemes are often encompassed in software which is either embedded in network devices or external to them. These software applications generally are known as intelligent network agents.
One network products vendor is presently marketing a 10Base-T network switch that allows any of its ports to be assigned to any one of four internal backplane segments. The product is described as having a load-balancing feature which can be activated in one of several ways:
On-demand by the user (ie. a manual trigger);
At a designated time or interval; or
When a user-defined threshold is reached (this is a threshold of a network value defined in the product's Management Information Base (MIB) and configured via the Simple Network Management Protocol (SNMP). For example, the MIB value could be a collision count, a segment utilization pattern, or a switch utilization pattern.
These mechanisms for triggering such actions are not new. The first two have existed in practice for some time, and the third is an aspect of recent implementations of the Remote MONitoring (RMON) suite of features. The algorithms that are actually used by the product's load-balancing feature to determine which ports to place on which of the four internal backplane segments have been informally described as being based on end-node communication patterns gleaned from address-pair information in the switch's address table. Thus, this approach requires that the switching fabric and intelligent network agent components are fully integrated into the intelligent agent's hardware environment.
Accordingly, the state of the art provides an automatic load-balancing solution that does not accommodate an inability on the part of the user or agent to communicate with the switching fabric (e.g., packet-forwarding, address table, etc.) hardware. This arrangement has a major disadvantage in that it is unusable with switches and other packet-forwarding devices which are fully external to the intelligent agent's hardware environment. It would therefore be advantageous to provide an agent that can intelligently improve network resource availability, without requiring special, integrated packet-forwarding hardware.
In addition, the prior method may result in unnecessary load-balancing. It would therefore be advantageous to reduce network disruption by performing such load-balancing only when it actually improves network resource availability.
Furthermore, it is advantageous to provide an agent that can intelligently improve network resource availability while requiring only a simple activate action on the part of the user, such that no further action or special knowledge of the agent, its environment, or networks in general (such as how to set a threshold level, or even which threshold values would be meaningful) is required on the part of the user.
SUMMARY OF THE INVENTION
The invention provides an intelligent network agent that implements a feature referred to as automatic load-balancing. The invention addresses both of the major issues of network resource impact and eas
Faulk, Jr. Robert L.
Kimball Karen E.
McGuire Robert M.
Finnegan Henderson Farabow Garrett & Dunner L.L.P.
Hewlett--Packard Company
Hsu Alpus H.
LandOfFree
Method and apparatus for automatic load-balancing on... 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 apparatus for automatic load-balancing on..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for automatic load-balancing on... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2971736