Communication system concentrator configurable to different access methods

Description Submit all comments and votes

FIELD OF THE INVENTION

The invention relates generally to a generic multichannel backplane bus architecture and its application thereof and more specifically to a Local Area Network (LAN) system architecture, and is also directly applicable to implementations of Wide Area Network (WAN) systems, and more particularly to a system for providing multiple LAN protocols, simultaneous multiple protocol-independent generic channels of communication on a backplane, as well as a unique means of LAN media connections to the backplane, and unique and useful systems of monitoring and controlling a LAN environment.

BACKGROUND OF THE INVENTION

Various systems for Local Area Networks are known from the prior art. However, these systems are all based on a particular medium and other particular standards which pose significant problems in the field.

The LAN standards in existence that pertains to this disclosure are listed below.

Digital Equipment Corporation/Intel/Xerox Ethernet.TM. Version 2.0

ISO/ANSI/IEEE 802.3 CSMA/CD,

ISO/ANSI/IEEE 802.4 Token Bus

ISO/ANSI/IEEE 802.5 Token Ring

ISO/ANSI X3T9.5 FDDI (Fiber optic Distributed Data Interface), a Token Passing Ring.

All of the above are networking protocols, and each standard specifies the Media Access methods (MAC), and Logical communication Link Control methods (LLC).

The concept of the "backplane bus" is well established in the computer and communications field; examples of standardized and proprietary backplane buses abound. In the most general terms, a backplane bus is a wiring center common to a single platform and shared by a number of users (generally implemented as printed circuit boards plugged into the bus via a single connector).

The backplane bus is generally used for communications between the users according to some common protocol. Prior art arrangements in this field require that the bus be dedicated to a single bus access protocol, and that all users of the bus communicate across the bus using only this protocol.

The protocol consists of an access method, data format and data rate, all of which must be common to all users of the bus. In general, only one user of the bus may generate data "packets" onto the bus at any instant; access to the bus is governed by a distributed mechanism common to all users of the bus, termed the access method.

Specifically in LAN applications of backplane buses, there are two well established access methods: Carrier Sense, Multiple Access with Collision Detection (CSMA/CD) and Token Passing. Token passing further distinguishes to a physical ring and physical bus manifestation. All of these access methods are used with multiple data rates and data formats, generating numerous protocols; in addition, there are other protocols which combine elements of both CSMA/CD and Token Passing, as well as protocols which use only some elements of the access methods (e.g. Carrier Sense, Multiple Access without Collision Detection).

Prior art in this field provides a single, unique backplane bus designed specifically for a single protocol. An example of an implementation of a single protocol backplane bus is found in the Multiconnect.TM. Repeater from 3Com Corporation. This product offers a backplane bus dedicated to the IEEE 802.3 10 MegaBit/Second CSMA/CD protocol (commonly known as Ethernet.TM.).

Additional prior art in this field provides multiple unique, separate backplane buses, each of which is specifically designed to support a single protocol. An example of this implementation is found in the System 3000.TM. Wiring Concentrator from Synoptics Corporation. This product offers four independent backplane bus protocols, each dedicated uniquely to one of the following protocols:

1. IEEE 802.3 10 Megabit/Second CSMA/CD (Ethernet.TM.)

2. IEEE 802.5 4 Megabit/Second Token Passing Ring

3. IEEE 802.5 16 Megabit/Second Token Passing Ring

4. ANSI X3T9.5 100 Megabit/Second Token Passing Ring (FDDI)

Additional prior art arrangements in this field provide a single backplane bus implementing a single protocol, and require the users of the bus to support protocol conversion between external multiple protocol interfaces and the internal protocol. All of these prior art arrangements suffer from the following limitations:

1. The backplane bus is dedicated to a single protocol, allowing no flexibility for growth or change.

2. The backplane bus supports only a single data path, allowing only one data packet to be generated onto the bus at any time.

3. Attempts to address these limitations in the prior art lead to higher costs: additional backplane buses or complex protocol converters (e.g. network bridges).

4. Each module within a system cannot be operating independently from the backplane network.

The concept of the logical network and the physical network configuration has been only attainable through a physical embodiment of physical network connections. All of the LAN/WAN (Local Area Network/ Wide Area Network) networking hub implementations are dedicated to, a single protocol, or to dedicated protocol channels to interconnect network users. By this pre-determined use of backplanes and their functional and physical definitions, the concept of module switching among generic channels was not possible.

Arrangements providing similar functions are known. The Access One.TM. from Ungermann-Bass, System 3000.TM. from Synoptics, MMNC.TM. from Cabletron, Multiconnect.TM. from 3Com corporations are examples of prior art arrangements.

Access One.TM. from Ungermann-Bass, System 3000.TM. from Synoptics, MMAC.TM. from Cabletron, Multiconnect.TM. from 3Com all employ dedicated protocol channels, and most offer one protocol and one channel only.

SUMMARY AND OBJECTS OF THE INVENTION

It is an object of the invention to provide a system and an arrangement for networks of all media types, including twisted pair, fiber optic, thin and thick coaxial cable and others by employing a concentrator which is modular and which provides a fault tolerant platform for one or more networks.

It is a further object of the invention to provide a system with communication channels which can be reconfigure at any time to a new protocol and support communication using the new protocol without providing a new management means.

According to the invention, a plurality of identical communication channels are provided in the form of a multiple channel concentrator with a plurality of slots. Each slot connects to the generic channels such that modules connected into the slot connect with the channels. Media channels and management channels are connected to the concentrator and provide the ability to operate several networks using different protocols concurrently with a single system concentrator.

According to the invention a unique means of connecting media to the backplane is provided by module switching to any channel. This uses the concept of a generic channel architecture. This module switching section provides means for switching each module, of any protocol or any media interface type, to any available channel thereby assigning the channel to a specific protocol dynamically. This invention is further improved by each port associated with a particular module switching to any channel. This allows building of logical network configuration with a physical network connections.

Several different types of modules exist in a LAN system hub according to the invention. The following describes and defines each type.

Control Module: A central module that performs functions uniquely related to a hub. Example of these type of cards are the repeater module in the Multiconnect.TM., and Re-timing module in the System 3000.TM.. The functions associated with this module is often integrated with other type of modules to lower the cost of a system.

Media Module: A module that interfaces to a LAN media specified in any of the aforementioned Standards. Each standard defines electrical, mechanical, functional, and optical, if applicable, interfaces for a particular media. A media module often has multiple ports.

Network Management Module: A module that interfaces the backplane that monitors LAN system status, controls system and module parameters, configures system and module parameters, and reports a LAN system status and network statistics.

Bridge Module: A module that implements any type of store-and-forward-function for any purpose. It either converts one protocol to another, or filters (receive all transmissions from one port, and selectively transmits to another port, or both).

According to the invention the connection of a given medium to the backplane is provided by a module switching element which has the ability to physically switch each module to any channel. This module switching provides modules, of any type listed above, with the capability to establish a LAN connection to the backplane, and to be connected or to any channel or to be isolated from the backplane allowing each module to continue to operate without any LAN connection to other modules.

This module switching element again uses the concept of the generic channel architecture. This allows for the switching of each port of a module of any protocol or any media interface card to any channel and allowing for the building of a logical network configuration with physical network connections.

The concept of logical network and the physical network configuration has been only attainable through a physical embodiment of physical network connections. Thus this invention provides dynamic control over physical network connections.

The module switching to any channel, allows a logical network by allowing any module to switch from one channel to another. This switching is restricted in that the switched module, switches all the parts connected to that module to another channel. This restriction forces users to pre-sort connections. Physical port switching does not have this restriction. Thus, module switching is further improved by allowing each port of a module to switch to any channel. For the purpose of economics, both module and port switching inventions are deemed uniquely and proportionally useful in applications needing network topological flexibility.

According to a further aspect of the invention, in order to organize the LAN/WAN, a management channel is provided. This management channel may be used for multi-management module arbitration with allowance to multiple agents.

Remote management of a LAN system is a well known function that most LAN system vendors offer. The intelligent entity within a LAN system that communicates to a host is called a Network Management Agent (NMA). All known implementations of the NMA integrate LAN management as well as vendor-specific hub system management.

System 3000.TM. from Synoptics, and MMAC.TM. from Cabletron, employ a dedicated management master for the purpose of network management functions in a Hub. All of the known multiplicity of management functions in these systems are, at most, provided for redundancy to the primary function.

This invention conceptually separates hub management functions from the network management functions. The architecture that separates the two functions which traditionally have been implemented as an integrated set of functions, provides the following useful features:

1. It allows multiple LAN Management Modules in one hub. This also allows redundant Management Modules.

2. It allows two multiple Network Management Agents pertaining to the operation with the Hub Management functions and its election.

This architecture of the present invention allows new and useful features to be incorporated in existing LAN/WAN protocols. These features are included in the protocols used according to the present invention.

According to the invention the Ethernet.TM. system has been improved. The invention provides a deterministic and precise method for collision detection on the backplane using slot-ID. Further precise collision counting is provided per-port. Still further, the invention allows a synchronous Ethernet.TM. channel, and half the repeater implementations of Ethernet.TM. modules.

According to the invention, collision-detection on the backplane using slot-IDs is provided allowing purely digital (binary) electronic circuits to detect contention of two or more simultaneous Ethernet.TM. transmissions on a backplane. The occurrence of this contention is called Collision.

Etherne.TM. is also called CSMA/CD, Carrier-Sense, Multiple Access/Collision Detection. The prior art is the Ethernet.TM. implementations by many companies such as Xerox, Digital Equipment Corporation, Intel and the ISO/ANSI/IEEE 802.3 standards implemented by numerous companies. The Ethernet.TM. and 802.3 standards each specify analog collision detection in coaxial media (10Base5 and 10Base2), serial bit-compare with analog collision detection in broadband coaxial media (10Broad36), and logical collision detection in the twisted pair and fiber point-to-point media (10BaseT, 10BaseF-Draft, respectively).

Most of the Ethernet.TM. HUB vendors employ analog collision detection on their backplane, in a similar way to that of 10Base5 or 10Base2, where either current or voltage is summed to represent the number of simultaneous transmissions. A voltage or current level that represents two or more simultaneous transmissions is deemed a collision.

The precision collision counting per-port allows accurate counting of collision statistics in an Ethernet.TM. network. In addition, this invention allows association of each collision to the port of its origination by the use of the port-ID.

This function is deemed useful for network statistics for management purposes, and this specific function is well known in the LAN industry. Many Ethernet.TM. implementations from numerous companies purport to provide this function.

Older implementations have used the collision condition sometimes detected when the monitoring LAN controller was not an active participant in the collision. This has been proven to be inaccurate and unreliable, because Ethernet.TM. LAN does not pass this information from one LAN segment to another, when separated by a repeater(s). Additionally, all known modern implementations use Ethernet.TM. controller ICs that can receive Ethernet.TM. transmissions promiscuously and receive transmissions less in length than a minimum sized packet.

The limitations with the above approach is that some collisions consist of all preamble patterns (repeating 1010..10 pattern) which is ignored and not reported by all known Ethernet.TM. controller ICs. In addition, some collisions consist of a preamble phase violation such as 1010..100. When a double zero is detected before a double one (last two bits of the Start Frame Delimiter, SFD), 10..1011, all known Ethernet.TM. controller ignore the subsequent reception, thereby ignoring the collision. Both of these cases are common in Ethernet.TM. based networks. This invention does not have the aforementioned limitations.

An additional limitation with the above approach resides in not distinguishing a locally occurring collision, i.e. among ports in a local hub, from a collision resulting elsewhere. Such collision statistics are less useful when these conditions are not distinguished. A locally occurring collision denotes congestion at the hub where the measurement is being made, whereas a remotely occurring collision denotes congestion at the hub other than where the measurement is being made. This distinction allows network users to optimize an Ethernet.TM. network with a use of Bridges or Routers. This invention distinguishes between local and remote collisions.

The Ethernet.TM. Statistics Counting per-Port supplements a Ethernet.TM. controller to extend its promiscuously received Ethernet.TM. packet statistics to associate with a particular port of a particular module.

This function is deemed useful for network statistics for network management purposes, such as fault isolation, traffic monitoring, etc., and implementations similar to this known in the LAN industry. These implementations use a set of registers that parallel the Ethernet.TM. controller IC's. Each network transmission records the slot-ID in the register. A local intelligence, usually through a microprocess, associates the contents of Slot-ID register to the received packet. When there is no discrepancy between the number of IDs in the Slot-ID register and the number of received Ethernet.TM. packets, this scheme functions properly.

The limitations with the above approach is that some Ethernet.TM. transmissions results in un-receivable packets in some network collision scenarios due to the inter-packet shrinkage, a well-known problem among knowledgeable users of the IEEE 802.3 standard. A collision may observe an inter-packet gap of a few bit times (a bit time is defined as 100 nano-seconds). When this type of network scenario occurs, the Ethernet.TM. controller IC may not receive the subsequent packet following the gap and ignore its activity. An additional limitation with the above approach is that an Ethernet.TM. controller IC does not have packet buffers to store incoming packets. All known Ethernet.TM. type controller IC's ignore the activities and do not receive incoming packets. In both of these cases, Slot-ID register is not guaranteed to be consistent with the received packet, and the resulting received packet statistics can be grossly incorrect. This invention closely monitors the Ethernet.TM. controller IC and its operation to associate the slot and port IDs to the received packet such that the device according to the invention is free of the aforementioned limitations.

The half-repeater implementations of Ethernet.TM. modules describes ISO/ANSI/IEEE 802.3 repeater implementations on the Ethernet.TM. channel. This invention causes approximately half the repeater limitations resulting from delay variability, thereby allowing a greater number of repeater functions to be present in hubs cascaded in series without violating the limit pertaining to the number of repeated sets by the 802.3 standard.

The repeater specified in 802.3 is limited to up to four in cascade. This limitation comes from preamble losses and its regeneration, and Ethernet.TM. device delay variations contributing to a dynamic packet gap (gap between two packets) space. The limitation comes from the shrinkages of the gap. This well-known limitation restricts Ethernet.TM. network topology, when interconnected via repeaters.

The restriction of four repeaters comes from the accumulated shrinkage of this gap through four repeaters and other media access devices. This phenomena and its restriction of Ethernet.TM. network topology to up to four repeaters in cascade, is well-known in the Ethernet.TM. LAN industry.

Another scheme known in the industry is to reduce the other device delay variations that contributes to the dynamic packet gap spaces thereby reducing the amount of gap shrinkage. This is called Inter-Packet Gap Shrinkage, IPGS. By reducing the shrinkage enough, more repeaters in cascade may be allowed.

This invention describes a novel and unique way to allow more hubs with IEEE 802.3 repeater functions in a direct cascade without exceeding the limits used to set the number of repeater allowances of four in direct cascade.

In the Token Ring and FDDI protocols, improvements have been made in establishing the Token rings.

A token ring comprises dedicated point-to-point links connecting one station to the next, until a physical ring is formed. Many products that emulate such physical connections are available. It is well known in the industry to provide a physical, electro-mechanical, bypass switches for establishing a ring.

Also, it is possible to establish multiple logical rings of compatible speeds of 4 MegaBPS and 16 MegaBPS through the configuration management and the data rate detection.

The improvements to the Token Ring protocol involve establishing the rings by means of slot-ID, and by slot-ID and speed detection. The improvements to the FDDI ring protocol involve establishing the rings of parallel data paths by means of slot-ID.

According to the invention token bus improvements are provided which allow the ISO/ANSI/IEEE 802.4 Token Bus protocol on the backplane.

The ISO/ANSI/IEEE 802.4 Token Bus standard is well understood and used in Manufacturing Automation Protocol. This is a Token Passing protocol in a physical bus and a logical ring topology. Most of the products implementing the Token Bus protocol uses bus topology. The bus topology's limitation is its difficulty to maintain, control, and configured the network, compared to the structured wiring made possible by this invention.

To implement Token Bus protocol, a Token Bus channel must provide a means of detecting data corruption and means of broadcasting data transmissions to all the other end-nodes. Both of these are common with the Ethernet.TM. network protocol. A bigger FIFO is provided to service increased packet size of 32 Kilo-bytes of Token Bus, compared to the 1.5 Kilobytes of Ethernet.TM., and an additional data line is provided to carry non-data symbols. Ethernet.TM. modules with these modifications carry packets with Token Bus Protocol.

The collision or in this case, data corruption detected is implemented by the detecting two or more simultaneous transmissions. Collision is detected in the same way as described in Ethernet.TM., as disclosed above. In addition, all data transmission from any port are broadcasted to all other ports, except during a collision. During a collision, all data transmissions from any port are broadcasted to all the ports.

According to the invention a Multiple Generic Channel, Multiple Protocol, Backplane Bus Architecture is provided which defines the Network Bus Architecture. This Multichannel Backplane Bus provides the following:

1. A single physical backplane bus which is logically partitioned into a number of generic data channels. Each of these channels consist of a