WO1996029802A1

WO1996029802A1 - Expandable port mobility for network repeater

Info

Publication number: WO1996029802A1
Application number: PCT/US1996/000774
Authority: WO
Inventors: Ian Crayford; Stephen Mcrobert
Original assignee: Advanced Micro Devices, Inc.
Priority date: 1995-03-23
Filing date: 1996-01-23
Publication date: 1996-09-26
Also published as: JPH11514508A

Abstract

An integrated repeater (100) for implementing per-port Port Mobility is provided as part of a semiconductor integrated repeater package. The integrated repeater includes a number of Pseudo-AUI ports for connecting to transceiver devices (104). Each repeater device is coupled into a collision domain. By connecting one transceiver device to multiple repeaters in different collision domains, and controlling the connection to the repeater, the network port associated with the transceiver device is moved from one collision domain into another. Corresponding Pseudo-AUI ports on repeaters not part of the desired collision domain are isolated (using a tri-state gate). Various repeater resources are shared across the various network ports, even though they may be in different collision domains.

Description

EXPANDABLE PORT MOBILITY FOR NETWORK REPEATER

BACKGROUND OF THE INVENTION The field of the invention relates generally to multiport repeaters for local area networks, and more particularly to bandwidth and reconfiguration management in a LAN based on multiport repeaters or hubs.

Evolution of 802.3 Standards;

Traditional Ethernet (820.3, 10BASE5, ISO/IEC 8802- 3:1990 (E) ANSI/EEE Std 802.3-1990 Edition, Section 8) and Cheapernet (802.3, 10BASE2 ISO/IEC 8802-3:1990 (E) ANSI/EEE Std 802.3-1990 Edition, Section 10), are coaxial wired systems, both standards are incorporated as reference for all purposes. The coaxial cable provides the linear bus to which all nodes are connected. Signaling is accomplished using a current sink technique wherein a center conductor is used for signal, and a shield used as a ground reference.

Fig. 1 is a diagram of a conventional local area network (LAN) 10 using coaxial bus topology. LAN 10 includes several nodes (or data terminal equipment (DTE) ) 12 coupled to a Media Attachment Unit (MAU) 14 by an Attachment Unit Interface (AUI) cable 16. Each MAU 14 is coupled to a coaxial cable 18 and is adapted for a particular type of media (e.g., coaxial cable 18) . Thus, MAU 14 is media dependent, having different configurations for different types of medium.

Twisted Pair Ethernet (802.3 10BASE-T ISO/IEC 8802- 3:1990 (E) ANSI/EEE Std 802.3-1990 Edition, Section 13/14) utilizes standard voice grade telephone cable (22-26 gauge) , employing separate transmit and receive pairs (4 wires) . Fig. 2 is a diagram of a conventional LAN 30 using twisted pair Ethernet. LAN 30 uses a star topology. At the center of the star is a "repeater" 32. Repeater 32 (or hub) performs signal amplitude and timing restoration. Repeater 32 takes an incoming bit stream received from a twisted pair MAU 34 attached to a 10BASE-T cable 36, and repeats the bit stream to all other ports connected to it (but not back to the originating port) . In this sense, repeater 32 acts as a "logical coax", so that any node connected to LAN 30 detects another node's transmission. Differential signaling is employed with one pair acting as the transmit path, and the other as a receive path.

In some implementations, node 12 may be integrated with MAU 34 to produce an integrated node/MAU 38. Repeater 32, in some embodiments, is divisible into a repeater logic function 40 coupled to a number of twisted pair MAUs 34, one twisted pair MAU 34 per port. Repeater logic function 40 is media independent.

While repeaters are used in traditionally wired coaxial Ethernet to extend the network's physical distance limits, 10BASE-T mandates the use of a repeater to actually provide the connectivity function if more than two nodes are required (clearly a requirement for any practical network) . Although the physical signaling on the cabling differs, the functionality of the repeater is identical in either coax or twisted pair Ethernet networks, as is the frame (or packet) format that is used to pass messages between the participating nodes on the network.

Multiple end stations are connected to a repeater via the ports of the repeater. A repeater may also be connected to other repeaters. When a repeater receives a packet on one of its ports, it retransmits the packet to all other ports.

Data is transmitted on the Ethernet LAN in the form of packets. Within the packet there is a destination address field, followed by a source address field, type/length field, and data field. The end station that transmits a packet to another station places its own unique address in the source address field of the packet, and the address of the recipient device in the destination address.

In a typical repeater based network, data that is transmitted from one end station to another may pass through one or more repeaters. Notice that regardless of where the 3 packet is destined, it is repeated to all devices on the network, whether or not it is intended for the recipient. This is by definition the way a "shared media" LAN operates. Note that there is nothing preventing anyone from connecting an unauthorized end station onto the repeater network and "eavesdropping" on the conversations on the network, nor in a standard based repeater is there anything preventing an unauthorized station from transmitting traffic onto the network. Several application relating to operation of "secure repeaters", which prevent such unauthorized access, have been made by the assignee.

Emerging Applications:

Ethernet networks have evolved from their initial deployment in expensive high-end connectivity applications, where there were relatively few stations which remained geographically static. The coax based Ethernet topologies were well suited to these environments, where each station connects to the coaxial bus. Now, Ethernet is being used as the preferred network solution for high volume PCs and workstations. As a result, network sizes have grown. This growth has been accompanied by some serious issues regarding network management (monitoring and keeping the network running) and reconfiguration (responding to additions and changes) .

In addition to the numbers of network nodes growing, so has the reliance of current and emerging applications on networking. Many applications make heavy demands on the network resource, such as file sharing, shared databases, and other forms of "collaborative computing". The combined growth in size of networks as well as the growth of network aware applications has in some cases led to the network itself becoming a communication (and hence computational) bottleneck, as aggregate bandwidth on the demands made on the network exceed physical bandwidth limitations.

As applications continue to demand additional bandwidth, the venerable first generation LAN solutions are reaching the limit of their aggregate bandwidth capabilities. Despite the fact that any station can for a small period of time maintain a burst bandwidth transaction with another station at (in the case of Ethernet) 10 Mb/s, this cannot be sustained as an aggregate bandwidth since many stations contend for the same 10 Mb/s channel, and ultimately (assuming complete fairness) are assigned a portion of the aggregate bandwidth according to their individual needs. This leads to perceived delay at some stations, particularly the users of those stations that are the most demanding of attention. A potentially more serious issue than mere response time delay, is that of emerging applications not being able to operate in this environment. As new interactive services become available, which are sensitive to time variance (the difference in packet-to-packet delay rather than absolute delay) , the problem becomes more acute. Video playback and

Real Time Video (RTV) applications are unlikely to be able to use highly loaded shared LANs, where stations (not directly involved in the real time application) are generating any significant traffic. Although these applications are "niche" at present, they stand the potential to become mainstream in the future as compute and communications power increases.

Twisted pair Ethernet (10BASE-T) overcomes many of the management and reconfiguration problems associated with shared media LANs, by providing a star topology based on a centralized repeater. The repeater acts as the central interconnect point for all the 10BASE-T connected stations, and can therefore be used as the concentration point to provide management functions such as fault detection and statistics gathering. However, such an architecture does nothing to address the needs of applications that require additional burst or aggregate bandwidth, due to either large numbers of users and/or time sensitive applications. Although each station has its own dedicated wiring, the repeater effectively provides the function of a "logical coax", and the entire aggregate bandwidth of the LAN is still shared exactly as it would be if the stations were connected to a multidrop coax bus.

For applications that require bounded time variance services, there is simply only one alternative - reduce the number of stations on the network contending for the aggregate bandwidth. In the limit, if the number of stations on any Ethernet LAN segment is reduced to a single station, then the aggregate and burst bandwidth become equivalent, and the station effectively receives the full 10 Mb/s allocation. An interesting trend emerging in the networking industry is a move to enhance Ethernet performance, using technologies such as "Switched Ethernet" and "Full Duplex Ethernet". Both of these technologies are aimed at providing additional migration paths for existing 10 Mb/s Ethernet installations. In the case of "switched Ethernet", the existing 10 Mb/s Ethernet controller is preserved at the end- station, and the repeater is replaced with a switch device, capable of supporting multiple simultaneous port-to-port conversations. In the case of "Full Duplex Ethernet", the end station is replaced or upgraded to support concurrent transmit and receive activity, and the repeater is replaced with a switch capable of both multiple simultaneous port-to-port conversations and concurrent transmit and receive activity on any of the ports.

Fig. 3 is a diagram of a LAN 50 implementing "micro- segmentation." In the case of large numbers of users contending for the aggregate bandwidth, a trend has been towards micro segmentation. In this approach, clusters of (normally work function related) end-stations 52 are configured as a "workgroup" 54_i, i-1 to n. Each workgroup 54 is then typically permitted access to a full 10 Mb/s aggregate bandwidth. In this case, stations 52 in each workgroup 54 are typically interconnected by a workgroup associated repeater 32. LAN 50 includes a switch 56 to selectively connect each workgroup 54 to a server 58. Note that each station 52 functionally includes a MAU and DTE as shown in Fig. 2, for example. Sometimes demands of certain end station applications exceed even the sharing of bandwidth with a small workgroup, then a "dedicated Ethernet" approach may be used. Fig. 4 is a diagram of a dedicated Ethernet LAN 60. Dedicated Ethernet LAN 60 is essentially micro-segmentation taken to its logical extreme, where only one end-station 52 is connected to each port of switch 56. Since each port of switch 56 achieves up to the full 10 Mb/s, each station 52 achieves an effective "dedicated" 10 Mb/s. An enhancement to dedicated Ethernet is the use of "Full Duplex Ethernet". This can be used for higher performance and enhanced support for interactive applications, where simultaneous transmit and receive activity is important. However, since most of the installed base of Ethernet end stations only operate in half duplex mode, this almost invariably requires the end station to be upgraded. Notice also that the use of switched Ethernet does not mandate full duplex operation, but the use of full duplex Ethernet mandates the use of a switch. At the network hub, this means that a repeater is inadequate (since it can only deal with a single active receive port at any one time) . The repeater therefore is replaced with a "switch" function, which is able to support multiple simultaneous transmit and receive operations, provide a required level of "store and forward" buffering, and route a data packet according to its source/destination address characteristics. This may be by means of the MAC address (in which case it is technically a bridge) , or an internetwork address (in which case it would be classed as a router) . The integration of this bridge/router functionality into a high performance multiport unit is effectively what the industry refers to as a "switch" in Ethernet terms. From an external perspective, a receiving end-station will be unable to determine if the packet has been delivered by a repeater, bridge, router or switch.

Not all users need full 10 Mb/s capability, hence the use of "micro-segmentation" (the capability to partition the LAN into groups of stations) , dependent on their bandwidth requirements. Under this architecture, some nodes will be part of a shared LAN, and some will have an entire segment dedicated to a single station. Most will be a compromise of these two scenarios.

Servers benefit the most from a dedicated 10 Mb/s service. For one thing, servers typically have a multi¬ threaded and multi-tasking operating system - the vast majority of client (end stations) which make up the majority of the "user population" do not (at least at present) . Note that the benefit to servers may not be limited to just allocating a server to a dedicated 10 Mb/s Ethernet segment. Other alternatives exist which may also alleviate bandwidth aggregation issues on congested devices. For instance, supplying a higher bandwidth "pipe" such as one of the emerging 100 Mb/s Ethernet standards (e.g., ANSI/IEEE 802.3a 100BASE-T draft suppl.), or providing multiple 10 Mb/s Ethernet connections to connect to a single server. In this latter example, application software effectively manages distribution of traffic across the multiple dedicated Ethernet connections, so that the load is distributed across (for instance) four dedicated connections to the server. This approach clearly increases the aggregate bandwidth of the overall network. It is undesirable to segment a network into "workgroups", only to subsequently force each workgroup to access the same server. What is desirable is separate workgroup servers, or additional bandwidth into one server, or a combination of both.

Further discussion of a "Switched Ethernet" hub function is outside the scope of this disclosure. However, since the use of LAN switching is becoming increasing more prevalent, the need for "switch capable" LAN components is becoming more important.

While the increase of bandwidth to specific groups or users is key to micro-segmentation, so is the need to be able to modify and/or adjust the network topology, to meet the changing needs of individual users and workgroups. For instance, the clients within a workgroup may need to be adjusted due to staff addition, reduction or location change. An individual user may also require different bandwidth requirements for a specific project, or may move from one workgroup (project or department related) to another.

Although micro-segmentation is useful to address increasing bandwidth needs, significant problems still remain in the area of reconfiguration. For instance, in the previous examples of Fig. 3, moving a user from Workgroup l to Workgroup 2 requires that the physical location of the repeaters (typically a wiring closet) and the appropriate cables to be identified, and cabling to be moved. Of course, the chance of error, and the continual upkeep of the changing configuration of the LAN 60 are major concerns in this scenario.

An automated approach to this reconfiguration function is required. One approach that has been taken is to provide multiple repeater units in a single enclosure, each with the capability to connect to an independent "collision domain". Fig. 5 is a schematic block diagram of a LAN 80 implementing Group Mobility™ functionality. LAN 80 includes a single enclosure 82 that houses three independent repeaters 84 (for example, AMD's IMR+ and HIMIB components manufactured and distributed by Advanced Micro Devices of Sunnyvale, California) , each repeater 84 having access to any one of three collision domains. In a typical implementation, since each IMR+/HIMIB repeater 84 combination supports 8 10BASE-T ports, each workgroup of 8 users is able to be effectively connected to an independent collision domain, hence each workgroup effectively accesses a full 10 Mb/s bandwidth for a single collision domain that is simply shared among other users. By providing the three backplanes as external connections, a switch 56 or similar function is able to be connected to these different collision domains, and traffic between the workgroups, or to external servers or backbone resources also connected to the switch (not shown) , is able to be forwarded based on appropriate rules. These rules may be a simple as MAC address filtering, or more complex including features such as protocol routing and/or security algorithms. Note that the number of separate repeaters, collision domains, and number of users within a single repeater are all conveniently scaleable, dependent on the requirements of the device. For instance, a hub with 5 collision domains, 3 separate repeaters, and 8 or 16 ports per repeater, could easily be manufactured. This approach is referred to as "Group Mobility™"', since it allows a defined workgroup 54x to be moved from one collision domain to another, under software control (by simply configuring switch 56 to connect workgroup 54x to the appropriate backplane) . While this approach allows groups of workers to be moved conveniently, it is inflexible to move individual users. Since movement or change in needs of a user is more representative of the real business environment, an alternative approach is required to address this.

SUMMARY OF THE INVENTION A solution to a problem of moving one or more desired users to a desired collision domain is referred to a "Port Mobility™". The preferred embodiment describes implementation of a repeater chip set that enables and/or performs "Port Mobility™".

A network designer or manager often has a dilemma. Additional bandwidth may presently be required by some users and/or workgroups. However, the cost of simply applying a switch (dedicated) Ethernet port to every user is prohibitively expensive and may be overkill for many non- demanding users. This combined with the need to adapt the network to dynamic project and/or workgroup needs imposes a significant challenge, especially as the users who require high bandwidth change locations and workgroup memberships. As suggested above, a static scheme that requires rewiring each time that the users and/or workgroups change their requirements, is undesirable.

If over time the cost of switched Ethernet is reduced to a modest premium over that of a repeater, the use of switched Ethernet could replace the traditional repeater. However, the functionality of a switch is significantly more complex than that of a repeater (i.e., packet buffering in a switch versus storing a small number of bits in a repeater) , and currently switch ports are considerably more expensive than repeater ports (in the order of 4 to 10 times more expensive) .

If the switch function is augmented with a Port Mobility™ capable repeater, the overall hybrid solution offers the scaleability needed for reconfiguration, as well as the availability of additional bandwidth to users and/or workgroups requiring it. This is the benefit of the "Switch Capable" or Port Mobility™ capable repeater:

While hubs with the Port Mobility™ capability have been available previously, this feature has been reserved to the domain of high-end, expensive devices. To date, these have been fixed configuration devices, where the decision to implement Port Mobility™ had to be taken at the onset of the hub purchase cycle. Where mobility and/or bandwidth were not of paramount importance, the use of a standard repeater was adequate. However, if at a later date either of these features were required, they are typically impossible to retrofit to a normal, single collision domain repeater.

Users who did not currently need the features, but who considered they may need these capabilities in the future and wanted to plan ahead, were faced with purchasing a port switching hub, incurring significant additional up-front cost, as well as management software complexity.

Some vendors upgrade path for switching is to retrofit a completely new backplane to the installed hub. The original backplane with a single collision domain is removed and replaced with a new backplane with the requisite number of collision domains. The vendor then typically supplies new modules to plug into this new backplane. After all this, the only things left of the original hub are the power supply and the chassis (metalwork) .

The "Switch Capable" design offers a low cost entry point, with only a small incremental cost over that of a single collision domain hub, but allows the system to be expanded when needed to accommodate additional bandwidth and/or management needs. In addition, it allows other network management and security features to be added in a scaleable, "pay-as-you-go" approach.

According to one aspect of the invention directed towards a system for implementing Port Mobility, it includes two repeaters, each having a port for connection to a transceiver providing a network port. Each repeater is connected into a collision domain, and the network port is directed into the desired collision domain by selectively activating a desired port of the repeater coupled to the desired collision domain.

Reference to the remaining portions of the specification, including the drawing and claims, will realize other features and advantages of the present invention. Further features and advantages of the present invention, as well as the structure and operation of various embodiments of the present invention, are described in detail below with respect to accompanying drawing. In the drawings, like reference numbers indicate identical or functionally similar elements.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a diagram of a conventional local area network (LAN) using coaxial bus topology;

Fig. 2 is a diagram of a conventional LAN using twisted pair Ethernet;

Fig. 3 is a diagram of a LAN implementing "micro- segmentation;"

Fig. 4 is a diagram of a dedicated Ethernet LAN; Fig. 5 is a schematic block diagram of a LAN implementing Group Mobility™ functionality;

Fig. 6 is a block diagram of an improved repeater implementing a preferred embodiment of the present invention;

Fig. 7 is a schematic block diagram of a hub implementing a twelve port repeater to selectively route any port into any one of three collision domains;

Fig. 8 is a schematic block diagram of a LAN implementing a Port Mobility™ function using hub shown in Fig. 7 in conjunction with a switch 56;

Fig. 9 is a perspective view of one implementation of hub/repeater shown in Fig. 8;

Fig. 10 is a block diagram of a repeater illustrating a preferred implementation for scaleable security; and Fig. 11 is an alternative preferred embodiment that provides Port Mobility™ by reducing the number of ports per backplane (collision domain) , with the possibility to offer multiple collision domains on a single repeater chip.

DESCRIPTION OF THE PREFERRED EMBODIMENT Fig. 6 is a block diagram of an improved repeater 100 implementing a preferred embodiment of the present invention. Repeater 100 may be used in enclosure 82 shown in Fig. 5 to provide Port Mobility™ to selected stations 52 in a workgroup. Repeater 100 is switch capable and includes two major components, a media independent repeater 102 and a number of transceiver devices 104. Repeater 102 performs the functions of a managed repeater circuit, and contains the repeater state machine(s) and the management counters (MIB variables), as described in the incorporated IEEE 802.3 standard, for a fourteen port repeater. Each transceiver device 104 in the preferred embodiment is a multi-MAU interface circuit (called the Quad Integrated Ethernet Transceiver or QuIET™ that is available commercially from Advanced Micro Devices of Sunnyvale, California) , that integrates four independent 10BASE-T transceivers with integrated waveshape control. Repeater 102 additionally includes ports for a reversible AUI port and an AUI port. Repeater 102 is the managed repeater circuit, with

12 Pseudo-AUI (PAUI) ports, one fully compliant AUI port, and one Reversible AUI Port (RAUI) . Repeater 102 also includes an expansion bus that facilitates multiple repeaters 102 to be cascaded together, in order to construct higher port density repeaters. One implementation of the expansion port is identified in US Patent 5,265,123 for Expandable Repeater. The expansion bus has both a synchronous and an asynchronous mode of operation. It is the synchronous mode of operation that is described in US Patent Number 5,265,123 (Expandable Repeater) . An asynchronous mode is described in US Patent

Application MANAGED ASYNCHRONOUS EXPANDABLE BUS FOR ETHERNET REPEATERS, Serial No. unknown, filed March 20, 1995, incorporated herein by reference for all purposes. The PAUI is a single ended interface that retains the normal voltage and timing characteristics of the 802.3 AUI specification, but allows use of single interface pins for the Data Out (DO) , Data In (DI) and Control In (CI) functions. This essentially halves the pin requirement for the PAUI over the 2 pin interface requirement for a compliant AUI, and hence substantially reduces the chip cost. The PAUI is described in copending patent application titled PSEUDO-AUI LINE DRIVER AND RECEIVER CELLS FOR ETHERNET APPLICATIONS filed February 21, 1995 and incorporated herein by reference for all purposes.

The AUI port is fully compliant 802.3 interface, using the conventional 6 pin implementation for DO, DI and CI. The RAUI operates as a differential AUI port of a DTE (CI is an input) or can be configured as "reversed" (CI is an output) , in which case it allows direct connection to a management MAC device with no external logic. RAUI is described in US Patent application titled REVERSIBLE AUI PORT FOR ETHERNET filed February 21, 1995 and incorporated herein by reference for all purposes. Each 4 port transceiver device 104 of the preferred embodiment is a PAUI to 10BASE-T interface, each 10BASE-T interface providing waveshaping that reduces the need for external filtering components. Transceiver device 104 may be replaced with any normal AUI based transceiver, for example to interface to 10BASE-F or 10BASE2 networks, by providing a simple single ended to differential conversion circuit in the PAUI between repeater 102 and the transceiver.

Fig. 7 is a schematic block diagram of a hub 200 implementing a twelve port repeater to selectively route any port into any one of three collision domains. Hub 200 includes one logical repeater 102x per collision domain (which may be a combination of repeaters synchronized by the expansion bus) , with each transceiver device 104 coupled to each repeater 102. This single-point to multi-point connection is not possible with an 802.3 compliant AUI connection. Repeater 102 provides a mechanism to "isolate" each of the PAUI ports. This isolation mechanism provides the ability to direct a physical port into any of the collision domains, under software control.

For instance, by un-isolating a selected port on repeater 102_χ coupled to Collision Domain 1, and isolating the same port on the other repeaters coupled to the other collision domains, e.g., repeater 102₂ and 102₃, the particular port is thereby connected only into Collision Domain 1. Any port may be moved under software control, using a simple "break-before-make" sequence of instructions wherein the port to be moved is first isolated from its collision domain before being coupled into another collision domain.

For example, the particular port coupled into collision domain 1 as identified above would be removed from Collision Domain 1 by isolating the particular port on repeater 102_1# and thereafter un-isolating the corresponding particular port on repeater 102₂ coupled to Collision Domain 2.

The port isolate function of repeater I02x for implementing Port Mobility™ may be implemented differently than the normal port enable/disable management function (PortAdminControl) that is described in Clause 19 of the incorporated 802.3 standard (Layer Management for 10 Mb/s

Baseband Repeaters) . Clause 19 defines the PortAdminControl action to enable and disable a port, and the PortAdminState attribute to reflect the state of the port. When the state of the port is changed from disabled to enabled, the auto- partitioning state machine will be reset. When a port is disabled, it neither transmits or receives. However, changing a port from an isolated state to an unisolated state, may not reset the partitioning state machine.

In the preferred embodiment of the present invention, when a particular port is isolated on repeater

102x, the PAUI outputs for the particular port are tri-stated, and the PAUI inputs are in a high impedance state. Data on the network is received by the appropriate transceiver device 104 port. The PAUI from transceiver device 104 port delivers received data from the network to the DI circuit. Each repeater 102x determines, based on the isolate state of the pertinent port, whether to accept the receive data on it's DI input and also forward it to the other devices in the collision domain. Repeater 102x will forward the received data to all enabled and un-isolated ports, as well as to the expansion bus (the appropriate collision domain coupled to the un-isolated port) . Repeater data placed by a controller onto the DO pin of a port, is received by the DO receiver on transceiver device 104, and transmitted to the network. A transceiver device 104 which detects simultaneous transmit and receive activity constituting a collision will drive the CI output circuit with a nominal 10MHz waveform. Each repeater 102x determines, based on the isolate state of the pertinent port, whether to accept and propagate the collision to other ports in the collision domain. The preferred embodiment of repeater 102 includes a control, such as an external pin, to set an operational mode when repeater 102 is dynamically inserted into an existing network. In one mode, a multi- collision mode, repeater 102 isolates all of its pseudo-AUI input ports, requiring a higher level control to un-isolate selected ports and steer them into the collision domain serviced by the newly added repeater 102. In another mode, a single collision domain mode, repeater 102 powers up in a ready state, automatically responding to any input signals present at any of its pseudo-AUI inputs. When repeater 102 includes only these modes, a single pin may be used for the control. Fig. 8 is a schematic block diagram of a LAN 250 implementing a Port Mobility™ function using hub 200 shown in Fig. 7 in conjunction with a switch 56. As described above, hub 200 incorporates a crosspoint switch function 252 to couple (using software in the preferred embodiment to control the isolate function) any station 52 from any workgroup 54 into any desired collision domain by switching desired ports into a repeater coupled to the desired collision domain. In the preferred embodiment, this crosspoint function 252 is implemented using the PAUI for single-point to multi-point connection, as well as controlling the isolation/un-isolation functions of each repeater.

There are some key advantages to the IMR2 chip set architectural approach, relating to scaleability, management, security, and bandwidth/fault management.

Scaleability

Repeater 102 allows a number of collision domains to be defined by adding additional repeater 102 devices. In practice, there may be a practical limit of six collision domains imposed due to bus loading effects in order to allow repeater 102 and transceiver devices 104 to adequately transmit and receive over the multi-drop PAUI lines. In addition, the number of ports per collision domain may be increased in increments of whatever number of ports repeater 102 is provided with, which in the preferred embodiment is 14 (12 10BASE-T, 1 AUI, 1 RAUI), by adding additional repeaters 102 into the same collision domain. Repeaters 102 in the same collision domain are preferably interfaced using the expansion bus and a single expansion bus interface, which is well- documented elsewhere and will not be further described herein. In the preferred embodiment, only a single transceiver port is required for each network port, although the number of repeaters 102 depends on the number of collision domains and the number of ports per collision domain. For example, a 24 port, 3 collision domain repeater would require 6 repeaters 102 and 6 transceiver devices 104. In addition, as transceiver device 104 of the preferred embodiment is replaceable with an alternate 802.3 compatible transceiver, then Port Mobility™ is also possible for fiber and coaxial networks. Using a coaxial medium, should multiple stations be present on a coax port as permitted by the standard, all stations on that coaxial port move from one collision domain to another using the Port Mobility™ function.

One benefit of this overall approach is that a single collision domain repeater, such as repeater 100 shown in Fig. 6, may be constructed. By using the present invention, this single collision domain repeater would have the ability to have additional collision domains retrofitted at some time in the future.

Fig. 9 is a perspective view of one implementation of hub/repeater 200 shown in Fig. 8. Hub 200 includes a single collision domain repeater standard module 300 for interfacing to each port 302 of eight ports. Also included are two optional expansion modules 304 to permit the eight ports to access any of three collision domains. By providing access to the PAUI, plug-in expansion module 304 (containing a repeater 102 and an appropriate expansion connector (not shown) ) , is insertable into hub 200 to augment the number of collision domains. This allows an inexpensive entry level repeater, hub 200 with only the standard single collision domain module 300, to be offered. Hub 200 has the ability to expand as network demands change.

Management

Each of repeater 102 of the preferred embodiment provides all of the required management counters for a fully compliant implementation of the 802.3 repeater MIB (defined in 802.3k, Clause 19, Layer Management for 10 Mb/s Baseband Repeaters) . Each repeater 102 also provides management counters specific to implementing some of the required RMON MIB functions (IETF, RFC 1757, formally RFC 1271). The provision of these hardware implemented counters ensure that any overhead of updating packet based statistics is hidden from the management function. As additional ports, and collision domains, are added, since all of the traffic related management counters are maintained within the repeater 102, it is not necessary to add additional performance to the management function. This also significantly influences the ability to be able to offer a scaleable function.

Repeater 102 also provides the RAUI port to allow connection of a management MAC. The management MAC is able to be connected to RAUI of a repeater 102 to allow it to receive and transmit in-band management packets, communicating with a remote network manager. A unique advantage that preferred embodiment for the repeater 102 architecture offers is a reduction in cost of the in-band management function. A single management module interfaces with multiple collision domains by simply connecting it to the appropriate repeater 102 associated with the desired collision domain. One way that this may be performed relatively easily is by use of a selector circuit (not shown) , that allows the single management MAC to connect to a single RAUI at any time. Essentially, this allows a "roving MAC" to be implemented, which, under control of the remote network manager, is selectively and controllably directed to connect to and analyze a specific collision domain. In addition, the management function may be implemented such that in the event of severe network errors or disruption, the "roving MAC" automatically moves to an alternate collision domain in order to inform the network manager via another route, that there is a problem. Note that these features also enhance the scaleability in that the "roving MAC" selector logic may be added to the base repeater functionality, with the appropriate logic to allow interface to the maximum number of available collision domains that can be implemented.

Security

Many new repeaters offer security features to prevent unauthorized receipt of data (eavesdrop protection) and/or the ability to connect to the network using an unauthorized location or piece of equipment (intrusion control) . An alternate preferred embodiment of the present invention provides a unique architecture to allow security features to be scaled easily in a Port Mobility™ application.

Fig. 10 is a block diagram of repeater 102 illustrating a preferred implementation for scaleable security. With this, only one security Logic 350 CAM, including one address detect/compare and embodiment disrupt circuit is required per repeater core 352. Repeater core 352 includes the required 802.3 repeater state machines, provides the expansion bus backplane for a plurality of PAUI ports 354 and the AUI and RAUI ports. Since security CAM Logic 354 is shared by all the possible network ports, the preferred embodiment provides a flexible and expandable security system. A single CAM location that stores (for instance) a destination address of an end station (or multicast address) , is mapped to one or more ports for the purposes of disrupting the output data stream. This is described in US Patent Application "PROGRAMMABLE ADDRESS MAPPING MATRIX FOR SECURE NETWORKS" filed December 30, 1994, hereby expressly incorporated by reference for all purposes. Security features will not be further described herein.

Fig. 11 is an alternative preferred embodiment that provides Port Mobility™ by reducing the number of ports per backplane (collision domain) , with the possibility to offer multiple collision domains on a single repeater chip 400. Fig. 11 illustrates a flexible approach that offers one network port per collision domain. Repeater chip 400 includes four separate repeaters implemented in a single device.

Despite the apparent flexibility, in that a switch matrix between the expansion buses of repeater chip 400 would allow any port 354x to be mapped to any arbitrary number of collision domains, the architecture of Fig. 11 is not a lowest cost entry repeater or solution, nor is it a lowest cost expandable security repeater having Port Mobility™.

Since it is unknown until a port count and a collision domain count is established, which ports will be in the same collision domain, and since any repeater can be in any collision domain, each repeater core must be supported with a dedicated Receive MAC Engine, as well as dedicated security logic. To build an N number port single collision domain repeater requires all of the expansion busses to be interfaced together (the approach of Fig. 10 required no external expansion bus logic for an N number port single collision domain repeater) .

In addition, for purposes of the following discussion, we assume that there are 2 entries (addresses) in each CAM that are dedicated to a single port. Addresses not used by a port (if it only requires a single individual address for instance) cannot be allocated to another port easily, since there is no simple way to associate the address locations with the port of another repeater core, which may not be in the same collision domain. Adding addresses to this architecture merely increases the potential wastage, but is the only simple way to have multiple address locations accessible for each individual port. In comparison, since the preferred embodiment for repeater 102 only stores addresses for a single collision domain, all addresses can potentially be shared, and single address locations can further be shared across multiple ports.

Bandwidth/Fault Management

Use of the variety of feature in combination make the preferred repeater 102 a unique solution for intelligent and dynamic management of network bandwidth and fault isolation. By merging the switch and Port Mobility™ repeater functions, and using an intelligent management function, the ports of the devices are allocated according to either traffic or error conditions. For instance, when a station continually communicates with other stations across the switch (i.e., in another collision domain) versus within the station's own collision domain, the management process could consider moving the station. Alternatively, if a particular collision domain exhibits any high error statistics, ports can be isolated and/or moved, so that the offending stations causing the error can be traced. Other applications include backing up devices by moving them to another collision domain for the period of the back-up (off-peak hours for instance) . This could essentially allow all stations to be moved, one at a time, to be backed-up, then returned to their original collision domain. Therefore, the above description should not be taken as limiting the scope of the invention which is defined by the appended claims.

Claims

WHAT IS CLAIMED IS:

1. A system for implementing Port Mobility™, comprising: a first repeater having a first expansion bus coupled into a first collision domain and having a first network port; a second repeater having a second expansion bus coupled into a second collision domain and having a first network port; and a first transceiver coupled to said first network ports, for coupling into one of said first and said second collision domains by activating one of said first network ports in one of said repeaters.

2. The system of claim 1 wherein said first repeater further includes a second network port and said second repeater further includes a second network port, and further comprising: a second transceiver coupled to said second network ports, for coupling into one of said first and said second collision domains by activating one of said second network ports in one of said repeaters.

3. The system of claim 2 wherein said second transceiver is coupled into a different collision domain than said first transceiver.

4. A method for adding an additional collision domain into a network having a hub that includes a first repeater coupled into a first collision domain, wherein the first repeater includes a first port coupled to a first transceiver device and a second port coupled to a second transceiver device, comprising the steps of: adding an expansion module into the hub, wherein said expansion module includes a second repeater having a first port, a second port, and an expansion bus; coupling said first port of said second repeater to the first transceiver device and coupling said second port of said second repeater to the second transceiver device when said expansion module is added into the hub; and coupling said expansion bus into the additional collision domain when said expansion module is added into the hub.

5. The additional collision domain adding method of claim 4 further comprising the steps of: isolating the second port of the first repeater; and thereafter unisolating said second port of said second repeater to couple the second transceiver device into the additional collision domain.

6. The additional collision domain adding method of claim 4 wherein said repeater includes a control for placing said second repeater into a multi-collision domain mode, the method further comprising the steps of: placing said second repeater into said ulti- collision domain mode prior to coupling the transceiver devices to said first and said second ports of said second repeater device.

7. The additional collision domain adding method of claim 6 further comprising the step of: unisolating said second port of said second repeater to couple the second transceiver device into the additional collision domain after adding said expansion module into the hub.

8. The additional collision domain adding method of claim 4 further comprising the steps of: adding a second expansion module into the hub, wherein said second expansion module includes a third repeater having a first port, a second port, and a second expansion bus; coupling said first port of said third repeater to the first transceiver device and coupling said second port of said third repeater to the second transceiver device when said expansion module is added into the hub; and coupling said second expansion bus into a second additional collision domain when said expansion module is added into the hub.

9. The additional collision domain adding method of claim 5 wherein said isolating and unisolating steps are performed under software control to issue instructions directly to the first repeater and to said second repeater.

10. The additional collision domain adding method of claim 5 wherein said isolating and unisolating steps are performed automatically in response to detection of a network condition.

11. The additional collision domain adding method of claim 10 wherein said network condition is detection of network traffic above a predetermined level on the first collision domain.

12. The additional collision domain adding method of claim 10 wherein said network condition is detection of network errors above a predetermined level on the first collision domain.

13. The additional collision domain adding method of claim 4 wherein the first repeater and said second repeater each include a resource, and wherein said resource is shared among transceiver devices coupled to unisolated ports of each repeater.

14. The additional collision domain adding method of claim 13 wherein said shared resource is a security system.

15. The additional collision domain adding method of claim 4 wherein the network includes a MAC function coupled to the first transceiver, the method further comprising the steps of : using said MAC function with respect to the first collision domain by unisolating the first port of the first repeater and isolating the first port of the second repeater; and using said MAC function with respect to the additional collision domain by isolating the first port of the first repeater and unisolating the first port of the second repeater.

16. The additional collision domain adding method of claim 4 further comprising the steps of: adding additional ports to the first collision domains by adding a third repeater having an additional port, wherein said third repeater is synchronized with the first repeater by an expansion bus.