This invention relates to communication networks, and more particularly, to an apparatus and method for Ethernet equipment in a ring topology.
As Ethernet is deployed in Metro and Access networks, and services are offered on these networks, there is a desire to maintain SONET-like resiliency (i.e. recover from a fault in less than 50 ms). One common means of providing resiliency is with a ring topology. However, Ethernet switches will not work properly if there is a ring or loop in the topology. Protocols such as IEEE 802.1d Spanning Tree Protocol (STP) or IEEE 802.1w Rapid Reconfiguration were invented to detect and remove loops. However, they are slow and cannot achieve path restoral within 50 ms as desired.
To solve this problem, the IEEE is working on 802.17 Resilient Packet Ring (RPR). Others are looking at Multiprotocol Label Switching (MPLS) with Fast Reroute capabilities. Both of these approaches are quite complex. RPR requires a new Media Access Control (MAC) Layer, and MPLS requires extensive signaling. Because of the complexities, these approaches will drive up the cost of the nodes on the ring.
This invention introduces a new way (Protected Switching Ring or “PSR”) of providing protection for Ethernet deployed in a ring topology with resiliency that does not require a new MAC layer, and that can be built using low cost Ethernet chips and methods.
This invention differs from some previous inventions. One of interest is described in U.S. Pat. No. 6,430,151, granted on Aug. 6, 2002. PSR is similar to '151 in that:
Some of the differences between PSR and the teachings of the '151 include:
Another approach to the problem is described in U.S. Pat. No. 4,354,267. The '267 patent describes a set of homogeneous layer 2 devices arranged in a ring. Each node in '267 forwards packets around the ring, and the originator removes the packet.
Some of the differences between PSR and the teachings of the '267 include:
Normal 802.3 Ethernet requires a tree topology. If a ring or a loop exists, then packets will be forwarded around the ring indefinitely. STP was created to solve this problem by detecting and breaking any rings. If the ring is broken, then there is no possibility of packets being propagated forever.
This invention shows how to virtually break the ring such that all nodes can communicate with each other, and how to remove the virtual break when a real failure occurs. This is accomplished by placing intelligent nodes on the ring that work together to virtually break and restore the ring.
In PSR, the nodes communicate between and among themselves to determine when and where a break occurs. The relevant state machines for a preferred embodiment of the present invention are contained within this disclosure.
The present invention will now be described more fully hereinafter with reference to the accompanying drawings, in order to disclose selected embodiments. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Like numbers refer to like elements throughout.
Overview
The Protected Switching Ring (PSR) enables building of partial or full ring topologies from low-cost Ethernet equipment, while providing for sub-50 millisecond recovery from equipment or link faults. PSR nodes support the transport of point-to-point port-switched connections across the ring topology. During normal, non-fault operation, one port in the ring will be blocked to user traffic, thus preventing a loop. In the event of a fault in the ring, the blocked port will be unblocked, allowing access to all nodes on the ring.
Two topologies using the present invention are described below. The first topology is the PSR Full Ring (“FR”) configuration that consists of a full ring of PSR nodes. Port-switch connections can be configured between any two subscriber ports on the ring. The second topology is the PSR High-Availability (HA) configuration. This configuration provides a partial-ring extension of a SONET or RPR ring, or a partial-ring addition to existing layer-2 switching equipment. In either case, a path is engineered through the existing equipment to complete the path for the PSR protocol traffic and user data.
PSR nodes are designated as Ring Arbiters or Relays. Each ring contains two Ring Arbiters. The Ring Arbiters communicate with a “hello” protocol to coordinate the blocking or forwarding of user traffic. In a preferred embodiment, the PSR Ring Arbiter ports take on the role of master or slave on the ring according to their relative node priority. In a preferred embodiment, the priority could be a unique identifier, such as a MAC address. In a highly preferred embodiment, the priority can be the concatenation of an operator-configurable priority with the MAC address (or other unique identifier) such that the priorities of two nodes would never be equal. In either of these preferred embodiments, the reception of a HELLO with the same priority would indicate a ring with only one arbiter, where that arbiter was receiving its own HELLOs.
In general, during normal fault-free operation of two Ring Arbiters, the slave Ring Arbiter will block one of its ring ports in order to terminate the ring loop. A ring may contain one or more Relay nodes. The Relay nodes may be distributed in any fashion around the ring, although some benefit is provided by distributing approximately equal numbers of Relays on each “side” of a full-ring configuration.
In addition to the “hello” protocol, each node performs a “discovery” protocol that allows each node to know about all the other nodes on the ring. The discovery protocol is also used to detect persistent ring faults and to generate the associated alarms. Both protocols operate at layer 2, employing reserved multi-cast MAC addresses.
IP connectivity between all ring nodes is accomplished over a control VLAN used only for that purpose. This allows Telnet and a UDP-based signaling protocol to operate between any nodes on the ring. (An explanation of Telnet is not critical to the understanding of the present invention but Telnet is a terminal emulation program used with TCP/IP networks that allows remote entry of commands that are treated as if input at the network device.) Bridging techniques are used to provide the connectivity for these IP-based applications; all user traffic is transported across the ring using port-switching. As such, all user traffic is point-to-point across the ring; traffic from a subscriber-port/VLAN on one node is connected to a subscriber-port/VLAN on another node.
An additional embodiment of the present invention addresses a partial failure of a network link so that the communication link is lost in only one direction. Yet, another embodiment uses a single arbiter to provide a high reliability connection of user ports to an existing network ring by creating a switching ring with the arbiter and two network ring access points.
Ring Nomenclature
When the PSR is configured, two ports are designated as the ring ports and may be called East and West ports. Also the node type is given to distinguish Ring Arbiter types and Relays (also called Ring Relays or Relay Nodes). The Ring Arbiter type may be High-Availability (HA) or Full-Ring (FR). The two Ring Arbiters on the ring must be of the same type. When speaking of a specific ring port, the partner port refers to the other port of the pair of ring ports on that Ring Arbiter or Ring Relay.
An additional distinction is made in the case of a HA Ring Arbiter. The port of the HA Ring Arbiter connected to the existing SONET or RPR ring is designated the “extension side” (ES) port. This port interfaces with the existing equipment for which we wish to extend a ring segment. The other Ring Arbiter port is referred to as the “ring side” port. It is connected to a string of one or more Ring Relays or directly to the other Ring Arbiter.
HELLO Protocol
Each PSR Ring Arbiter periodically issues a “HELLO” protocol packet out each ring port. In a preferred embodiment each PSR Ring Arbiter issues a “HELLO” protocol packet out each ring port every 10 milliseconds. The packet uses a special multicast MAC address as the destination address. The Relay nodes are configured to have the data plane pass the packet from one ring port to the other, so a Relay node adds only a small amount of delay as the packet moves from one Ring Arbiter to the other. The remote Ring Arbiter node will terminate the packet and send the packet to the control plane. The control plane uses the presence of the new packet and some control information to drive its state machine for the Ring Arbiter ports. The absence of a new HELLO message for 30 milliseconds constitutes a ring timeout. If the timeout persists for 1.5 seconds, a ring failure is declared and the appropriate alarm is issued.
The significantly longer period used as a trigger for a ring failure keeps a short intermittent problem from being deemed ring failures though the problems may be handled by the declaration of ring timeouts. In one embodiment, the ring failure is detected by loss of Discovery messages, described below. One of skill in the art could implement the ring failure to be based on the absence of HELLO messages rather than Discovery messages. One of skill in the art would appreciate that the HELLOs are not processed at the RR nodes, whereas the Discovery messages are. HELLOs therefore propagate around the ring faster than Discovery messages. A timeout threshold for loss of HELLOs can be set lower than an equivalent threshold for Discovery messages.
A ring timeout causes the state machines to transition a slave Ring Arbiter port to a FORWARDING state. This response ensures that any loss in connectivity due to a single failure across the ring will only persist for 50 milliseconds or less.
In a preferred embodiment the sequence number in the HELLO PDU is used at the receiving Ring Arbiter to distinguish the arrival of a new HELLO PDU. Those of skill in the art will recognize that other methods could be employed to detect the arrival of a new HELLO PDU. The Relay nodes do not process the HELLO PDUs; they only forward them between ring ports.
Discovery Protocol
The discovery protocol is an optional protocol that can be implemented in order to add functionality. Note since the discovery protocol is not a necessary requirement of the state machines for any of the Ring Arbiters, Protected Switching Rings in accordance with the teachings of the present invention could be implemented without implementing the discovery protocol.
The discovery protocol also uses a special multicast destination MAC, but runs every 500 milliseconds. The discovery PDU is originated by the Ring Arbiters, appended to by intervening Relay nodes, and terminated at the remote Ring Arbiter. As the discovery PDU traverses the path between Ring Arbiters, each node in the path appends its management IP address, egress port for the PDU, and node type to the PDU. Since the discovery messages are flowing in both directions on the ring, each node on the ring can see the path of nodes to each Ring Arbiter on each of its ring ports. For example in
Additionally, as each Ring Arbiter constructs the discovery message to send out a ring port, the Ring Arbiter adds the completed node list received at its partner port. This allows every node in the PSR to know all the IP addresses of the nodes in the ring.
In the event of a ring or node failure, the Relay nodes closest to the point of failure will originate the discovery message. In other words, if a relay fails to receive a discovery PDU from its upstream neighbor (due to a link or node failure), then the relay will create and send a discovery PDU in the downstream direction. All downstream nodes will detect that the Ring Arbiter is no longer the originator of the discovery message and declare a fault alarm. If a node either does not receive a Discovery message or receives a Discovery message without a ring Arbiter as the originator, a ring failure is declared after 1.5 seconds. The fault is cleared when the node receives a Discovery message with a ring Arbiter as the originator.
PSR Data Plane for User Traffic
User traffic may enter and leave the PSR at any Ring Arbiter or relay node. A PSR connection defines the entry and exit points for a full-duplex flow of user traffic across the ring. This flow is defined by a pair of port/VLAN ID/PSR Node Address tuples. The connection defines a path through the ring between 2 user ports, each residing on a PSR node, configured to carry the user traffic for specific or all VLAN IDs on that port.
As the user traffic enters the ring, a ring tag is added to the packet. The ring tag is a VLAN tag and is unique on the ring. The ring tag defines a given connection between two ring nodes. At the egress node of the PSR connection, the ring tag is removed from the frame before forwarding to the user port. In this way, the VLAN tags present in the user data are transparently transported across the ring. VLAN IDs used on one user port do not interfere with IDs used on another user port.
A PSR node is either an endpoint of a given connection or a transit node for that connection. A PSR node is an endpoint for a connection if one of its user ports is specified in the definition of the given connection. The node is a transit node if neither endpoint of the connection resides on the node. In either case, a switch table used by the data plane is configured on each PSR node to either terminate one end of a given connection or to act as a transit node for that connection. When a node is a transit node for a given connection, the node simply transfers frames from one ring port to the other, based on the ring tag, without modification. When a node is an endpoint node for a given connection, the data plane directs the data arriving on a ring port to the correct user port and removes the ring tag. Conversely, the node's data plane directs user packets from the given user port with the given VLAN ID to the ring ports, adding the correct ring tag in the process.
PSR Control Plane for Control Traffic
A PSR requires a mechanism to transport HELLO PDUs, discovery PDUs, and IP traffic for ring control applications between PSR nodes. While user traffic transport is transported using switching techniques, in a preferred embodiment the control functions are transported using bridging techniques. By using bridging techniques, full PSR node control connectivity is attained with all nodes appearing on the same IP subnet. This makes configuration much simpler.
One ring tag is reserved for PSR control traffic. The data plane uses learning procedures and forwarding table lookups to direct control traffic to the correct PSR node. Note that the use of learning procedures and forwarding table lookups for the direct control traffic imposes an overhead burden that is orders of magnitude smaller than the overhead needed to use learning techniques for user data traffic. In the preferred embodiment the HELLO and discovery messages use known multicast MAC addresses and thus do not add additional learned database entries to be forwarded. Flushing is not needed for the control traffic upon failure, recovery, or reconfiguration of the ring, as the new port entries are learned from bidirectional traffic after a short period of time.
While the use of bridging for control traffic is preferred, it is not required in order to implement the present invention. The present invention could be implemented to use switching techniques for data packets and some or all types of control traffic. Care must be taken in creating this variation that the control traffic described in this application as passing when data packets are blocked, must be allowed to pass.
Fault Detection
Assuming RA node 300 was the master Ring Arbiter, when RA node 325 detects the loss of communication; RA node 325 will start forwarding traffic to the right hand side of the ring onto link 1330 and accepting user traffic arriving on link 1330 and relaying the traffic to link 1335 and to the subscriber ports of RA 325. This will restore communications between all of the nodes on the ring. At this point, RA 325 is forwarding traffic on both ring ports. The ring port that is facing link 1335 is in MASTER FORWARDING state, and the ring port that is facing link 1330 is in SLAVE FORWARDING state.
Link Restoral
When link 1325 is restored, RA node 325 needs to block one of its ring ports to prevent a loop in the ring. When RA node 325 receives the first HELLO on link 1335 (due to the restoration of link 1325), RA node 325 will see that the partner port to the port that is facing link 1335 is in SLAVE FORWARDING state. RA node 325 will move the port that is facing link 1335 to the BLOCKING state. Assuming that the Ring ports of nodes 310 and 320 connected to link 1325 went to an OPER DOWN state during the failure, the TIMING state in the relay nodes 310 and 320 will prevent forwarding of traffic until the slave Ring Arbiter has time to switch from MASTER FORWARDING to BLOCKING on the 325 side of the Ring Arbiter. OPER DOWN is an indication from the physical or transport layer that a link is not operational. It is normally based on the detection of loss or corruption of the incoming electrical or optical signal.
The advance to the TIMING state is triggered by the reception of a HELLO message. This TIMING state delay in the resumption of operation of relay nodes 310 and 320 prevents duplicate packets from reaching a given destination when the failed link is restored. To illustrate the value of this delay in the Ring Relay ports, consider a message coming to Ring Relay 305 to a subscriber port connected with Ring Relay 310 just before link 1325 is restored. Ring Relay 305 operating normally will send the same message onto link 1300 and link 1320. The message traveling counterclockwise reaches Ring Relay 310 in a conventional way. The message traveling clockwise to Ring Relay 310 will pass through Ring Arbiter 325 onto link 1335 as the West Port is operating in MASTER FORWARDING. When link 1325 is restored, there is a path for a duplicate message to cross link 1325 to Ring Relay 310. This potential is eliminated if the Ring Relay ports undergo a suitable delay between receipt of the first HELLO and the ultimate state of FORWARDING. Note that the HELLO messages from Ring Arbiter 300 to Ring Arbiter 325 and from Ring Arbiter 325 to Ring Arbiter 300 will pass over link 1325 as soon as it is restored as the HELLO messages are not blocked at any port in any state.
The preferred embodiments disclose using a timing delay to ensure that a port progressing from OPER DOWN to operational delays sending data packets long enough for the slave arbiter to impose a virtual break. One of skill in the art will recognize that the use of the timer could be replaced by a control signal sent by the slave arbiter after it has successfully imposed the virtual break. In either case, the port does not go to fully operational until after the virtual break has been imposed to preclude the creation of a ring for data packets.
Fault Detection
Link Restoral
In a preferred embodiment, the HA ring favors the Ring Side once the fault is removed. Instead of leaving the slave Ring Arbiter ES port (the port facing link 1540) in the forwarding state and blocking the Ring Side port (the port facing link 1530), the HA slave Ring Arbiter 530 always returns to a FORWARDING state on the Ring Side segment and blocks the ES port.
The Ring Side segment of the HA ring is favored in order to minimize HA ring traffic on the existing SONET or RPR ring as this will cut some of the user traffic on the SONET ring segment between the Ring Access Equipment as one direction will be blocked (thus counterclockwise traffic from port 1440 will be blocked while clockwise traffic from 1400 will continue to travel on the SONET Ring.
Nomenclature for State Diagrams
In the following descriptions, “isMaster” is based on the last received HELLO. If no HELLO was ever received on the port, then isMaster is based on the partner's last HELLO. If no HELLOs have ever been received by either port, then isMaster is “true”. The Boolean variable “isSlave” is the logical negation of “isMaster”.
The term “PartnerHelloTimeout” indicates that the partner port's age timer has timed out. The designation “RxHello<Node” means a HELLO message has been received whose priority is less than that of the receiving node. This event would cause the receiving node to consider itself a master.
Full Ring Mode—Ring Arbiter
Description of States
The state machine of
Description of State Transitions
The table below describes the transitions of the state machine shown in FIG. 6.
Note the fd timer reference below runs using a time that is a small fraction of the time used for the age timers in the RA and Relay nodes. This ensures that the relays are timed for a period long enough after the expiration of the fd timer for the loop to be broken. For example, one embodiment uses a 10 millisecond timer for the RA and Relay nodes and the fd timer at just one “tick” (a single 10 millisecond delay). This 10 millisecond interval is a small fraction of the 30 millisecond interval used to detect a ring timeout when a new HELLO message does not arrive within that interval.
Note that the state machine for ring arbiters in the preferred embodiment does not wait indefinitely to see a HELLO as long as the ports of the ring arbiter are operationally UP. The goal is to let the parts of the network ring operate even if other portions of the network ring cannot operate.
The following table shows the complete state transitions for the Full-Ring Arbiter as shown FIG. 6.
In a preferred embodiment, every 10 milliseconds, the two ports are checked in the same order. The combination of variations in when the HELLOs were generated plus transit delays may cause one HELLO on one port to arrive before the other HELLO on the other port. In any case, since one port is checked before the other then the other, it always appears as though one HELLO arrives before the other. The order that the ports are checked does affect which slave port is set to BLOCKING on the full ring.
One of skill in the art will recognize that any embodiment that does not check one port before the other would need to address the case of two HELLOs arriving essentially simultaneously at the two ports.
Full Ring Mode—Ring Relay
Description of States
The state machine of
Description of State Transitions
The table below describes the transitions of the state machine shown in FIG. 7.
High Availability Mode—Ring Arbiter—Ring Side
Description of States
The state machine of
Description of State Transitions
The table below describes the transitions of the state machine shown in FIG. 8.
High Availability Mode—Ring Arbiter—Extension Side
Description of States
The state machine of
Description of State Transitions
The table below describes the transitions of the state machine shown in FIG. 9.
As shown in the sequence of events reported in the table below, the RS ports of the Arbiters are always forwarding, unless the ports are OPER DOWN. The point of interest is the state of the ES port of the slave Arbiter. In essence, the ES slave port is FORWARDING if there is a HELLO timeout on either the RS or ES.
Unidirectional Break
The control system described above assumes that a break in a network ring will be a bidirectional break as it connects both the clockwise and counterclockwise virtual breaks upon failure to receive a HELLO. This bidirectional response could cause a loop in the event of a unidirectional failure.
For example, when the network ring is fully operational, Master Arbiter 1000 can receive HELLOs from Slave Arbiter 1025 via link 11037, relay 1020, link 11027, relay 1010, and link 11012. Likewise, Slave Arbiter 1025 can receive HELLOs from Master Arbiter 1000 via link 11010, relay 1010, link 11025, relay 1020, and link 11035.
If link 11027 was cut but link 11025 was left in service, then the West port on Master Arbiter 1000 would soon stop receiving HELLOs from Slave Arbiter 1025, while Slave Arbiter 1025 continued to receive HELLOS from Master Arbiter 1000. In the previously described embodiment, this unidirectional cut at link 11027 would not trigger the Slave Arbiter 1025 to unblock as it continues to receive HELLOs from Master Arbiter 1000 across intact link 11025. Thus ring relay 1020 as well as connected subscriber ports 1045would be cut off from the east side of the ring as Slave Arbiter 1025 is still blocking data, including data that would otherwise travel from Slave Arbiter 1025 to ring relay 1020.
One alternative embodiment is to react to a port going to an OPER DOWN state by stopping the transmission of HELLOs and all data from that port in the opposite direction, effectively creating a virtual unidirectional break in the other direction. Hence when ring relay 1010 observes an OPER DOWN associated with link 11027, ring relay 1010 would stop sending HELLOs and all data on link 11025. After Slave Arbiter 1025 fails to receive HELLOs in an allotted time, the Slave Arbiter 1025 would remove the virtual break on its west side to allow data traffic from link 11035 to proceed towards link 11030 or the user ports 1060 and to allow traffic from link 11032 or user ports 1060 to proceed onto link 11037.
Dual Homing Using a Single Node Ring
In this alternative embodiment, the Slave Arbiter node 1130 would see its own HELLOs. As described in Table C, one side of 1130 (for example the West side of the Slave Arbiter connected to link 1160) would go to the SLAVE FORWARDING state and one side (for example, the East side of the Slave Arbiter connected to link 1170) would go to the BLOCKED state.
Now, in response to a fault on the Ring Access Equipment 1110 or the link 1160, the East side of the Slave Arbiter 1130 would unblock, and the User Ports 1140 would continue to have access to the network. The network access for User Ports 1140 is therefore protected against faults in either the access links (1160 and 1170) as well as in the Ring Access Equipment nodes (1110 and 1120).
The preferred embodiments disclose using a timing delay to ensure that a port progressing from OPER DOWN to operational delays sending data packets long enough for the slave arbiter to impose a virtual break. One of skill in the art will recognize that the use of the timer could be replaced by a control signal sent by the slave arbiter after it has successfully imposed the virtual break.
One of skill in the art will recognize that alternative embodiments set forth above are not universally mutually exclusive and that in some cases alternative embodiments can be created that implement two or more of the variations described above.
Those skilled in the art will recognize that the methods and apparatus of the present invention have many applications and that the present invention is not limited to the specific examples given to promote understanding of the present invention. Moreover, the scope of the present invention covers the range of variations, modifications, and substitutes for the system components described herein, as would be known to those of skill in the art.
The legal limitations of the scope of the claimed invention are set forth in the claims that follow and extend to cover their legal equivalents. Those unfamiliar with the legal tests for equivalency should consult a person registered to practice before the patent authority which granted this patent such as the United States Patent and Trademark Office or its counterpart.
This application claims priority to U.S. Provisional Application 60/490,764 filed Jul. 29, 2003 and U.S. Provisional Application 60/468,325 filed May 6, 2003. This application incorporates by reference these two provisionals.
Number | Name | Date | Kind |
---|---|---|---|
4209666 | Lawton | Jun 1980 | A |
4354267 | Mori et al. | Oct 1982 | A |
4462099 | Braun | Jul 1984 | A |
4527270 | Sweeton | Jul 1985 | A |
4538264 | Bahr et al. | Aug 1985 | A |
4573044 | McConachie et al. | Feb 1986 | A |
4594709 | Yasue | Jun 1986 | A |
4633468 | Skatrud et al. | Dec 1986 | A |
4654857 | Samson et al. | Mar 1987 | A |
4683563 | Rouse et al. | Jul 1987 | A |
4709365 | Beale et al. | Nov 1987 | A |
4800559 | Florea et al. | Jan 1989 | A |
5003531 | Farinholt et al. | Mar 1991 | A |
5113398 | Howes | May 1992 | A |
5136589 | Konishi | Aug 1992 | A |
5218600 | Schenkyr et al. | Jun 1993 | A |
5317198 | Husbands | May 1994 | A |
5410542 | Gerbehy et al. | Apr 1995 | A |
5446725 | Ishiwatari | Aug 1995 | A |
5469503 | Butensky et al. | Nov 1995 | A |
5491696 | Nishimura | Feb 1996 | A |
5495472 | Ohara | Feb 1996 | A |
5532862 | Tada et al. | Jul 1996 | A |
5651000 | Lee et al. | Jul 1997 | A |
5737370 | Hetzel | Apr 1998 | A |
5781530 | Segal | Jul 1998 | A |
5796720 | Yoshida et al. | Aug 1998 | A |
5944798 | McCarty et al. | Aug 1999 | A |
5959990 | Frantz et al. | Sep 1999 | A |
6052210 | Nathan | Apr 2000 | A |
6111853 | Goodman | Aug 2000 | A |
6199174 | Norizuki et al. | Mar 2001 | B1 |
6307652 | Fassih-Nia et al. | Oct 2001 | B1 |
6430151 | Glas et al. | Aug 2002 | B1 |
6581126 | Regula | Jun 2003 | B1 |
6678241 | Gai et al. | Jan 2004 | B1 |
6717909 | Leroux et al. | Apr 2004 | B2 |
6717922 | Hsu et al. | Apr 2004 | B2 |
6766482 | Yip et al. | Jul 2004 | B1 |
6813413 | Bisson et al. | Nov 2004 | B2 |
20030039243 | Parker | Feb 2003 | A1 |
20030099191 | Kim | May 2003 | A1 |
20030223379 | Yang et al. | Dec 2003 | A1 |
20040132409 | Arnold et al. | Jul 2004 | A1 |
Number | Date | Country |
---|---|---|
0052390 | Oct 1981 | EP |
0494695 | Jul 1992 | EP |
0519712 | Dec 1992 | EP |
0403763 | Sep 1994 | EP |
0472380 | Jun 1999 | EP |
07095227 | Apr 1995 | JP |
Number | Date | Country | |
---|---|---|---|
20040223503 A1 | Nov 2004 | US |
Number | Date | Country | |
---|---|---|---|
60490764 | Jul 2003 | US | |
60468325 | May 2003 | US |