Two priority fair distributed round robin protocol for a network having cascaded hubs.

<u>Background</u>

0001The present invention concerns a two priority fair distributed round robin protocol for a network having cascaded hubs.

0002Various topologies may be used to connect network stations together in a local area network. In one topology, a hub may be directly connected to several network stations, called end nodes. Each end node connected to the hub communicates to other end nodes connected to the hub by sending network packets to the hub which the hub forwards to one or more of the other end nodes.

0003When, at one time, more than one end node desires to forward network packets to the hub, some form of arbitration scheme is necessary to assure that no transmissions are lost or garbled since only one packet can be transmitted over the network at any one time. When there is only a single hub in the local area network, the hub can, with limited overhead, enforce a fair and effective arbitration scheme.

0004It is desirable, however, to allow for expansion of a hub network by adding additional hubs. This is done, for example, by cascading the hubs. See, U.S. Patent Number 5,018,139, issued to Rémi Despres for <u>Communication Network Between User Equipment</u> or U.S. Patent Number 4,716,408 issued to Stuart O'Connor, et al. for <u>Data Transmission System of the Star Type with Enablement Passing</u>.

0005However, prior arbitration schemes could not simultaneously provide priorities, cascaded hubs and both fair and effective arbitration. What is meant by a fair arbitration scheme is one which treats requests by all end nodes in an even-handed manner. That is, no end nodes are put at a disadvantage because they are connected to one hub and not to another hub. What is meant by an effective arbitration scheme is one in which network throughput is maximized. That is, when end nodes are contending for network resources, the amount of time the network is idle or in arbitration is minimized.

0006Fair and efficient arbitration schemes for a network with cascaded hubs have generally handled requests within only a single priority level. See, for example, U.S. Patent Number 5,018,139, cited above. When requests with a second priority level have been allowed, this has resulted in arbitration schemes which are less fair and/or less efficient. For example, such two priority arbitration schemes have handled only one network packet per hub per cycle. This results in end nodes connected to hubs with fewer active end nodes having a larger share of network bandwidth than end nodes on hubs with many active end nodes. Additionally, arbitration schemes which handle only one network packet per hub per cycle require more overhead for each network packet sent across the network. This reduces network throughput.

<u>Summary of the Invention</u>

0007In accordance with the preferred embodiment of the present invention, a network system includes end nodes which are connected to a plurality of cascaded hubs. A two priority protocol is provided for selecting end nodes to send network packets over the network. Network packets are, for example, high priority network packets or low priority network packets. When an end node has a high priority packet to send it makes a high priority request to its hub. When an end node has a low priority packet to send it makes a low priority request to its hub. Once an end node makes a high or low priority request, the end node waits for an acknowledgment from the hub before transmitting the packet. When at least one end node in the network system has a high priority network packet to send, each end node with a high priority network packet is acknowledged in turn. Each end node, when acknowledged, sends its high priority network packet through the network.

0008For example, a root hub acknowledges in turn each second level hub which is connected to an end node which has a high priority network packet to send. Each second level hub, when acknowledged, proceeds to acknowledge in turn each end node which is connected to the second level hub and which has a high priority network packet. Each end node, when acknowledged, sends its high priority network packet.

0009When no end nodes have a high priority network packet, and at least one end node in the network system has a normal priority network packet to send, each end node with a normal priority network packet is acknowledged in turn. This includes the root hub acknowledging in turn each second level hub which is connected to an end node which has a normal priority network packet to send. Each second level hub, when acknowledged, proceeds to acknowledge in turn each end node which is connected to the second level hub and which has a normal priority network packet. Each end node, when acknowledged, sends its normal priority network packet. However, when the root hub receives a request for an end node to send a high priority packet, the root hub signals to the second level hub that a high priority request has been received. The second level hub stops acknowledging the end nodes connected to the second level hub which have a normal priority network packet. Then, each end node with a high priority network packet is acknowledged in turn. Each end node, when acknowledged, sends its high priority network packet.

0010In the preferred embodiment, during transmission of network packets between each second level hub and the root hub, a link between the second level hub and the root hub acts as a half duplex link and during the interpacket gap between transmission of network packets, the link between the second level hub and the root hub acts as a dual simplex link. Thus signaling between the root hub and the second level hubs is done during interpacket gaps. For example, each link is formed by a plurality of twisted wire pairs. During transmission of network packets between each second level hub and the root hub, the plurality of twisted wire pairs are used as a half duplex link. During the interpacket gap between transmission of network packets, half the twisted wire pairs are used to forward signals from the second level hub to the root hub and half the twisted wire pairs are used to forward signals from the root hub to the second level hub.

0011During the interpacket gaps, the second level hub and the root hub signal to each other using tones which are transmitted at a transmission frequency which is lower than a transmission frequency used to transmit the network packets. For example, in the preferred embodiment of the present invention, the root hub acknowledges the second level hub by silence during the interpacket gap between transmission of network packets. When a second level hub is acknowledging end nodes with normal priority network packets, during an interpacket gap the root hub signals the second level hub that there are end nodes with high priority packets by forwarding a combination of tones which signify a preempt command.

0012The protocol described herein provides fair and efficient control of the transfer of network packets for a two priority system operating on a network architecture which includes cascaded hubs.

<u>Brief Description of the Drawings</u>

0013Figure 1 shows end nodes connected to a single hub in a network configuration.

0014Figure 2 illustrates the basic operation of a protocol for the network configuration shown in Figure 1.

0015Figure 3 illustrates the further operation of a protocol for the network configuration shown in Figure 1.

0016Figure 4 illustrates in more detail operation of the protocol shown in Figure 3.

0017Figure 5 shows end nodes connected to a plurality of cascaded hubs in accordance with a preferred embodiment of the present invention.

0018Figure 6 illustrates the operation of a protocol for the network configuration shown in Figure 5 in accordance with a preferred embodiment of the present invention.

0019Figure 7 further illustrates the operation of a protocol for the network configuration shown in Figure 5 in accordance with a preferred embodiment of the present invention.

0020Figure 8 additionally illustrates the operation of a protocol for the network configuration shown in Figure 5 in accordance with a preferred embodiment of the present invention.

<u>Description of the Preferred Embodiment</u>

0021Figure 1 shows a hub 10 connected to an end node 11, an end node 12 and an end node 13. In the embodiment shown in Figure 1, each of end nodes 11, 12, and 13 is connected to hub 10 using four twisted wire pairs. For more information on connection of a hub to end nodes, see co-pending patent serial number 07/972,694 filed on November 6, 1992 by Alan Albrecht, Steven Harold Goody, Michael Peter Spratt, Joseph Anthony Curcio, Jr. and Daniel Joseph Dove for HIGH SPEED DATA TRANSFER OVER TWISTED PAIR CABLING, which is hereby incorporated by reference.

0022The protocol for sending messages is summarized as follows. Each four twisted wire pair link between hub 10 and each end node has three stable states.

0023In the first stable state, hub 10 is sending a packet to the end node on all four twisted wire pairs. In the preferred embodiment, in this transmission there is a total bandwidth of 100 megabits per second (Mb/s). In the second stable state, the end node is sending a packet on all four twisted pairs. In the preferred embodiment, in this transmission there is also a total bandwidth of 100 Mb/s. In the third stable state, the end node is sending low frequency tones on two twisted wire pairs (twisted wire pair one and twisted wire pair two) and the hub is sending low frequency tones on the other two twisted wire pairs (twisted wire pair three and twisted wire pair four).

0024There are two possible tone frequencies on each twisted wire pair. Thus, the end node can transmit four possible tone combinations on twisted wire pair one and twisted wire pair two. For example a first possible tone combination is to send a one megahertz (MHz) tone on twisted wire pair one and a one MHz tone on twisted wire pair two. A second possible tone combination is to send a one MHz tone on twisted wire pair one and a two MHz tone on twisted wire pair two. A third possible tone combination is to send a two MHz tone on twisted wire pair one and a one MHz tone on twisted wire pair two. A fourth possible tone combination is to send a two MHz tone on twisted wire pair one and a two MHz tone on twisted wire pair two.

0025Similarly, the Hub can transmit four possible tone combinations on twisted wire pair three and twisted wire pair four. A first possible tone combination is to send a one megahertz (MHz) tone on twisted wire pair three and a one MHz tone on twisted wire pair four. A second possible tone combination is to send a one MHz tone on twisted wire pair three and a two MHz tone on twisted wire pair four. A third possible tone combination is to send a two MHz tone on twisted wire pair three and a one MHz tone on twisted wire pair four. A fourth possible tone combination is to send a two MHz tone on twisted wire pair three and a two MHz tone on twisted wire pair four.

0026The tone combinations are selected of a low enough frequency to avoid contributing in any significant way to crosstalk. These tone combinations are used to convey certain well defined information. For example, in the preferred embodiment, tone combinations from end nodes 11, 12 or 13 to hub 10 indicate one of the following: "Idle", "ReqH" (high priority request) or "ReqN" (normal priority request). The fourth tone combination is reserved for future use. The tone combinations from hub 10 to end nodes 11, 12 or 13 indicate one of the following: "Idle", "Incoming" or "Preempt" message. The remaining tone combination is reserved for future use. In the preferred embodiment the message signal is used as described below. In an alternate embodiment, the idle message can be used to where preempt is used.

0027Additionally, a special tone combination from hub 10 to end nodes 10, 11, 12 is "Ready to Receive". This special tone combination is signaled by silence (no tones) on twisted wire pair three and twisted wire pair four. The Ready to Receive functions as an acknowledge (Ack) by hub 10 to a request received from one of end nodes 10, 11 or 12.

0028Figure 2 illustrates the operation of the protocol for use in the transfer of a network packet. A first end node, represented by a line 21, sends to hub 10, represented by a line 20, a request, as represented by an arrow 31. The request is either a ReqH or ReqN depending on the priority of the network packet. After receiving the request, hub 10 sends an acknowledge in response, as represented by an arrow 32. As discussed above, this is silence on twisted wire pair three and twisted wire pair four. Thus, the end node actively transmits the request (ReqH or ReqN) on twisted wire pair one and twisted wire pair two. Upon detecting the acknowledge on twisted wire pair three and twisted wire pair four, the first end node starts sending the network packet to hub 10, as represented by a start of packet (SOP) arrow 33.

0029Hub 10 starts receiving the network packet and temporarily stores the first part of the network packet until the destination address is completely received by hub 10. At this time, the destination port to which the packet is to be sent can be quickly ascertained. Upon receiving the start of the network packet, hub 10 sends the Incoming message tone combination to all other end nodes, as represented by an arrow 34. This causes the other end nodes to stop sending request (ReqH or ReqN) or Idle on their twisted wire pair one and twisted wire pair two. Thus, when hub 10 starts sending the network packet on all four twisted wire pairs to a destination end node, as represented by arrow 35, the destination end node is not actively transmitting on any twisted wire pair. The destination end node is represented by a line 22. An arrow 36 represents the end of packet (EOP) sent from the first end node to hub 10. An arrow 37 represents the end of packet (EOP) sent from hub 10 to the destination end node.

0030By allowing an end node to originate back-to-back network packets, network throughput may be increased. This is illustrated by Figure 3. In Figure 3, after the first end node has completed sending the first network packet, as described in the discussion of Figure 2 above, the first end node sends to hub 10 another request, as represented by an arrow 41. The request is either a ReqH or ReqN depending on the priority of the second network packet. Immediately upon sending the end of the network packet to the destination end node, as represented by arrow 37, hub 10 sends another acknowledgment to the first node, as represented by an arrow 42. (In the preferred embodiment, when hubs are cascaded, the second message is not immediately sent, but rather the end node must wait for its next turn to send the second message). As discussed hub 10 acknowledges the first node by silence on twisted wire pair three and twisted wire pair four. Upon detecting the acknowledge on twisted wire pair three and twisted wire pair four, the first end node starts sending the second network packet to hub 10, as represented by a start of packet (SOP) arrow 43.

0031Hub 10 starts receiving the second network packet and temporarily stores the first part of the second network packet until the destination address is completely received by hub 10. At this time, the destination port to which the second network packet is to be sent can be quickly ascertained. In the example illustrated by Figure 3, the second network packet is sent to the same destination; however, the destination may be any end node, other than the first end node, connected to hub 10.

0032Upon receiving the start of the second network packet, hub 10 sends the Incoming message tone combination to all other end nodes, as represented by an arrow 44. This causes the other end nodes to stop sending request (ReqH or ReqN) or Idle on their twisted wire pair one and twisted wire pair two. Thus, when hub 10 starts sending the second network packet on all four twisted wire pairs to a destination end node, as represented by arrow 45, the destination end node is not actively transmitting on any twisted wire pair. An arrow 46 represents the end of the second network packet sent from the first end node to hub 10. An arrow 47 represents the end of the second network packet sent from hub 10 to the destination end node.

0033Figure 4 shows additional detail of network operation in the interpacket gap between the end of the first network packet and the start of the second network packet. The interpacket gap has two parallel functions. First, in order to determine which end node the second network packet should be sent to, hub 10 uses the interpacket gap to temporarily store the start of the second network packet, and look up the destination address for the second network packet.

0034Second, the interpacket gap allows the destination end node to send a request to hub 10, before the request is cut off by the Incoming signal from hub 10. This is represented by arrows 52. If the destination end node does not desire to send a network packet, the destination node forwards an Idle signal to hub 10. End nodes which do not receive the first network packet also can send a request to hub 10 during the interpacket gap. However, in addition, the end nodes which do not receive the first network packet can also send a request to hub 10 at the same time the first network packet is being sent to the destination end node.

0035The request window extends from the time hub 10 transmits the end of the first packet, represented by arrow 37, until time hub 10 begins sending the incoming signal, represented by arrow 44. During the request window, hub 10 continuously sends the idle signal, as represented by arrows 53 to the end nodes. At the completion of the request window, hub 54 sends the incoming signal represented by arrow 44 and arrows 54, to all the end nodes except the end node which will transmit the network packet. Each end node, upon the receiving the incoming signal from hub 10 stops sending the idle signal or the request signal in order to be ready to receive a network packet from hub 10.

0036The request window needs only to be long enough to allow the destination end Node to send out a request of sufficient duration for hub 10 to recognize the request. The request window duration is therefore largely determined by the time for the hub to acquire the request, and is independent of the propagation time required to send a message from hub 10 to an end node. For example, in the preferred embodiment, the request window has a duration of approximately 850 ns. The period during which hub 10 sends the incoming signal to the end nodes, called the line turnaround period, is determined largely by the maximum round-trip propagation delay between hub 10 and the end node, plus the time to receive enough of the incoming packet to determine the destination address. The round-trip propagation delay from hub 10 to an end node, in the preferred embodiment, is approximately 1.8 microsecond for a cable distance of 180 meters.

0037The request window and the line turnaround period are only important for embodiments of the present invention which utilize half-duplex links. In the case of fiber optic links, for example, between cascaded hubs, the connection will be dual simplex, and so there is no requirement to allow additional time for a second level hub to pass a request to a root hub. This is because while the root hub is sending a packet to the second level hub, the second level hub can pass requests back to the root hub.

0038In an alternate embodiment, hub 10 sends the "Incoming" signal at the same time the "Acknowledgment" signal is sent. In this case, any end node with a packet sends a "Request" immediately at the end of a last packet, even though it is receiving an "Incoming" signal. The end node continues sending the "Request" signal only for short time, and then removes the "Request" in preparation to receiving a packet.

0039Figure 5 shows an example of a network system with cascaded hubs according to the preferred embodiment on the present invention. In Figure 5, a root hub 60 is connected to an end node 71, an end node 72, an end node 81 and an end node 82. Root hub 60 is also connected to a second level hub 61, a second level hub 62 and a second level hub 63. Second level hub 61 is connected to an end node 73, an end node 74 and an end node 75. Second level hub 62 is connected to an end node 76 and an end node 77. Second level hub 63 is connected to an end node 78, an end node 79 and an end node 80.

0040In the preferred embodiment, the interconnections are with non-bundled four-pair Category 3, or better, unshielded twisted pair cable, two-pair shielded twisted pair cable or optical fiber. A second level hub will communicate on half-duplex links with root hub 60, using the protocol illustrated by Figure 3 and Figure 4. Alternately, dual simplex fiber optic links may be used when longer distances separate root hub 60 from the second level hubs. In the preferred embodiment, the interhub connections are not through bundled cable.

0041The protocol used to transfer network packets from end nodes 71, 72, 81 and 82 is the same as in the single hub case described above. The protocol used to transfer network packets from end nodes 73, 74 and 75 to second level hub 61, to transfer network packets from end nodes 76 and 77 to second level hub 62, and to transfer network packets from end nodes 78, 79 and 80 to second level hub 63 is also the same as in the single hub case described above. The protocol used to transfer network packets from second level hubs 61, 62 and 63 to root hub 60 is also the same as in the single hub case described above, with slight modifications, as described below.

0042Addressing for the network shown in Figure 5 is summarized as follows. All network packets are seen by all hubs. Each hub, however, employs an address filter so that unicast network packets are only sent on to an end node if the destination address of the network packet matches the address of the end node. If a bridge or network monitor is attached to the network, all packets are also forwarded to the bridge or network monitor.

0043In the preferred embodiment of the present invention, the cascaded hubs maintain an enhanced two priority round robin arbitration. In the two priority round robin arbitration, end nodes send a request at either high priority or normal priority. Once requests have been received, proceeding to each in turn, each end node with a high priority network packet is allowed to send the network packet. This continues until all high priority network packets have been sent. Then, while there are no requests to send high priority network packets, proceeding it each in turn, each end node with a normal priority network packet is allowed to send one network packet. In the preferred embodiment, only one network packet is sent, however, it is contemplated that in alternate embodiments of the present invention, more than one network packet may be sent.

0044During operation, all network packets are sent to root hub 60 which relays them onto the attached second level hubs. Request and other tone combinations can be sent between end nodes and the root hub, between end nodes and the second level hub, and between the second level hub and root hub 60 during interpacket gaps. The interpacket gaps are forced by the protocol to be at least a minimum time duration sufficient to allow requests and tone combinations to be sent and received.

0045For example, during a request window of an interpacket gap, the end nodes attached to each second level hub will send to the second level hub request tone combinations. Each second level hub which receives a request, forwards the request to root hub 60. The request forwarded by a second level hub to root hub 60 reflects the highest priority level of the requests the second level hub received from its end nodes. If a second level hub receives one or more high priority requests from one or more end nodes, the second level hub sends a high priority request to root hub 60. If the second level hub receives no high level requests from its end nodes, but receives one or more normal priority requests from one or more of its end nodes, the second level host will forward to root hub 60 a normal priority request. If the second level hub is not receiving any requests from attached end nodes, then the second level hub sends an idle during the request window of the interpacket gap.

0046When root hubs 60 sends an acknowledge (i.e., silence) to a second level hub, the second level hub becomes the acknowledged second level hub. If the acknowledged second level hub has been sending high priority requests to root hub 60, then the second level hub sends an acknowledge to one of its end nodes which has made a high priority request. The end node receiving the acknowledge will send a network packet which the second level hub will send to root hub and to any end node connected to the second level hub to whom the network packet is addressed. When the acknowledged end node has completed sending the network packet, the second level hub will continue a sweep of its end nodes until all its end nodes that have made a high priority request have had a single opportunity to send a high priority network packet.

0047During the sweep, while there are still end nodes attached to the second level hub which have not yet had their opportunity to send their high priority network packets, after a network packet is forwarded by the second level hub to root hub 60, the second level hub forwards to the root hub a high level request. The second level hub also sends an acknowledge to the next end node to send a high priority network packet, without waiting for an acknowledge from root hub 60.

0048Once the second level hub has completed a single sweep of its end nodes, the second level hub will return an idle to root hub 60. On receiving the idle, root hub 60 recognizes that the second level hub has finished its part of the high priority round robin sequence. Root hub 60 then continues by sending an acknowledge to the next (in turn) second level hub or end node directly coupled to the root hub that has made a high priority request. This will proceed until all high priority network packets have been sent. Then, while there are no requests to send a high priority network packets, proceeding in turn, each end node with a normal priority network packet is allowed to send the network packet

0049Operation of requests and acknowledges in the network shown in Figure 5 is further illustrated by way of example. In the example, end node 71, end node 73, end node 75, end node 77 and end node 81 each request to send a network packet with high priority. Likewise, end node 72, end node 74, end node 76, end node 78, end node 79, end node 80 and end node 81 each request to send a network packet with normal priority.

0050Thus, for the example set out above, presuming root hub 60 is at the start of its cycle, the order in which the high priority requests will be serviced is end node 71, end node 73, end node 75, end node 77 and end node 81. Assuming that there are no more high priority packets queued at the end nodes, and therefore no more high priority requests, the normal priority requests will be serviced in the following order: end node 72, end node 74, end node 76, end node 78, end node 79, end node 80 and end node 81.

0051In this example, root hub 60 will first grant an acknowledge to end node 71. After end node 71 has sent its high priority network packet, end node 71 will send an idle signal to root hub 60. Root hub 60 will then send an acknowledge to second level hub 61. Second level hub 61 will then send an acknowledge to end node 73.

0052Figure 6 illustrates what happens next. Line 93 represents end node 73. End node 73 forwards a network packet 98 to second level hub 61. Second level hub 61 is represented by a line 91. Second level hub 61 forwards network packet 98 to root hub 60. Root hub 60 is represented by a line 90. Root hub 60 will forward network packet 98 to second level hub 62 and second level hub 63. Also the end node on the network which is the destination for network packet 98 will receive network packet 98. In Figure 6, the end node which is addressed by the network packet is represented by line 94 and the second level hub to which the destination hub is addressed is represented by line 92.

0053After second level hub 61 has completed forwarding the message to root hub 60, second level hub will send an acknowledge 96 to end node 75. End node 75 is represented by a line 95. Second level hub 61 will also send a high level request 97 to root hub 60 indicating another high level priority network packet will follow. End node 75 will send a network packet 99 to second level hub 61. Second level hub 61 forwards network packet 99 to root hub 60. Root hub 60 will forward network packet 99 to second level hub 62 and second level hub 63. Also the end node on the network which is the destination for network packet 99 will receive network packet 99. The end node which is addressed by the network packet is represented by line 94 and the second level hub to which the destination hub is addressed is represented by line 92.

0054As illustrated by Figure 7, after second level hub 61 forwards network packet 99 to root hub 60, second level hub 61 will forward an idle signal 101, indicating that second level hub 61 has finished the sweep of all its end nodes which had a high priority network packet to send. Upon receiving idle signal 101, root hub 60 will send an acknowledge signal 102 to second level hub 62. In Figure 7, second level hub 62 is represented by a line 105. Second level hub 62 will send an acknowledge 103 to end node 77. End node 77 is represented by a line 104. End node 77 will then send its high priority network packet 100 to second level hub 62. Second level hub 62 forwards network packet 100 to root hub 60. Root hub 60 will forward network packet 100 to second level hub 61 and second level hub 63. Also the end node on the network which is the destination for network packet 100 will receive network packet 100.

0055After all high priority network packets have been sent, the normal priority network packets will be serviced. In the example, root hub 60 will first grant an acknowledge to end node 72. After end node 72 has sent its normal priority network packet, end node 72 will send an idle signal to root hub 60. Root hub 60 will then send an acknowledge to second level hub 61. Second level hub 61 will then send an acknowledge to end node 74. After end node 74 has sent its normal priority network packet, end node 74 will send an idle signal to second level hub 61. Second level hub 61 will send an idle signal to root hub 60. Root hub 60 will then send an acknowledge to second level hub 62.

0056If in middle of a second level hub serving normal priority packets, root hub 62 becomes aware of an end node sending high priority request (ReqH) or of a second level hub sending a high priority request (which would have originated from one of the end nodes of the second level hub), root hub 60 needs to ensure that the high priority request is serviced quickly. It does this by sending a burst of Preempt during the request window of the inter packet gap to the second level hub. This is different than the silence root hub 60 normally sends to an acknowledged second level hub during the interpacket gap. In the preferred embodiment, the "Preempt" signal is used, in an alternate embodiment, the "Idle" signal may be used instead of the "Preempt" signal. The burst of Preempt forces the second level hub to send control back to the root hub after the current packet has been sent.

0057For example, Figure 8 illustrates this. In the example discussed above, once second level hub 63, represented by a line 121, is acknowledged for normal priority packets, second level hub 63 acknowledges end node 78. End node 78 forwards a normal priority network packet 111 to second level hub 63. Second level hub 63 forwards network packet 111 to root hub 60. Root hub 60 is represented by line 90. Root hub 60 will forward network packet 111 to second level hub 61 and second level hub 62. Also the end node on the network which is the destination for network packet 111 will receive network packet 111. In Figure 6, the end node which is addressed by the network packet is represented by line 124 and the second level hub to which the destination hub is addressed is represented by line 123.

0058After second level hub 63 has completed forwarding network packet 111 to root hub 60, second level hub 63 will send an acknowledge 114 to end node 79. End node 79 is represented by a line 125. Second level hub 63 will also send a normal level request 113 to root hub 60 indicating another normal level priority network packet will follow. If root hub 60 is aware of an end node which desires to send a high priority packet, root hub 60 will send a burst preempt signal 115 to second level hub 63 at the completion of receipt of network packet 111.

0059Because end node 79 has already been acknowledged, end node 79 will send a network packet 112 to second level hub 63. Second level hub 63 forwards network packet 112 to root hub 60. Root hub 60 will forward network packet 112 to second level hub 61 and second level hub 62. Also the end node on the network which is the destination for network packet 112 will receive network packet 112. The end node which is addressed by the network packet is represented by line 124 and the second level hub to which the destination hub is addressed is represented by line 123.

0060After second level hub 63 has forwarded network packet 112 to root hub 60, second level hub 63 sends an idle signal 116 to root hub 60. Even though second level hub 63 has not serviced the normal priority network packet from end node 80, second level hub 63 gives idle signal 116 to root hub 60 in response to preempt 115, sent in the previous inter packet gap. Receipt of idle signal 116 allows root hub 60 to give an acknowledge for the high priority packet.

0061The foregoing discussion discloses and describes merely exemplary methods and embodiments of the present invention. As will be understood by those familiar with the art, the invention may be embodied in other specific forms.

0062Accordingly, the disclosure of the present invention is intended to be illustrative, but not limiting, of the scope of the invention, which is set forth in the following claims.