CA2079340A1

CA2079340A1 - Packet switching communications system

Info

Publication number: CA2079340A1
Application number: CA002079340A
Authority: CA
Inventors: Pavel Cerna; Robert F. Hay
Original assignee: Individual
Current assignee: Republic Telcom Systems Corp
Priority date: 1991-02-01
Filing date: 1992-01-29
Publication date: 1992-08-02
Also published as: JPH06501365A; US5444707A; BR9204115A; AU1440592A; WO1992014320A1; EP0530345A1; AU652814B2; KR100249112B1; EP0530345A4; KR930701039A

Abstract

A multi-channel telephonic communications system, (PBXA-PBXD), where voice, data, facsimile information is packetized and switched to the desired destination based upon a call number embedded in each packet which is based upon the digits dialed at the source.
Aswitching architecture (44) is used to minimize the trunk line resource needed and to simplify system configuration and maintenance. Connection between the source and the desired destination are allowed if there is sufficient bandwidth to support the connection. Flow control is also provided, which dynamically varies the bandwidth provided. Flow control is also provided, which dynamically varies the bandwidth used by voice packets to adjust to changing traffic levels. Voice bandwidth is reduced when needed by the traffic volume, and voice bandwidth is increased when possible to improve voice quality.

Description

Wo 92/14320 PCI/US92/00718 PACKET SWITCHING COMMUNICATIONS SYSTEM

TECHNICAL FIELD

The present invention pertains generally to the field of multi-channel telephonic communications. In particular, it pertains to a communications system which switches packetized information fro~n a 10 source node to a desired destination node based upon the dig~ts dialed, which have been embedded in each of the packets in the forrn-of a unique call number.
.
BACKGROUND ART
; :
Telephonic cornmunication between remotely located offices of a single organization is traditionally accomplished through use of public phone lines or by leasing private lines. For occasional co~ununication traffic, public telephone lines are quite adequate. As the Yolume of traffic increases, 20 it may become economically advantageous to lease private lines. While reducing the cost compared to the public lines, the leasing of private lines remains quite expensive. In an a~tempt to further reduce the cost, techniques have been developed for multiplexing rnultiple channels of communication across a reduced nurnber of communication lines. These techniques are well 25 known in the art, and include synchronous tim~division multiplexing (TDM) as well as information packetizing.
Multiplexers performing these techni~ques are placed at opposite ends of a communication path, increasing the amount of commun~cation SUBYITUTE S~EFT

wo 92/14320 ~ , c~ 2 - pcr/us92/oo718 traffic a particular line can handle, thus reducing the cost per communication event. A large organization may have several offices scattered nationally or even globally. A private communication system meeting the needs of such an organ~ation would typically provide one or more multiplexers at each 5 location, leased trunk lines to interconnect the locations to each other, ard a local PBX system at each location to interface internal telephones to the public phone system and the leased trunk lines.
The difficulty with this approach is t~at multiplexers are point to point devices, in that they are intended to be connected at the tw~ nodes (or 10 end points) of a single communication path. Interconnection of more than two nodes requires trunk lines to interconnect every pair of nodes, ancl that the multiplexers at each node be capable of handling multiple trunk lines. As the co~nplexity of this type of system increases through the addition of nodes, the number of trunlc lines required for interconnection of all pa*s of nodes 15 increases exponentially, with a corresponding exponential increase in cost.
At some pomt, the complexity of the system calTies with it a cost which is prohibitive, and in an attempt to reduce cost, some of the contemplated trunk lines are eliminated. Communications between a pair of nodes not directly connected by a trunk line are accomplished by routing 20 through an alternate node which is connected to both nodes in the pair between which communication is to take place. The PBX at the alternate node receives the traffic from the originating node, and routes it back through its multiplexer to another trunk line and on to the destina~on node.
While roubng in this manner reduces the number of 25 interconnecting trunk lines required and the associated eos~, other problems .
.

. .

. ~ , : . ' . . . . . . . ' ' . . . ~ .. . . : .

o 92/14320 Pcr/uss2/oo7 and limitations are introduced. For example, if the multiplexers use a digital packetizing scheme including speech compression, then routing th~ough multiple PBX systems would likely invokre tandem encoding. Tandem encoding occurs when an analog signal is packetized and compressed to 5 digital at the source node, decompressed and restored to analog at the intermediate node's PBX, re-packetized and re-compressed to digital, and decompressed and restored again to analog at the destination PBX. Each generation of compression and decompression introduces distortion into the resulting analog signal, and more than two generations is likely to result in lû distortion sufficient to render the co~s~nunication unintelligible. This means that ~e system must provide a number of t~ lines sufficient to allow every node to communicate with every other node with no more than one inteImediate routing through a PBX. This requirement lirruts the ability of the organization to reduce cost through the elimination of trunk lines.
In addition, trafflc routing requires that each node have sufficient trunk resource not only for its own communication, but also for that of the traffic it must pass on to other nodes. Determining the amount of ;;
trunk line resource necessary for any given node thus becomes rather complex process, in that it depends upon traffic ~at the node will have no 20 direct involvement with. Further complicating the problem is the fact that , each node has multiple trunk lines, each corresponding to a given remote node, and that the multiplexer to trunk line cormections are "nailed up", .
meaning that the PBX and multiplexer use the trunk line corresponding to the desired destination node. The result of ths architec~ure is that ~he traffic ':
':' `' wo 92/14320 P~r/Us92/00718

2 ~ ~ - 4 -on each trunk line depends upon the node to node routing paths, as well as the volume of traffic being handled by that nc~e.
The previous discussion presumes the use of multilplexers with "nailed up" PBX connections and no direct connections for passing thrs ugh-5 tra~fic. However, even if the multiplexers being used support networking (i.e., pass the through-traffic without PBX tandem encoding), the connections through which ~e through-traffic is routed are sta~c and at best can be altered based upon the tine of day or other statistical parameters. This tends to mitigate the problems associated with "nailed up" connections, but does not 10 eliminate them. The amount of trunk line resource needed at a node still depends upon trafflc the node is not involved in and upon the routing paths being used.
These problems associated with communication routing systems of the prior art are exacerbated as the comple~aty of the system increases. The 1~ addition of more nodes either increases the number of trunk lines required or makes routing more complicated, or both. The more complex the routing patterns, the more dependant a node is on o~er node's resources, and the more difficult it is to determine the amount of resource needed for any given trun~ line at each node. Increased complexity also increases the 20 administrative burden involved in managing the line resources and the equlpment involved.
It is clear that there is a need ~or a communicatlon system which makes optimum use of trunk line resource, while ha~Lng an arch~tecture ~at minimizes trunk line expense and that allows for expansion wit~out undue 25 complexity or expense.
.:

WO 92/14320 ~ ? PCI/US92/00718 DISCLOSURE OF THE INVF.NTION

The present invention is directed toward a packet sw~tching communications system capable of conveying voice, facsimile (FAX) and data 5 between nodes. The system operates by packetizing amd compressing the information (voice, FAX, or data) and embedding a header within each packet containing a unique call number based upon the digits dialed at the source location. By including this routing information (the call number) with the voice, FAX or computer data being sent, a packe'cizing and compressing 10 system can use a st~tching architecture rather than a rou~ng architecture, thus making optimal use of the digital circuits required to interconnect the various nodes. This architecture eliminates the " nailed up" and static connection problem, in effect being able ~o alter the connections on a call-by-call basis.
Operation begins when a telephone number is dialed at a source node, and the RLX at the source node sends a request to the packet switc~ing unit for a connection. The packet switching unit reads the digits dialed at the ~ :
source node and determines whe~er there is sufficient bandwidth for communication over the requested connection to occur. If there is sufficient 20 bandwidth, then the packet switching unit requests a connection to the desired destination node. If accepted, ~e ~ the connection is complete and communication between the nodes can begin. ~:
The packet switching unit consists of a master board and one or more slave boards, connected together over a cornmon internal bus. Each ':';

~ . .

o 92/14320 PCr/l IS92/0071 slave board connects to as many as four RLX devices, each located at one of the various nodes of the system.
Packeti~ed data arrives from one RLX to a slave board, the header is read to determine its destination, and the packet is switched to the 5 desired destination RLX as indicated by the call number. The packet is then switched onto a trunk line associated with the desired destinatiors node.

Accordingly, it is an object of the present invention to provide a communication system which makes optimal use of digital trunk lines which 10 interconnect the various commurlication nodes of the system.
Another object of the present invention is to provide a communication system which uses packetized information, and switches packets from the source node to the desired destination node.
These and other objectives of the present invention will become 15 apparent with referenoe to the drawings, the detailed description of the preferred embodiment, and the appended claims.

BRIEF pESCRIPTION OP DRAW~GS

Figure 1 shows a m ultiple node communications system architecture according to ~e prior art.
~igure 2 shows a m~ltiple node packet switching communications system architecture according to the present invention.
Figure 3 shows a block diagram of the packet switching unit of 25 Figure 2 and its connections to multiple nodes.

wo 9~/1 4320~ 3 Pcr/ US92/007 1 8 Figure 4 shows a block diagram of the Master module of the packet switching unit of Figure 3.
Figure 5 shows a block diagram of a Slave module of the packet switching unit of Figure 3.
5Figure 6 shows a flow chart of the packet switch communications connection protocol of the presenl: in~ren~on.
Figure 7 shows a flow chart of packet flow control of the present invention. :

10BEST MODE FOR CARRYING OUT THE ~VENTION
.'~ '.

The assignee of the present invention is also the assignee of U.S.
Patent No. 4,782,485, which is addressed to digital packetizing techniques, and the disclosure of which is hereby incorporated by reference.
1~ :
Referring to Figure 1, a block diagram of a prior art telecommunication system 10 is shown. Each location or node within the system has a PBX 12 and a Tl multiplexer 14 for interfacing to Tl trunk lines, and which connects to a set of trunk lines 16 interconnecting the various 20 locations. For example, trunk line 16a connects multiplexers 14a and 14c, trunk line 16b connects multiplexers 14a and 14d, etc. To fully interconnect all four locations, six trunk lines 16a-16f are required. In the shown example, f~y interconnecting four locations requires six trunk lines. As ~e number of locations to be interconnected increases, the required number oF trunk W0 92/14320 p~ PCr/US92/00718 lines increases exponentially. In order to hllly interconnect all locations, the required number of trunk lines is:

~ n~- n)/2 where 'n' is the number of locations to be interconnected.
Each multiplexer 14 may be a single device capable of handling all the ~unk lines 16 connected to the node in question, or may be multiple devices, each handling one or more of the trunk lines 16. Either way, there 10 must be sufficient multiplexing equipment 14 at eac h location to interface to each of the trunk lines 16 connected to that location. The same holds ~ue for the PBX equipment at each location. Since the multiplexer connections are "nailed up", the PBX must be capable of connecting to as many trunks as are connected to the location. I~e costs associated with the PBX ;md 15 multiplexing equipment at each location, and the correspcnding maintenance costs are function of the number of trunk lines. If the communications system to be installed includes many locations or as ~e size of an existing system grows, the cost of the tl unk lines and the associated equipment costs increase exponentially as the n~nber of ~unk lines 20 increases.
One way of limiting the costs associated with system size, is to reduce the number of ~ lines. For example, if it is dete~nined that the .
;~ ~ amount of ~raffic be~ween PBX A 12a and PBX B 12b is small, the trunk line connecting the two locations 16e may be Pl ated. This would also allow 25 ~r reduced mul~plexing resource 14a, 14b and PBX resource 12a, 12b at both .

wo 92/14320 PCI/US92/00718 g ~
locations. Communication between PBX A 12a and PBX B 12b would need to be routed through ai. alternate node connected to both. For example, à call fron~ PBX A 12a would be sent through multiplexer 14a on the channel connected to trunk line 16a, and received by PBX C 12c. PBX C 12~ would 5 detect that the call was intended for PBX B 12b, and route the communication back out ~rough multiplexer 14c through a PBX connection 18. I~e multiplexer charmel chosen is one connected to trunk. line 16d, which connects to ~e multiplexer 14b and PBX 12b at location B. This req~Lires tandem encoding if the multiplexers 14 do not directly pass through-traffic, 10 and in any event would cause the resource problems previously described d-1e to the connections being static.
Referring now to Figure 2, a conununication system 40 according to the preferred embodiment of the present invention is shown. Each location has a standard Pl~X 12 for managing the ;nternal corrununication 15 system. Interface to external trunk lines is ac~omplished by a multiplexed digital packetizirLg unit (RLX) 42 located at each location. Ea :h RLX is connected to a trunlc line 16g-16j to a packet switch system 44.
In operatlon, communication originating at PE~X A 12a, for example, is packeti zed by RLX 42a in a manner known in the art (see U.S.
20 Patent No. 4,782,485). Briefly, digital packetization of a voice signal involves digital sampling of the analog signal, compressing ~e data through use of speech compression algorit~uns, and organizing ~e data into a se~es of groups or packets. Facsimile (FAX) and computer data can also be packetized, and compressed using known data compression algD~ithms. Since all types of 25 co~ununication is packetized, the system can easily n~ix voice, FAX, and , : ~ :, ~ . . . . ~ :
., , ~ ;. . ~ . :
::. ~ ... . . .

w(~ 92/14320,~ ~ _ PCr/US92/00718 computer data. Within each packet is a header which contains information identifying the type of data contained w;thin the packet. Of importance to the present invention is that each RLX 42 embeds wiWn each packet a call number, uniquely derived as a function of the digits dialed by the originator 5 of the communication.
After a group of data is packetized, the packet is sent to the packet switch 44 over a trur~ line 16. The packet switdl 44 reads from each packet the call number, and switches the packet onto a tr-mk line 16 according to the desired destination as indicated by the call num~er. The first advantage 10 provided by this system is the fact that no more than one trunk line is required per location, since each location need only connect to a single switch 44. l~us, increases in systen complexity do not result in exponential increases in trunk line and equipment costs, as is true of prior art systems.
There are other significant advantages to the system of Figure 2 15 over the prior art. In addition to limiting the impact of additional nodes on ~e number of additional trunk lines, the number of trunk lines needed for : -any given number of nodes is also reduced. There is also a corresponding reduction in the number of circuits needed in the PBX and multiplexers present at each node. Ihese advantages also lead to a reduced system 20 administration burden, and make planning easier since the amount of trunk line resource needed depends upon the annount of co~ununication traffic for each node, and is no longer dependant upon the traffic patterns between nodes.
Those skilled in the art will recognize that rnany modifications 25 may be made to system sho~m on Figure 2 without departing from the scope :;:

WO 92/14320 ~ PCI/US92/00718 of the present invention. For example, the packet switch 44 would preferably be located at one of the system nodes, allowing the packet switch 44 and the RLX 20 at that location to be connected with local lines, obviating ~e need for the leased trunk line. In the example of Figure 2, if the packet switch 44 was S located at location A, then the connection between RLX A 20a and the packet switch 44 represented by 16g, could be inexpensive local wiring rather than a more expensive leased line.
In addition, those skilled in the art will recogruze that the trunk lines are contemplated to be leased Tl lines, but could be any other lû communication medium such as El lines, a satellite link, modems connected to analog lines, leased or switchecl digital lines, or Integrated Services Dig~tal Network ~ISDN) lines. Further, these trunk lines can be leased in whole or in part, depending upon the amount of usage desired. I~us, references to a ~r~mk line in the present disdosure refer equaLJ to a portion of a h~unk line 15 or multiple trunk lines, as required by the specific application and the expected amount of traffic.
Referring now to Figure 3, a blo~ diagram of a packet switch 44 is shown according to the pre~erred embodiment of the present invention.
One or more slave boards 60 are connected to RLX packe~zing multiplexers 2û
20 over trunk lines 16. The slave boards 60 are connected to a mas~er board 62 over a system bus 64 which in the preferred embodiment is an industry-standard Multi-Bus, but which may be any o~er appropriate bus or interface.
The master board 62 contrs)ls ~e operation of the slave boards 6û, manages connection protocol between the slave boards 60 and the multiplexers 2û, and 25 can monitor and log system ope~a~ions. The en~re system can be con~olled ~ ; ... : ~ . .... . . . . .. .

WO 92/~'~320 ~ '` '3 P(~/U592/00718 and monitored from either the console 66 or from a PC 68. The console 66 is used to directly control the system, such as for installation, maintenance, or in the event that the PC 68 is inoperative or inaccessible The PC 68 is preferably an industry standard personal computer 5 using an Intel 80386 microprocessor and running the UNV( operating system, although a variety of computers or workstations running various operating systems could be used. The PC 68 executes a Network Control System (NCS) program which handles a variety of activities, including down~oading the packet switch's operating firmware to the master 62 and slaves 60 on power 10 up, providing a user interface to a system administrator, logging of errors, and logging of call activity parameters such as calls made, begin time, end time, source node, destination node, digits dialed, etc. In addition, netlwork configuration parameters are stored in the PC 68 such as routing tables, RLX
parameters, and any ac~ons to be taken based upon time of day.
In opera~ion, comrnunication is ini~ated by a ~elephone, FAX
machine, or computer at one system node dialing a number corresponding to a party located at another system node. To establish com~nunica~on, the RLX
20 at ~e source node first requests a connection to the packet switch 44, which be~sins a connection protocol exchange which will be further described in 20 reference to Figure 6. Once connection is established between the source RLX
and t~e desired destination RLX, packetized information can be sent from the source RLX to the destination RLX under ~e direc~ion of the slave board 60 involved, without the need for intervention by ~e master board 62. If the source and destination RLXs are connected to different slave boards, packets ~ :
25 are moved from one slave to the other by the master board 62, although those , . , . , ,, :

- . . . ,: . .. . - . . .............. - , . .. . . . . . ............ .. . .. .
~ . , . , - . . . . .. .. .. .. . .. . .

o 92/1~320 p.~ Pcr/US92/00718 skilled in the art will recognize that it would be possible to have the slave boards 60 control packet movement directly over the system bus 64.
Referring now to Figure 4, a block diagram of a master board 62 is shown. A buffer 72 isolates ~he system bus from the rnaster processor bus 5 74. A m~croprocessor 76 such as an Intel 80C186 is connected over the master processor bus 74 to a va~iety of standard processor peripherals, including a Programmable Read Only Memory ~PROM) 78, Random Access Memory (RAM) 80, a Time of Day Clock 82, and an Interrupt Controller 84. In addition, parallel ports 86 are provided for connec~don to parallel busses or 10 devices as desired in any given implementation. A Serial Channel Controller 88 providers four serial ports 90 through which the master board connects to the console 66, the PC 68, and any other desired serial devices. Finally, a Watchdog Timer 92 is provided to issue a reset to the processor 76 in the event that the software malfunctions.
In the prefelTed embocliment, the PROM 78 contains bootstrap code which serves to download the operating firmware from the PC 68. This facilitates easier maintenance of operating firmware, in that new firmware can be installed using a floppy disk or a file transferred via modem, rather than having to replace an internal device. Those skilled in the art will 20 recognize that many modifications can be made to the master board 62 without departing from the scope of the present inven~on. ~ivr example, firmware for the master board 62 could reside in EPROM, EEPROM, or any other conventionally used non-vola~le storage media. In addi~on, one of ~e serial ports 90 could be connected ~o a data mul~dplexer (not shown) ' - 1:

- . .

o 92/14320 pcr/us92loo7l8 which, under firmware control, could split ~e available bandwidth of trunk lines 16 between packe~zed data and bit interleaved multiplex data.
Referring now to Figure 5, a bloclc diagram of a slave board 60 is shown. A slave processor 96 is connected to its peripherals via a slave 5 processor bus 98. The interfaoe to the master board 62 (not shown) is through a shared Dual-Port RAM 100 to the system bus 64. Comman~ls from the master board 62 to the slaYe boards 60 are written in the shared RAM 100 by the master board 62, and retrieved for execution by the slave processor 96, which in the preferred embodiment is also an lntel 80C186. Firmware for the 10 slave board processor 96 resides in the slave's local RAM 102, where it is do~mloaded to during the power up sequence by the master board 62, ,after receiving the slave firmware from the PC 68. The slave processor bus also connects to a Direct Memory Access (DMA) controller 104, which controls eight DMA channels 106, which support serial communications controllers so 15 as to minimize microprocessor involvement in data transfer. Board control loglc 108 is also provided to control typical operational aspects of the slave board 60, such as power on reset.
Referring now to Figure 6, a flowchart of the connection protocol between the source RLX and the packet switch is shown. In order to ensure 20 that all connections remain reliable cmd intelligible, there must be ade~uate trunk line bandwidth. ~ there is ins~cient bandwidth, packets will be lost, reslilting in reduced intelligibility of the communication. In step 120, the destirlation node of t~e attempted connectiorl is determined from the digits dialeà at ~he source node. Current ac~vity is then monitored in step 122 to 25 deter~une if ~ere is sufficient bandwidth to support the connec~ion. If there .
'' ~' .

,~. , :,...... . .. . . . . ... . . .. . . ... .

wo 92tl4320 Pcr/uss2/oo718 is sufficient bandwidth, a request for connection is sent to ~e destination node at step 124. If the request is granted at step 126, the source node is informed that the connection request has been granted at step 128. If there is insufficient bandwidth at step 122 or if the connection request i5 denied by the5 destination node at step 126, the source node will be lnforrned that the connection request was rejected a~ step 130.
Referring now to Figure 7, a flowchart of the packet flow control according to the preferred embodiment of the present invention is shown.
Flow control allows the bandwidth being used by the commuIucations traffic 10 to be dynam~cally adjusted, sacrificing voice quality when necessary and increasing voice quality when possible. The result is that the voice quality remains at or near the optimal quality possible given the available bandwidth and the current traffic.
I~ at any point in time, there is insufficient bandw~dlth to handle 15 current traffic at a given node, outgoing voice packets will be discarded to ease ~e congestion. When a packet is recelve by the des~na~on node a~ step 140, the RLX che~s the sequence number embedded in the packet to see if the current packet æquentially follows the previous packet. Any gap in the sequence nurnbers of successive packets indicates that one or more packets 20 were discarded as a result of an overflow condition on one of the links in the coIrununication path. If the packet is in sequence (step 14~), the destination RLX signals the source RL)( to decrease the compression slightly at step 144 which increases the bandwidth ~eing used while improving voice quality.
The packet is processed in the usual manner in step 146. Howeves, in the 25 event that a packet is received out of sequence, (step 14'~), lt is presumed that -- . .
. ~
..
, , . . : :: : .

wo 92/14320 ~ Pcr/us92/oo718 the bandwidth needed by the current ~¢affic exceeds the available bandwidth.
The destination RLX then signals the source RLX to increase the compression factor at step 148 which results is a reduced bandwiclth requirement. Aga~n, the packet is prc~essed in the usual manner in step 146.
The amounts by which the compression factor of voice packets is altered depends upon many factors. What is important is that detection of an overflow must increase compression im~ediately and by a rela~dvely large amount so as to eliminate the loss of packets. The clecrease in compression can occur more gradually, and need not occur on every colTectly sequenced 10 packet.

Although the description of the preferred embodiment has been presented, it is contemplated that various changes could be made without deviating from the spirit of the present invention. Accordi;ngly, it is intended15 ~at the scope of the present invention be dic~ated by the appended claims, rather than by the description of ~e preerred embodiment. ~
~ ..

.. ...

Claims

- 17- What is claimed is:

1. A communication system comprising:
a plurality of nodes including a source node and a desired destination node, said source node comprising means for packetizing information and said desired destination node comprising means for restoring packetized information; and switching means operably connected to said plurality of nodes for receiving packets from said source node and switching said packets to said desired destination node.

2. A method of switching packetized information, comprising the steps of:
embedding destination information within a header portion of an information packet;
selecting a destination node for the information packet based upon the destination information contained in the packet header; and switching the information packet to the selected destination node.

3. A method of controlling packet flow in a packetized communication system having a source and a destination wherein packets contain compressed voice information, the method comprising the steps of:

embedding a unique sequence number within each of a plurality of packets being sent from the source and received at the destination;
at the source, discarding a packet if there is insufficinet bandwidth to send it;
at the destination, determining from the sequence numbers contained in the packets received to determine if any packets have been discarded; and increasing compression at the source if a packet has been discarded so as to reduce bandwidth.

4. A method according to claim 3, wherein compression is increased by an amount sufficient to prevent the discarding of packets.

5. A method according to claim 3, further comprising the step of decreasing compression by an incremental amount at the source if packets are being received in correct sequence.