This invention pertains generally to telephony, and more particularly to packet-switched telephony.
Packet-switched telephony is an alternative to traditional circuit-switched telephony. In circuit-switched telephony, a circuit is dedicated to each active telephone call. In packet-switched telephony, the audio signal is partitioned into data “packets” that are individually routed over a data network, sharing the network's pathways with packets from many other sources. Before a packetized audio segment can be played out, its packets must be received and re-assembled.
Packet telephony has several advantages when compared with circuit-switched telephony. Packet telephony uses network resources efficiently, since packets need not be sent when no audio signal is present on the line. Packet telephony also often employs audio codecs (coder/decoder pairs) in order to compress the amount of data that must pass over the network. And with at least some packet telephony systems, users may avoid long distance charges that would be incurred with a circuit-switched connection.
Packet telephony also has disadvantages. Whereas a circuit-switched connection—once established—usually provides fairly consistent connection quality over the length of the call, a packet telephony connection may deliver widely-varying connection quality. Quality variations generally mirror the instantaneous packet delivery performance of the packet-switched connection, which in turn depends on the level of competing traffic. As traffic increases, packets may be delayed, increasing the latency of the connection. An increase in the interarrival jitter, which is an expression of the variation in packet transit times, also increases latency. In some cases, the network or receiver may throw away packets that have been delayed too long, causing dropouts in the reconstructed speech.
A typical packet-switched telephone connection is illustrated in
A typical gateway 28 is illustrated in
It is recognized herein that the uncertain quality of service associated with packet telephony is a major detriment to its widespread acceptance. The present invention attempts to provide a more uniform quality of service by providing a mechanism for re-routing calls from a packet-switched connection to a circuit-switched connection, and preferably, from a circuit-switched connection to a packet-switched connection. In other words, if the quality of a packet-switched connection is poor, the present invention enables a call to be switched to a circuit-switched connection transparently and with no apparent loss of connection. Preferably, if conditions later warrant re-establishing the packet-switched connection, this can also be accomplished.
In one aspect of the invention, a packet voice switch is disclosed. The switch comprises a telephony interface and a packet data interface, and coding/decoding means for converting data from circuit-switched to packet-switched formats (and vice-versa) as it passes between the interfaces. The switch further comprises call switching means that routes calls between the telephony interface and the coding/decoding means. The switch also comprises network switching means (preferably implemented on a host processor) for selecting call routing—the network switching means can route calls received at the telephony interface either back out through the telephony interface over another circuit, or through the coding/decoding means to the packet data interface.
In a second aspect of the invention, a method of operating a telephony switch connected to both a circuit-switched network and a packet-switched network is disclosed. The method comprises calculating the estimated quality of a packet-switched connection for a given call, and selecting, based on that quality, either a packet-switched connection or a circuit-switched connection for that call. The method further comprises connecting the call over the selected network.
In another aspect of the invention, a method of operating a packet voice switch connected to both a circuit-switched network and a packet-switched network is disclosed. The method comprises the steps of receiving an original call request over the circuit-switched network and routing the call request over the packet-switched network to a remote packet voice switch for call connection to a called party. The method further comprises responding to a network change request by rerouting the call over a long-distance circuit-switched network to the same remote packet voice switch.
In yet another aspect of the invention, a second method of operating a packet voice switch connected to both a circuit-switched network and a packet-switched network is disclosed. The method comprises the steps of receiving and acknowledging network switch requests from a second packet voice switch. The method further comprises matching an incoming circuit-switched connection from the second packet voice switch with an existing packet-switched connection from the same second switch, and then switching call voice data from the packet-switched to the circuit-switched connection.
The invention may be best understood by reading the disclosure with reference to the drawing, wherein:
The invention is described below with reference to specific implementations that show, for purposes of illustration, calling and called parties connected to local circuit-switched networks. Those of ordinary skill in the art will immediately recognize that the teachings contained in this disclosure apply equally to parties that establish telephone connections in other ways, e.g., by a local circuit- or packet-switched PBX (Private Branch eXchange), dedicated circuit connection to a central office, gateway, or switch, and/or direct connection to a data network.
For purposes of this disclosure, a packet voice switch is a device that can take call data from a circuit-switched telephony connection, packetize it, and route the packets to a remote destination (as well as accepting and de-packetizing such data and placing it on the circuit-switched connection). This definition thus includes packet voice gateways and gateway/circuit switch combinations. A telephony switch includes packet voice switches and circuit switches that operate in conjunction with, and at least partially control, packet voice gateways.
Without loss of generality,
Operational Overview
The operation of a first, relatively uncomplicated embodiment of the invention is depicted in
A more preferable operational environment is shown in
Hardware
The hardware necessary to implement one embodiment of the invention is shown in
Signal processor 70 also communicates with TDM data bus 72. Processor 70 selects incoming voice data from one timeslot on data bus 72, and supplies outgoing voice data to data bus 72 during a separate timeslot. Processor 70 performs operations such as voice compression/decompression, voice activity detection, and DTMF detection, and supplies data to packet data transceiver 68 for packetization and transmittal. Processor 70 also accepts incoming packet data from packet data transceiver 68 and processes this data to create an outgoing voice data stream. A typical gateway 60 comprises an array of such processors similarly connected to TDM data bus 72 and packet data transceiver 68. Host processor 78 controls overall system operation, and preferable implements the network selection means for the system.
Considering the one-way voice stream originating at a local circuit-switched caller, gateway 60 has the flexibility to transmit the voice stream along one of several paths. One path through gateway 60 is the conventional packet voice application path—through circuit-switched interface 74 to one receiver 62, TDM switch 66, signal processor 70, and packet data transceiver 68 to a packet-switched network. Additionally, gateway 60 can transmit the voice stream along at least one alternate path.
One alternate path is through one receiver 62, TDM switch 66, and back out through one of transmitters 64. Once set up, this path performs nothing more than a timeslot interchange and retransmission of the received PCM data across the circuit-switched network. Note that the transmitter may be connected to the local circuit-switched network (e.g., as shown in
A second alternate path is through one receiver 62, TDM switch 66, signal processor 70, back to TDM switch 66, and then back out through one of transmitters 64. This path can utilize the compression capabilities of the gateway to reduce the data bandwidth as it would be reduced for packet-switched transmission. The compressed data can then, optionally, share an outgoing timeslot with other voice or data transmissions bound for the same destination gateway. Note that such a configuration most likely requires that a compatible gateway exist at the other end to comprehend the shared-timeslot data and re-partition it, and that the intervening connection remain digital. A third possible path is similar to the third path, but includes the packet data transceiver in the loop, and sends the information in packets out over the circuit-switched network.
Determining Network Switch Conditions
Preferably, a gateway or switch operating according to the present invention performs relatively few network switches. Two general methods for determining when a switch is appropriate are comprehended by the invention—a network performance-based approach, and a caller-initiated approach. In the network performance-based approach, a system function trades various aspects of the operational condition of each network pathway against each other, using caller and/or system preferences. In the caller-initiated approach, a caller requests a network switch, preferably by entering a pre-defined in-band signal on their keypad.
Performance-Based Switching
Performance-based switching encompasses a wide range of flexibility, allowing only a few (or many) factors to be considered before a decision to switch a call to a different network is reached. Generally, the results of such a consideration can be quantified as a “figure of merit” (FOM) for each possible network path. The figures of merit can then be compared to reach a decision. An alternative approach is a rule-based approach—in essence, rules avoid computation of an express figure of merit by fitting the factors into a preset decision path, although the results of this approach can be similar to those obtained with a figure of merit approach. Hybrid approaches and other decision-making equivalents are also possible.
For a packet-switched network path, several factors are pertinent to connection performance. One measurable factor is latency. As the estimated one-way latency for a call increases beyond about 250 ms, the perceived quality of the connection decreases rapidly. Of course, one offsetting factor is physical distance, as most callers will tolerate greater latency, e.g., when they call across a continent, than they will tolerate when they call to a neighboring state.
Many latency measurement methods are known for estimating average packet transit times and packet interarrival jitter, which are the main components of latency. See, e.g., W. Montgomery, “Techniques for Packet Voice Synchronization”, IEEE J. on Selected Areas in Comm., vol. SAC-1, No. 6, pp. 1022-1028, December 1983; R. Ramjee, et al., “Adaptive Playout Mechanisms for Packetized Audio Applications in Wide-Area Networks” in Proceedings of the Conference on Computer Communications (IEEE Infocom), (Toronto, Canada), pp. 680-688, June 1994. The particular latency measurement method selected for an application will depend on the transport and packet protocols and network capabilities.
Another measurable factor is packet loss, which includes packets lost or corrupted during transit, or packets received too late to be played out. Lost packet rates of several percent or greater tend to cause noticeable degradation in connection quality.
Latency and packet loss can be used to express a figure of merit, or to trigger rules for network switching. For example, ranges of latency are used in one embodiment to correspond to different levels of quality of service. These levels can also be indexed to ranges of call physical distance.
For a circuit-switched network path, assuming a circuit is available, the main factors to be considered are latency and cost. Latency may be fully comprehended in the distance-indexed quality of service discussed in the previous paragraphs, as the index can account for the expected latency for a circuit-switched connection. Cost is determined by the tariff structure of the circuit-switched network operator, and can most easily be determined by accessing a tariff database.
User preferences can also figure into a network-switching decision. Some users may prefer to use only packet-switched network paths, no matter what the quality of service. Other users may select different service plans that provide for switchover at different cost/quality breakpoints. The appropriate service plan for a caller is preferably retrieved from a database at call set-up. Databases can reside with the system, or can be remotely accessed, e.g., over the packet-switched network.
System considerations may also dictate network selection. For instance, if signal processing resources for a gateway are taxed, the gateway may switch higher quality-of-service level customers to a circuit-switched connection without increasing their cost—thus off-loading calls from their processor and allowing more connections to be serviced. On the other hand, the system may choose not to make circuit-switched connections when its circuit-switched interface is taxed. Such considerations can also be used to limit network switching during periods of high usage to only high QOS customers.
A system may also monitor its own health and switch active calls to a circuit-switched connection upon partial system failure. For instance, if the port used for egress to the packet-switched network fails, active calls can be switched to the circuit-switched network where possible. Or, in a system with multiple processors, each processing a portion of the calls entering the system, failure of some subset of the processors may leave the system unable to handle the signal processing required to process all calls over the packet-switched connection. The system can automatically reduce its processing requirements upon such a failure by switching some calls to circuit-switched connections. When, e.g., the faulty processing board is replaced, these circuit-switched connections can be switched back.
Caller-Controlled Network Switching
Caller-controlled network switching can be used in conjunction with performance-based switching, or in a standalone implementation. With caller-controlled switching, the caller has the ability to perform their own quality/cost analysis and “manually” switch networks. For example, a call may be initiated over the packet-switched network, and at some point the caller may determine that they want better quality for that call than their normal QOS provides, and that they will pay more for it. The caller can enter a DTMF code on their telephone keypad to signal the gateway that they are requesting a network switch.
Initiating Gateway Operation
The initiating gateway is the gateway accessed by the calling party. Normally, this gateway performs path selection (although no technological barrier prevents the receiving gateway from selecting a new path). The path selection operations are at least partially dependent on the functionality offered by the receiving gateway. The receiving gateway's level of functionality can be determined from stored information, or by querying the gateway at call initiation.
Operating Without a Matching Network-Switching Gateway
If the receiving gateway does not support network switching according to the invention, the initiating gateway may be limited to a one-time initial network selection.
Referring to
If circuit-switched connections are allowable, the gateway next determines which network path is most desirable. First, the gateway gathers information. Round trip latency can be quickly estimated for the path to the receiving database. It may also be determined from recent past statistics, or by looking at other in-progress calls to the same ending gateway, if they exist. The gateway also finds the distance to the destination and the applicable tariff for a circuit-switched connection to the called party. These factors are then combined with QOS and system considerations, if applicable, to compute figures of merit for the packet-switched and circuit-switched potential connections. The figures of merit are then compared, and the appropriate connection is attempted.
Operating With a Matching Network-Switching Gateway
If the receiving gateway can perform network-switching operations, the initiating gateway is not limited to a one-time network decision, and can follow a decision flowchart such as the flowchart of
If the QOS level allows network-switching, the gateway retrieves information similar to that retrieved in the embodiment above. Periodically, the gateway examines a time-filtered estimate of latency and computes a packet-switched figure of merit. As long as the packet-switched FOM remains greater than the circuit-switched FOM, no action is taken. If the packet-switched FOM drops below the circuit-switched FOM for the call, the initiating gateway attempts a parallel connection to the remote gateway over the circuit-switched network.
The new connection preferably requires some handshaking between the gateways. The initiating gateway first requests the switch by sending a control packet containing a network switch request to the receiving gateway. If the request is granted by an acknowledgment, the initiating gateway completes a circuit-switched connection by dialing out through its circuit-switched interface to the circuit-switched interface of the receiving gateway. When the receiving gateway answers, the initiating gateway identifies the connection as a parallel of the call's packet-switched connection. Once the new connection is acknowledged (using either connection for acknowledgment), the voice data is re-routed over the circuit-switched connection. The packet-switched connection can then be terminated.
The dashed pathway of
Other signals can also be used to control user preferences in a system according to the invention. For instance, a caller may indicate a QOS level for the current call that is different from their default QOS, without explicitly requesting a circuit-switched connection. The system then uses the new QOS level to make network-switch decisions.
Receiving Gateway Operation
When a receiving gateway acknowledges a request, and knows that it should expect a remote gateway to attempt a circuit-switched connection, it monitors newly-initiated circuit-switched connections. When it identifies the new connection that corresponds to the existing packet-switched connection, the gateway makes arrangements to associate this new pair of timeslots with the resources committed to the packet-switched connection. At an appropriate moment, the connection is then switched.
One of ordinary skill in the art will recognize that many advantageous modifications to the disclosed embodiments are enabled by the concepts taught herein. For example, the gateway/circuit switches can implement PBX (Private Branch eXchange) functions at different sites for a common enterprise. The gateways can decide whether traffic between the sites should use circuit-switched or packet-switched connections. If the enterprise pays for a certain number of dedicated circuits, these can be allotted in a predefined manner as call load between the sites varies. Optionally, compressed voice traffic can use the dedicated circuits instead of the packet data network when conditions warrant.
The present invention also is not limited to implementations where the circuit-switch is performed at the same physical location as the packet-switch. For instance, a circuit switch can gather latency and/or other statistics from an attached packet voice gateway over a control channel, and initiate switches of calls to and from the gateway using this same control channel. The control channel allows the switch to be logically connected, although not directly physically connected, to the packet-switched network for purposes of initiating network switches.
Other modifications to the disclosed embodiments will be obvious to those of ordinary skill in the art upon reading this disclosure, and are intended to fall within the scope of the invention as claimed.
This application is a continuation of prior U.S. patent application Ser. No. 09/304,032, filed May 3, 1999, now U.S. Pat. No. 6,781,983, issued Aug. 24, 2004.
Number | Name | Date | Kind |
---|---|---|---|
5726984 | Kubler et al. | Mar 1998 | A |
5903558 | Jones et al. | May 1999 | A |
5953312 | Crawley et al. | Sep 1999 | A |
5974043 | Solomon | Oct 1999 | A |
6064653 | Farris | May 2000 | A |
6078582 | Curry et al. | Jun 2000 | A |
6115580 | Chuprun et al. | Sep 2000 | A |
6118780 | Dunn et al. | Sep 2000 | A |
6122255 | Bartholomew et al. | Sep 2000 | A |
6137792 | Jonas et al. | Oct 2000 | A |
6192045 | Williams et al. | Feb 2001 | B1 |
6282192 | Murphy et al. | Aug 2001 | B1 |
6307853 | Storch et al. | Oct 2001 | B1 |
6324280 | Dunn et al. | Nov 2001 | B2 |
6452922 | Ho | Sep 2002 | B1 |
6574216 | Farris et al. | Jun 2003 | B1 |
6590869 | Beyda et al. | Jul 2003 | B1 |
6597686 | Smyk | Jul 2003 | B1 |
6628666 | Pickering et al. | Sep 2003 | B1 |
6674713 | Berg et al. | Jan 2004 | B1 |
6731625 | Eastep et al. | May 2004 | B1 |
20030039237 | Forslow | Feb 2003 | A1 |
20030095542 | Chang et al. | May 2003 | A1 |
Number | Date | Country |
---|---|---|
WO 9718665 | May 1997 | WO |
Number | Date | Country | |
---|---|---|---|
20050025134 A1 | Feb 2005 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 09304032 | May 1999 | US |
Child | 10916054 | US |