System and method for maintaining a plurality of media conferences

This invention relates to the field of telecommunications and, more particularly, to a system and method for maintaining a plurality of media conferences.

Many hardware and software applications use digital signal processors (DSPs) to process real-time data communicated over packet-based networks. These applications typically receive input data packets from end-user devices, process the input data packets to generate output data packets, and communicate the output data packets to the end-user devices.

Unfortunately, applications that manipulate real-time data require a minimum amount of memory, processing, or other resources for the communication and processing of the data packets to appear transparent to users. Once the applications allocated to a DSP exhaust the DSP's available resources, current systems restrict the dynamic growth of the applications to prevent performance degradations. For example, devices that process media conferences do not allow media conferences to add new participants if the DSP supporting the media conferences lacks the available resources to support the new conference participants. Hardware and software developers need a system that supports the dynamic nature of many real-time applications, such as media conferencing.

In accordance with the present invention, a system and method for maintaining a plurality of media conferences is provided that substantially eliminates or reduces disadvantages or problems associated with previously developed systems and methods.

In one embodiment, a system for maintaining a plurality of media conferences includes digital signal processors (DSPs) and a system resource management (SRM) module coupled to the DSPs. Each DSP receives input data packets from conference participants, processes the input data packets to generate output data packets, and communicates the output data packets to the conference participants. The SRM module may migrate a media conference from a first DSP to a second DSP.

Technical advantages of the present invention include a system and method for maintaining a plurality of media conferences. By migrating a media conference from a first DSP to a second DSP, the SRM module may free processing, memory, or other resources that the first DSP may use to support new media conferences or new participants in existing media conferences. As a result, new participants may join a media conference allocated to the first DSP without degrading the real-time quality of the media conference. In addition, by avoiding predetermined and fixed limits on the number of participants that may join a media conference and the number of conferences that may be allocated to each DSP, the SRM module improves the DSPs' resource utilization. Other technical advantages are readily apparent from the attached description, figures, and claims

For a more complete understanding of the present invention, and further features and advantages, reference is now made to the following description, taken in conjunction with the accompanying drawings, in which:

FIG. 1 illustrates a system that supports media conferences using a data network;

FIG. 2 illustrates a conferencing device for processing media conferences;

FIG. 3 illustrates a table that includes status information and resource information relating to DSPs in the conferencing device;

FIG. 4 illustrates two bar graphs demonstrating the resource utilization of two DSPs over time;

FIG. 5 illustrates a timeline demonstrating the migration of a media conference from one DSP to another over a mirroring time period; and

FIGS. 6A and 6B illustrate a flowchart of a method for allocating media conferences among DSPs.

FIG. 1 illustrates a system 2 that supports media conferences using a data network 4 . A media conference is a real-time or near real-time communication among three or more participants. To establish and maintain media conferences, system 2 includes data network 4 , user devices 6 a , 6 b , and 6 c (collectively, end-user devices 6 ), gateway device 8 , and conferencing device 10 . End-user devices 6 communicate data streams to conferencing device 10 and receive data streams from conferencing device 10 using data network 4 . Conferencing device 10 receives input data streams from end-user devices 6 , processes the input data streams to generate output data streams, and communicates the output data streams to end-user devices 6 . As described in further detail below with reference to FIG. 2, a system resource management (SRM) module in conferencing device 10 allocates the media conferences to digital signal processors (DSPs) that process the media conferences and migrates media conferences among the DSPs according to the dynamic resource requirements of the media conferences.

End-user devices 6 provide a user interface for conference participants. Each end-user device 6 receives audio, visual, or other sensory input from a conference participant and, in response, generates voice, image, picture, video, or other media information. End-user device 6 communicates an input data stream including the media information to conferencing device 10 and receives from conferencing device 10 an output data stream that conferencing device 10 generates by processing input data streams from other participants' end-user devices 6 . Using voice, image, picture, video, or other media information from the output data stream, end-user device 6 produces audio, visual, or other sensory output for a conference participant. In a particular embodiment, end-user device 6 may be a telephone device 6 a , a computer 6 b , video conferencing equipment 6 c , or any other suitable processing or communications device.

Data network 4 communicates data streams between end-user devices 6 and conferencing device 10 . More specifically, data network 4 communicates data packets using an Internet protocol, an Ethernet protocol, an Asynchronous Transfer Mode (ATM) protocol, or any other suitable network protocol. Data network 4 may include a local-area network (LAN), a wide-area network (WAN), the Internet, or any other suitable packet-based network. In a particular embodiment, end-user devices 6 are coupled to data network 4 by wireless, wireline, or other suitable communication paths. End-user devices 6 communicate data streams to conferencing device 10 by encapsulating the data streams in data packets and communicating the data packets to data network 4 . End-user devices 6 also receive data packets from data network 4 and process the data packets to reconstruct the output data streams generated by conferencing device 10 .

In an alternative embodiment, a gateway device 8 links end-user devices 6 to data network 4 . Gateway device 8 receives data streams from end-user devices 6 , encapsulates the data streams in data packets, and communicates the data packets to conferencing device 10 using data network 4 . Gateway device 8 also receives data packets from data network 4 , processes the data packets to reconstruct the output data streams generated by conferencing device 10 , and communicates the output data streams to end-user devices 6 .

FIG. 2 illustrates in more detail conferencing device 10 for processing media conferences. Conferencing device 10 may be either a stand-alone, networked device or a component operating in conjunction with a networked device. In a particular embodiment, conferencing device 10 is a circuit board coupled to a backplane in a networked, telecommunications device. Conferencing device 10 includes digital signal processors (DSPs) 12 a , 12 b , and 12 c (collectively, DSPs 12 ), a system resource management (SRM) module 14 , an interface 16 , and a memory 18 coupled together by links 20 . Links 20 may be shared or dedicated communication paths that support serial, parallel, or any other suitable form of communication. DSPs 12 receive input data packets from interface 16 , process the data packets to reconstruct input data streams generated by end-user devices 6 , mix the input data streams to generate output data streams, encapsulate the output data streams in data packets, and communicate the output data packets to end-user devices 6 using interface 16 . SRM module 14 allocates media conferences to DSPs 12 and migrates media conferences among DSPs 12 according to the dynamic resource requirements of the media conferences.

Each DSP 12 includes a processing module 22 and a memory 24 . Processing module 22 performs various communication and processing operations according to instructions programmed using hardware, software, or a combination of both hardware and software. Memory 24 may include volatile memory, non-volatile memory, or both volatile and non-volatile memory. Although memory 24 appears internal to each DSP 12 in FIG. 2, memory 24 may be internal to or external from each DSP 12 according to particular needs.

In response to SRM module 14 allocating a media conference to DSP 12 , processing module 22 creates a conference context for the media conference by devoting memory, processing, or other resources to the media conference. Processing module 22 receives from SRM module 14 initiation information indicating a number of participants in the media conference and, in response, creates in memory 24 one or more jitter buffers 26 a , 26 b , and 26 c (collectively, buffers 26 ) for each participant. Once DSP 12 begins processing the media conference, processing module 22 will store in each buffer 26 data packets from an associated conference participant and will process the data packets stored in each buffer 26 to reconstruct an input data stream generated by the associated conference participant. Because DSP 12 may support more than one media conference at a time, memory 24 may include several groups 30 a , 30 b , and 30 c (collectively, groups 30 ) of buffers 26 , each group 30 associated with a separate media conference. In a particular embodiment, processing module 22 dynamically allocates memory 24 to create buffers 26 as DSP 12 receives data packets from conference participants. As part of creating a conference context, processing module 22 may also devote particular processing resources to the media conference.

After creating a conference context for a media conference, DSP 12 begins processing the media conference. Processing module 22 receives data packets associated with the conference participants and stores the data packets in buffers 26 . To reconstruct the input data streams generated by each participant's end-user device 6 , processing module 22 stores the received data packets associated with each conference participant in separate buffers 26 . In a particular embodiment, processing module 22 receives a data packet from interface 16 , examines the source address of the data packet, identifies one of buffers 26 associated with the source address, and stores the data packet in associated buffer 26 . Because the source address of the data packet identifies end-user device 6 that generated the data packet, processing module 22 may use the source address to associate the data packet with a conference participant and to store the data packet in buffer 26 also associated with the conference participant.

To reconstruct the input data streams generated by end-user devices 6 , processing module 22 may resequence the received data packets. Because the data packets are transmitted individually over data network 4 , the data packets may travel different paths from each end-user device 6 to conferencing device 10 . As a result, although the data packets encapsulating each input data stream are typically transmitted sequentially from the same location, the data packets may arrive at conferencing device 10 out of order due to different communication delays associated with the different communication paths in system 2 . The received data packets include identifiers that indicate the proper order of the data packets associated with each conference participant, and using the identifiers, processing module 22 resequences the received data packets to place the packets in their proper order. In a particular embodiment, end-user devices 6 or gateway devices 8 sequentially number the data packets associated with each conference participant to allow processing module 22 to resequence data packets received out of order. When processing module 22 receives data packets, processing module 22 may store the data packets in buffers 26 in order of receipt and then later sort the data packets stored in buffers 26 to resequence the data packets. Alternatively, when processing module 22 receives data packets, processing module 22 may initially store the received data packets in buffers 26 in their proper sequence according to the already stored data packets. The latter embodiment avoids the need to later sort the data packets stored in buffers 26 to resequence the data packets.

After re-sequencing the received data packets, processing module 22 may also identify and regenerate missing data packets. By examining the identifiers of the resequenced data packets in buffers 26 , processing module 22 can identify missing data packets. If processing module 22 does not receive a missing data packet within a threshold period of time, processing module 22 may regenerate the missing packet based on information included in preceding data packets, succeeding data packets, or both preceding and succeeding data packets. The threshold period of time allows processing module 22 a limited amount of time in which to receive any missing data packets without impairing the real-time quality of the media conference. After expiration of the threshold period of time, processing module 22 continues processing the media conference without the missing packets to maintain the real-time quality of the media conference. In a particular embodiment, instead of regenerating missing data packets, processing module 22 inserts null packets in place of the missing data packets.

Processing module 22 retrieves data packets from buffers 26 and processes the data packets to generate one or more output data streams. First, processing module 22 retrieves data packets from buffers 26 and decodes, or linearizes, the data included in the packets. A data packet typically includes media information encoded in a format that facilitates efficient communication over data network 4 . Processing module 22 decodes the data in the data packets to regenerate the encoded media information. For example, in a voice telephone conference, data packets may include voice information encoded according to various standard protocols, such as G.711, G.723, or G.729. In a particular embodiment, a data packet includes compressed data, and processing module 22 decompresses the data before decoding. Processing module 22 then mixes the media information from two or more input data streams associated with a media conference to produce one or more output data streams. In a particular embodiment of a voice telephone conference, processing module 22 produces separate output voice streams for each speaking participant: each speaking participant receives an output voice stream that includes a mix of the other participants' voices but not his or her own voice. Processing module 22 may encode the output data streams for each conference participant according to a coding format and packet size used by the participant's end-user device 6 or gateway device 8 . Processing module 22 encapsulates the encoded data streams in data packets and communicates the data packets to interface 16 for further communication to the conference participants.

SRM module 14 allocates media conferences to DSPs 12 . SRM module 14 may be implemented in hardware, software stored in a computer readable medium, or both hardware and software. In response to receiving a request to create a new media conference, SRM module 14 selects one of DSPs 12 for service and allocates the new media conference to selected DSP 12 by communicating initiation information to selected DSP 12 . As described above, DSP 12 uses the initiation information to create a conference context for the new media conference. After allocating the new media conference to selected DSP 12 , SRM module 14 receives data packets associated with the new media conference and communicates the data packets to selected DSP 12 . As explained above, DSP 12 processes the data packets to generate output data streams and communicates data packets encapsulating the output data streams to the conference participants using interface 16 .

In a particular embodiment, SRM module 14 allocates new media conferences to DSPs 12 according to resource information relating to DSPs 12 . Each DSP 12 must devote a minimum amount of memory, processing, or other resources to a media conference for the mixing of data streams to appear transparent or near transparent to conference participants. Otherwise, processing and buffering delays between the receipt of an input data packet and the communication of a corresponding output data packet degrade the quality of service below an acceptable threshold for media conferences. To avoid such a performance degradation, SRM module 14 may allocate new media conferences to DSPs 12 according to resource information relating to DSPs 12 .

In response to selecting one of DSPs 12 for service, SRM module 14 examines resource information relating to selected DSP 12 . The resource information may identify the amount of resources that DSP 12 already has devoted to existing media conferences, or the resource information may identify an amount of resources available at DSP 12 for processing the new media conference. SRM module 14 may receive the resource information from DSP 12 , or SRM module 14 may generate the resource information based on status or other information stored in memory 18 . As described in more detail below with reference to FIG. 3, memory 18 may store status, resource, or other information relating to the operation of conferencing device 10 . In a particular embodiment, the resource information relates to processing resources, commonly measured in millions of instructions per second (MIPS), and SRM module 14 generates the resource information according to a number of media conferences allocated to DSP 12 , a number of participants in each media conference, and the encoding formats used by the participants' end-user devices 6 or gateway devices 8 . In such an embodiment, the SRM module 14 may generate the resource information by calculating the number of MIPS required to decode, mix, and encode the data streams associated with each participant in the media conferences allocated to DSP 12 . Generally, encoding data streams requires more MIPS than decoding or mixing data streams. For example, processing module 22 may use only 5 MIPS to decode a G.729 data stream and to mix the decoded data stream with other data streams to generate an output data stream, but processing module 22 may need to devote 25 MIPS to encoding the output data stream according to G.729. By calculating the number of MIPS required to support each conference participant, SRM module 14 can generate resource information indicating the number of MIPS devoted to existing media conferences and the number of MIPS available to process new media conferences.

After examining the resource information, SRM module 14 determines whether selected DSP 12 has sufficient resources available to devote to the new media conference. If selected DSP 12 has sufficient available resources according to the resource information, then SRM module 14 may allocate the new media conference to selected DSP 12 . Otherwise, SRM module 14 may allocate the new media conference to another DSP 12 with sufficient available resources. In a particular embodiment, SRM module 14 maintains at each DSP 12 a buffer of available resources so that each DSP 12 may process allocated media conferences in a real-time or near real-time manner even if the resource requirements of the media conferences fluctuates. In such an embodiment, SRM module 14 may allocate the new media conference to selected DSP 12 if doing so would not compromise the buffer of available resource at selected DSP 12 . In response to allocating the new media conference to selected DSP 12 , SRM module 14 modifies the status, resource, or other information stored in memory 18 relating to selected DSP 12 and the new media conference.

SRM module 14 also may migrate media conferences among DSPs 12 . As described above, each DSP 12 must devote a minimum amount of memory, processing, or other resources to a media conference for the real-time mixing of data streams to appear transparent or near transparent to conference participants. The resource requirements of a media conference may increase over time. For example, a new participant may join the media conference, or an existing participant may change the coding format used by his or her end-user device 6 , communicate additional media information to conferencing device 10 , or take other actions that affects the resource requirements of the media conference. SRM module 14 may migrate media conferences among DSPs 12 to ensure that DSPs 12 can meet the changing resource requirements of the media conferences.

In response to an increase in the resource requirements of media conferences allocated to a first DSP 12 a , SRM module 14 may migrate a media conference from first DSP 12 a to a second DSP 12 b to free processing, memory, or other resources at first DSP 12 a . SRM module 14 selects a media conference for migration from first DSP 12 a and allocates the selected media conference to second DSP 12 b by communicating initiation information to second DSP 12 b . Using the initiation information, second DSP 12 b creates a conference context for the selected media conference. Next, SRM module 14 communicates input data packets associated with participants in the selected media conference to both first DSP 12 a and second DSP 12 b for a mirroring period of time. As described in further detail below with reference to FIG. 5, although second DSP 12 b receives the input data packets, second DSP 12 b cannot immediately generate output data streams for communication to the conference participants due to buffering and processing delays. Thus, during the mirroring period, SRM module 14 continues to communicate the input data packets to first DSP 12 a so that first DSP 12 a can continue to generate output data streams associated with the selected media conference while second DSP 12 b fills its buffering and processing pipeline. After the mirroring period, second DSP 12 b begins to generate output data streams associated with the media conference, and SRM module 14 may de-allocate the media conference from first DSP 12 a . SRM module 14 may stop communicating input data packets associated with the media conference to first DSP 12 a and may free processing, memory, or other resources of first DSP 12 a previously devoted to supporting the media conference. In a particular embodiment, SRM module 14 de-allocates the media conference from first DSP 12 a in response to receiving matching output data packets associated with the media conference from second DSP 12 b . In another embodiment, SRM module 14 de-allocates the media conference from first DSP 12 a after a pre-determined number of clock cycles. In response to migrating a media conference from first DSP 12 a to second DSP 12 b , SRM module may modify status, resource, or other information stored in memory 18 relating to first DSP 12 a and the migrated media conference.

In a particular embodiment, SRM module 14 may migrate a media conference from first DSP 12 a to second DSP 12 b according to resource information relating to DSP 12 a . As described above, SRM module 14 may receive resource information from first DSP 12 a or may generate resource information based on status or other information stored in memory 18 . SRM module 14 examines the resource information and, based on the resource information, determines whether to migrate a media conference from first DSP 12 a to second DSP 12 b . SRM module 14 may examine the resource information in response to receiving a request to add a participant to a media conference allocated to first DSP 12 a , and if the resource information indicates that first DSP 12 a does not have sufficient resources available to support the new participant, SRM module 14 may migrate a media conference from first DSP 12 a to second DSP 12 b to free additional processing, memory, or other resources at first DSP 12 a . Alternatively, SRM module 14 may maintain a buffer of available resources at first DSP 12 a , and if the amount of resources devoted to existing media conferences exceeds a resource threshold or if the amount of available resources drops below a resource threshold, SRM module 14 may migrate a media conference from first DSP 12 a to second DSP 12 b to restore the buffer of available resources at first DSP 12 a . SRM module 14 may evaluate the state of the resource buffer in response to adding a new participant to a media conference, changing a coding format used by a participant, or any other action that substantially affect the resource utilization of first DSP 12 a . By migrating media conferences among DSPs 12 according to resource information, SRM module 14 improves resource utilization while allowing conferencing device 10 to adjust to the dynamic resource requirements of the media conferences.

FIG. 3 illustrates a table 30 that includes status information and resource information relating to conferencing device 10 . The status information relates to the allocation of media conferences among DSPs 12 in conferencing device 10 , and the resource information relates to the utilization of processing and memory resources by DSPs 12 to maintain the media conferences. As described above, SRM module 14 may store the status information and resource information in memory 18 . Although FIG. 3 is a table for purposes of illustration, memory 18 may store status and resource information using any suitable data structures and may maintain associations between the data structures using arrays, linked lists, pointers, or any other suitable programming techniques.

A first column 32 , a second column 34 , and third column 36 store status information relating to the allocation of media conferences to DSPs 12 . First column 32 lists media conference identifiers for the media conferences allocated to DSPs 12 . Using participant identifiers, second column 34 associates each media conference identified in first column 32 with the participants included in the associated media conference. In the illustrated embodiment, the participant identifiers are IP addresses (shown in dotted decimal notation) assigned to the participants' end-user devices 6 . Using DSP identifiers, third column 36 associates each media conference identified in first column 32 with DSP 12 that processes the associated media conference. Although the media conference identifiers and DSP identifiers are numbers and the participant identifiers are IP address in table 30 , SRM module 14 may use numbers, letter, addresses, or any other suitable information to identify media conferences, conference participants, and DSPs 12 in conferencing device 10 .

SRM module 14 uses the status information stored in table 30 to support the media conferences allocated to DSPs 12 . For example, SRM module 14 may use the status information to communicate input data packets to appropriate DSPs 12 in conferencing device 10 . In response to receiving an input data packet from interface 16 , SRM module 14 may examine the data packet's source address, identify a conference participant associated with the source address using second column 34 , identify DSP 12 associated with the conference participant using third column 36 , and communicate the input packet to associated DSP 12 for processing.

In table 30 , a fourth column 38 and a fifth column 39 store resource information relating to the resource utilization of DSPs 12 . Fourth column 38 associates each media conference identified in first column 32 with processing information indicating an amount of processing resources, measured in MIPS, that DSP 12 must devote to the associated media conference for the real-time processing of the media conference to appear transparent or near transparent to conference participants. Fifth column 39 associates each media conference identified in first column 32 with memory information indicating the percentage of memory 24 in DSP 12 that is devoted to maintaining the associated media conference. Although table 30 includes processing and memory information in FIG. 3, table may include information relating to any resources that affect the quality of service provided by DSPs 12 .

As described above, SRM module 14 may allocate media conferences to DSPs 12 or migrate media conferences among DSPs 12 according to the resource information stored in table 30 . Using column 38 , SRM module 14 may calculate to amount of processing resources that DSP 12 is devoting to existing media conferences. For example, DSP 12 associated with identification number “1” is devoting 25 MIPS to media conference “1” and 70MIPS to media conference “2” for a total of 95MIPS. If SRM module 14 sets a resource threshold of 90MIPS, then because 95MIPS exceeds the resource threshold of 90MIPS, SRM module 14 may migrate a media conference from DSP 12 to restore the resource buffer at DSP 12 . In a similar manner, SRM module 14 may allocate media conferences to DSPs 12 or migrate media conferences among DSPs 12 according to the memory information stored in column 39 .

FIG. 4 illustrates two bar graphs 40 a and 40 b that demonstrate the resource utilization of DSPs 12 a and 12 b , respectively, over time. Bars graphs 40 a and 40 b represent the utilization of processing, memory, or other resources by DSPs 12 a and 12 b , respectively. Although bar graphs 40 a and 40 b include only a single measure of resource utilization, alternative embodiment may use many different measures to separately represent processing, memory, or other resources that affect the quality of service provided by DSPs 12 a and 12 b . Each bar 42 is divided into a solid bar 44 and a hashed background 46 . Solid bar 44 represents the portion of the resources devoted to maintaining existing media conferences, and hashed background 46 represents the portion of the resources available to support new media conferences or new participants in existing media conferences. At each identified time period, bar 42 has a height 48 indicating a total amount of resources, solid bar 44 has a height 50 indicating the amount of resources devoted to maintaining existing media conferences, and hashed background 46 has a height 52 indicating the amount of resources available to support new media conferences or new participants in existing media conferences. Solid bars 44 are sub-divided into one or more segments 54 representing the portion of resources devoted to separate media conferences. In bar graphs 40 a and 40 b , segments 54 include numbers identifying the separate media conferences. Dotted lines 56 indicate resource thresholds set by SRM module 14 . In a particular embodiment, if the amount of resources devoted to existing media conferences exceeds resource threshold 56 , then SRM module 14 may migrate a media conference from associated DSP 12 .

At time T 1 , SRM module 14 allocates a first media conference to DSP 12 a , and bar graph 40 a indicates that DSP 12 a devotes a portion of its resources to maintaining the first media conference. In contrast, bar graph 40 b indicates that DSP 12 b is not utilizing any of its resources. At time T 2 , SRM module 14 allocates a second media conference to DSP 12 a , and bar graph 40 a indicates DSP 12 a devotes resources to both the first and second media conferences. Again, DSP 12 b is not utilizing any of its resources. As shown in bar graphs 40 a , the amount of resources required to maintain the first and second media conferences is below resource threshold 56 .

At time T 3 , a new participant joins the first media conference allocated to DSP 12 a . As a result of the additional participant, the amount of resources required to support the first and second media conferences exceeds resource threshold 56 . As a result, SRM module 14 selects the second media conference to migrate from DSP 12 a to DSP 12 b to restore a buffer of available resource at DSP 12 a . SRM module 14 maintains the buffer of available resource so that DSP 12 a may have sufficient resources available to support the first media conference if a new participant joins the media conference or an existing participant changes his or her coding format, communicates an additional data stream, or takes any other action that substantially affects the resource requirement of the media conference. At time T 4 , both DSP 12 a and DSP 12 b are using resources to process the second media conference. As described in more detail with reference to FIG. 5, both DSPs 12 a and 12 b receive input data packets associated with the second media conference during a mirroring period. After the mirroring period, DSP 12 b generates output data packets associated with the second media conference, and SRM module 14 de-allocates the second media conference from DSP 12 a . At time T 5 , DSP 12 a supports the first media conference, and DSP 12 b supports the second media conference. As a result, the resources devoted to the media conferences by DSP 12 a and DSP 12 b are below resource thresholds 56 .

FIG. 5 illustrates a timeline demonstrating the migration of a media conference from one DSP 12 a to another DSP 12 b over a mirroring time period. Timing diagrams 30 a and 30 b illustrate the states of DSPs 12 a and 12 b , respectively. over several time periods (labeled t 7 through t 14 ). To demonstrate the migration of a three-participant media conference from DSP 12 a to DSP 12 b , entries 32 identify data packets stored in buffers 26 , entries 34 identify data packets being processed by processing modules 22 , and entries 36 identify data packets outputted by DSPs 12 a and 12 b . To illustrate the movement of data packet through the buffering and processing pipelines of DSPs 12 a and 12 b , time diagrams 30 a and 30 b label data packets using numbers that indicate a time period in which DSPs 12 a or 12 b received the data packets. For purposes of simplifying timing diagrams 30 a and 30 b , DSPs 12 a and 12 b store, process, and output the data packets in order of receipt without re-sequencing data packets or regenerating of missing data packets.

At time t 7 , DSP 12 a outputs data packets “3,” processes data packets “4” and stores data packets “5,” “6,” and “7” for later processing. SRM module 14 selects the media conference handled by DSP 12 a to migrate to DSP 12 b . At time t 8 , SRM module 14 allocates the media conference to DSP 12 b by communicating initiation information to DSP 12 b . Using the initiation information, DSP 12 b creates a conference context, as indicated by empty entries 32 , 34 , and 36 . Also during this cycle, DSP 12 a receives and stores data packets “8” in buffers 26 , and the other data packets progress through the buffering and processing pipeline so that DSP 12 a outputs data packets “4,” processes data packets “5,” and stores data packets “6,” “7,” and “8.” At time t 9 , SRM module 14 begins communicating data packets associated with the selected media conference to both DSPs 12 a and 12 b , as indicated by data packets “9” stored in jitter buffers 26 for both DSPs 12 a and 12 b . In addition, DSP 12 a outputs data packets “5,” processes data packets “6,” and stores data packets “7” and “8.” As shown, due to the buffering and processing pipeline, DSP 12 b cannot immediately generate output data packets in response to receiving data packets “9.” Thus, SRM module 14 continues to communicate the input data packets associated with the selected media conference to DSP 12 a so that DSP 12 a can continue to generate output data packets. During time periods t 10 through t 12 , the data packets progress through the buffering and processing pipelines in DSPs 12 a and 12 b.

At time t 13 , both DSPs 12 a and 12 b output data packets “9,” process data packets “10,” and store data packets “11,” “12,” and “13.” Because outputs from DSP 12 b and 12 a match, SRM module 14 de-allocates the media conference from DSP 12 a at time t 14 , as indicated by the absence of entries 32 , 34 , and 36 . At time t 14 , DSP 12 b maintains the media conference by outputting data packets “10,” processing data packets “11,” and storing data packets “12,” “13,” and “14.”

FIG. 6 is a flowchart illustrating a method of allocating media conferences among DSPs 12 . The method begins at step 100 , where SRM module 14 may receive a request to create a new media conference. If SRM module 14 receives a request to create a new media conference, the method continues at step 124 . If SRM module 14 receives a request to add a participant to an existing media conference at step 102 , SRM module 14 adds the new participant to the existing media conference at step 104 . Because the resource requirements of the media conference increase due to the new participant, the method continues at step 106 . Similarly, if SRM module 14 detects an increase in the resource requirements of a media conference in response to an action by an existing participant, the method continues at step 106 .

At step 106 , SRM module 14 examines resource information relating to first DSP 12 a supporting the media conference and, at step 108 , determines whether the resource information indicates that the amount of resources devoted to media conferences at first DSP 12 a exceeds a resource threshold. If the resources devoted to existing media conferences does not exceed the amount of resource threshold, then the method returns to step 100 . Otherwise, to reduce the resources devoted to existing media conferences, SRM module 14 selects a media conference to migrate from first DSP 12 a to second DSP 12 b at step 110 . SRM module 14 creates in second DSP 12 b a conference context corresponding to the selected media conference at step 112 . During a mirroring period, SRM module 14 communicates to first DSP 12 a and second DSP 12 b input data packets associated with the selected media conference at step 114 , and first DSP 12 a generates output data packets associated with the selected media conference at step 116 . At step 118 , SRM module 14 determines whether second DSP 12 b has begun to generate output data packets associated with the selected media conference. If second DSP 12 b has not begun to generate output data packets associated with the selected media conference, the method returns to step 114 . If second DSP 12 b has begun to generate output data packets associated the selected media conference, then SRM module 14 de-allocates the selected media conference from first DSP 12 a at step 120 , and SRM module 14 receives from second DSP 12 b output data packets associated with the selected media conference at step 122 . In a particular embodiment, SRM module 14 examines the data packets generated by second DSP 12 b and determines whether the data packets match the data packets generated by first DSP 12 a . The method returns to step 100 .

At step 124 , SRM module 14 selects one of DSPs 12 to support the new media conference and examines resource information relating to selected DSP 12 at step 126 . At step 128 , SRM module 14 determines whether the resource information indicates that selected DSP 12 has sufficient available resources to support the new media conference. If selected DSP 12 does not possess sufficient available resources to support the new media conference, SRM module 14 selects another DSP 12 to support the new media conference at step 130 , and the method returns to step 126 . If selected DSP has sufficient available resources to support the new media conference, SRM module 14 allocates the new media conference to selected DSP 12 at step 132 , and the method returns to step 100 .

Although the present invention has been described with several embodiments, a myriad of changes, variations, alterations, transformations, and modifications may be suggested to one skilled in the art, and it is intended that the present invention encompass such changes, variations, alterations, transformations, and modifications as fall within the scope of the appended claims.