Patent Grant
8385937
1. Technical Field
Aspects of the present invention relate to antennas. More particularly, aspects of the present invention relate to modifying beams from antennas to maximize throughput through a network while maintaining quality of service requirements.
2. Related Art
Mobile operators and suppliers constantly search for ways to respond to the increasing demand for ubiquitous mobile services. Mobile operators adjust a network's architecture so that they can introduce new higher speed technologies quickly, while suppliers are working to devise ways for improving the capacity of their wireless products. Current trends indicate:
The current approaches are cumbersome. They do not dynamically adapt directions of frequency channels relevant to at least one of locations and traffic characteristics.
Conventional analytical beam forming techniques usually adjust/control the relevant parameters of a smart antenna such that the signal to noise and interference ratio (SNIR) of each frequency channel is minimized, and its capacity is maximized. The prevalent “optimality” criteria for beam forming techniques are the minimum mean square error (MMSE), and least square (LS) techniques. These techniques use optimal filtering theory to devise a recursive spatial filter that minimizes the square of the difference between the antenna array output and locally generated estimate of the desired signals of subscribers (i.e., a local reference signal) at the transceiver. The MMSE and LS techniques require that the transceiver have either a-priori knowledge or an estimate of the desired signals of subscribers. These estimates are usually obtained using methods such as periodic training sequences, decision directed adaptation, etc. However, they do not dynamically address traffic concerns or locations.
Aspects of the present invention relate to allocating beams to provide improved service to mobile users as their needs change, thus addressing one or more issues with conventional techniques.
Aspects of the present invention relate to controlling beams to load balance frequency channels. Aspects of the present invention may be used in conjunction with one or more databases that contain information on the locations and types of services of the users.
Aspects of the present invention may use present uplinks between access points and mobile terminals that recognize the start of frames. Various approaches may be used including strictly synchronous technologies (e.g., Bluetooth, CDMA) and not as strictly synchronous technologies including explicit interactions between 802.11 access points and mobiles on the uplink that provide relevant framing. At least one advantage may include, when using aspects of the present invention in conjunction with a “packet steering” approach on the uplink, reducing power consumption of the mobiles without any degradation in the overall throughput of the antenna system.
The following describes aspects of the present invention that attempts to increase throughput while satisfying the QoS requirements of real-time services (e.g., voice) of smart antenna systems (WLAN and others) whose controller entity may be connected to a database that contains information on the locations and type of services of the users.
Referring to
Aspects of the invention may be used with systems that have or may obtain a-priori knowledge of each user location, i.e., (ri and θi), as well as the user's type of service, and updates its database dynamically as the users move around. Here, ri is the distance of the i-th user from the array antenna, and θi its azimuthal angle. As one example, the session manager (e.g., SIP server) may provide type of service information to a smart antenna controller upon receiving session set-up and disconnect requests (i.e., “INVITE” or “BYE” messages). Various conventional approaches exist to determine a mobile's location and are not addressed in detail here. For instance, in mobile operator networks, one can easily obtain location information via the global positioning system and ranging or hyperbolic position algorithms.
Here, a smart antenna system is expected to provide N distinct frequency channels, i.e., it has N beam-formers (for instance 3 or 4 beam-formers). The maximum, ωmax, and minimum, ωmin feasible values of beam-widths are known (for instance, values such as ωmin=15° to ωmax=80° may be realized). The system is also expected to support both types of handoff (i.e., mobile assisted hand-off (MAHO) or network assisted hand-off (NAHO)).
The WLAN or other network may support at least one of real-time (e.g., voice) and non-real-time (e.g., best effort data) services. The loads may or may not be symmetric. The network may further exert admission control on real-time services at the session set up time to properly limit the number of simultaneous real-time sessions in the WLAN environ. In general, the network may exert admission control on all sessions except those transporting best effort data traffic. Here, the equivalent capacity of a session may be the estimated amount of capacity that a network should allocate to a session so that the QoS requirements of the session are met.
The N parallel servers 207-209 model the N frequency channels of the WLAN smart antenna. The actual service time of a packet equals the waiting time for access to the channel plus the time it takes to transmit the packet itself. For instance, in 802.11, whenever a mobile has a packet, its MAC layer may send an RTS (request to send frame) asking for “permission” to send its packet. The customers in each queue represent packets of the user population who fall within that frequency channels coverage. In
The antenna controller 201 uses users'locations (211) and relevant information about their sessions such as the desired capacity of a session and its service time (212) to partition the user population, through adjustments in directions of antenna beams, into N subset that “maximize” the throughput of the system, while satisfying quality of service (QoS) requirements of real-time services, e.g., stringent bounds on the delay and jitter of real-time (e.g., voice) packets.
To satisfy the QoS requirements of real-time services, the real-time traffic may or may not be segregated (virtually, not necessarily physically) and may have either pre-emptive priority over non-real-time traffic or receives a periodic service through scheduling.
Also, to maximize the throughput of a queuing system but with only one class of packets, it is helpful to keep the servers as busy as possible.
The following describes the antenna system controller first in view of non-real-time services then addresses real-time services. Since aspects of the described controller distribute the user's traffic load across the frequency channels relatively equally, the controller is referred to as a “load equalizer” or “spatial load equalizer.”
Real-Time Services and Non-Real-Time Services
First, systems with only non-real-time services are addressed. Here, for instance, WLAN environment only supports only non-real-time services that have no delay requirements. To “maximize” the smart antenna throughput and ensure that it is equally likely to have a packet or more in each queue, waiting for transmission on each frequency channel, the controller should partition the user population among the N frequency channels such that the traffic load on the frequency channels are relatively equal, and the beam-widths of all frequency channels are in the admissible range for the antenna array.
If the total number of ongoing sessions is U, and the equivalent capacity of k-th session is Ck bps, then the controller may distribute the ongoing sessions into N neighboring regions/zones such that each of the regions contains approximately one-N-th of the total traffic load, i.e.,
bps, where ┌z┐ represents the smallest integer that exceeds Z. This partitioning process may carried out only based on the azimuthal angles of users location because there is only one type of service and users appear in the location database of the smart antenna (provided they are within the antenna range). In general, partition of the coverage area such that each region/zone contains exactly one-N-th of the total traffic load, i.e., S bps is impractical because each region/zone may only contain neighboring users, and sessions may have different equivalent capacities. Thus, instead of looking for strict equality in regions/zones'loads, the approach described herein opts for a spatial load equalizer (i.e., zoning scheme) that distributes the traffic load relatively equally among the regions/zones.
The following describes a process for the spatial load equalizer. First, the equalizer starts with packing neighboring users/terminals in the 1st region/zone until either the load of the 1st region/zone exceeds one-N-th of the total traffic or the resulting beam-width exceeds ωmax. This is shown by the allocation of mobile users in arc 301 defined by the beam width ω0.
Next, the spatial load equalizer continues with assigning the remaining (i.e., not a member of the 1st region/zone) neighboring users/terminals to the 2nd region/zone until either the load of the 2nd region zone exceeds one-(N−1)-th of the total load of the terminals that has not been assigned to the 1st region. This is shown by users in arc 302 with beam width ω1.
Next, the spatial load equalizer continues with assigning the remaining (i.e., not a member of the 1st and/or 2nd regions/zones) neighboring users/terminals to the 3rd region/zone until either the load of the 3rd region zone exceeds one-(N−2)-th of the total load of the terminals that has not been assigned to the 1st and 2nd regions or the resulting beam-width exceeds ωmax. This is shown here by users in arc 303 with beam width ω2.
This process continues until all N regions or zones are designated.
The width and direction of each beam is set such that it covers one of these regions/zones, and its beam direction points to the azimuthal center of the region.
The spatial load equalizer may be realized with various algorithms that sort the users in view of capacity needed and/or location. For instance, a quick sort algorithm among others may be used. A quick sort algorithm creates a list of ongoing sessions in the descending order of the azimuthal angles of their users. The spatial load equalizer assigns neighboring sessions to a zone/region until either the total load of the zone exceeds S or the resulting beam-width exceeds ωmax. Assuming that the azimuthal angle of i-th session's user location is θi, and the beam-width of the j-th frequency channel is ωj, the pseudo code of the spatial load equalizer may be represented follows:
First, if a smart antenna is capable of forming beams of arbitrary width in the range of (0-π], designated neighboring “load-equalized” regions/zones should not overlap. However, in practice, when the load equalization algorithm/heuristic designates a region/zone whose beam-width is smaller than ωmin, an antenna that directs a beam with width of ωmin to this zone that will overlap with neighboring zones. In other words, if the traffic density at an area is high, this approach may partition the coverage area into narrower non-overlapping zones whose realization in practice results in directing multiple overlapping beams to the heavy traffic area.
Second, as soon as the location database is updated, the controller updates the partitioning accordingly and adjusts the beams directions and beam-widths such that traffic load is almost equally distributed among the N frequency channels. Needless to say when the load equalizer adjusts the smart antenna beams, it may force some users to hand-off even though they may have not moved at all and their locations may not have changed.
The implication of the latter scenario is that a MAHO environment (e.g., IEEE 802.11) also requires a NAHO capability that allows the load equalizer to unequivocally inform those mobiles that are affected by the re-configuration of beams. To ensure correct operation of the load equalizer, the MAHO and NAHO schemes may be used as follows:
The next question is what protocol does the access point use for its NAHO mechanism informing the affected mobiles about beam re-assignments. Since the cause of the beam re-assignment is access point reconfiguration, it may use network management protocol, i.e., SNMP in an IP environment, to instruct the affected mobiles to hand-off to a new beam/channel. The detail specifications of the NAHO hand-off protocol is not part of this disclosure. Alternatively, one can invoke lower layer protocols to perform the required NAHO process. The general approach of the NAHO hand-off process is as follows:
Third, regardless of differences in the actual loads of resulting zones, the “load-equalizing” approach primarily assigns contiguous users and their sessions to a zone to ensure smallest possible beam-width and maximum possible range extension even though non-contiguous distribution of users may result in a less difference among the loads of each region/zone.
Fourth, a user may have several simultaneous ongoing sessions. When such a user has only one interface, the load-equalizing algorithm may assign all sessions of this user to one of the frequency channels, even at risk of increasing the “equalization” error (that is intentionally making the frequency channels unequal).
In a special case where all users have identical non-real-time sessions (i.e., equivalent capacity of all sessions are identical), the load-equalizer may partition the coverage area of the smart antenna into N regions/zones such that each region with “equal” session populations, i.e., each frequency channel serves almost 1/N of the sessions. More specifically, if the total number of ongoing session is U, and U=(MN+r), and assuming that the equivalent capacity of k-th session is Ck=1 for all 0≦k<U, in the pseudo-code of the load equalizer algorithm, the result is a zoning that distributes these sessions into N neighboring zones such that r of the zones contains M+1 sessions, and the remaining (N−r) zones each have M sessions.
The described spatial equalizer may determine the beam-widths in a WLAN environment that supports real-time and non-real-time services concurrently as well. The spatial equalizer performance is acceptable if the load of ongoing real-time sessions in each frequency channel (or beam) does not exceed the maximum real-time load that it can support. However, if the load of real-time sessions exceeds the maximum that a channel can support, then the QoS requirements of these sessions may not be satisfied. The system may or may not then use only locations of real-time users/sessions to adjust the antenna beams. This approach may satisfy the QoS requirements of the real-time sessions, though it may degrade the performance (e.g., throughput, packet loss) of the non-real-time services because it may result in skewed distribution of non-real-time sessions across the beams. The following describes another approach to addressing multiple classes of service.
The temporal-spatial load equalizer may be a time-division variant of the spatial load equalizer that time-shares the smart antenna system between the real-time and non-real-time services such that the throughput is “maximized” and the QoS requirements of the real-time services are satisfied. The temporal-spatial load equalizer divides the time into equal frames of size T, where every frame comprises a real-time and a non-real-time cycle. During the real-time cycle, only real-time sessions that have packets can transmit. Similarly in the non-real-time cycle only sessions with non-real-time services can send packets. During each real-time (or non-real-time) cycle, the behavior of the temporal-spatial equalizer is similar to that of a spatial load equalizer that supports only real-time (or only non-real-time) sessions. In other words, during the real-time (or non-real-time) cycle, it distributes the real-time (or non-real-time) sessions into N regions/zones whose traffic loads are relatively equal, and tailors the beam-widths and directions of the frequency channels to the regions/zones.
The frame size T is usually equal to the time it takes the slowest real-time application to generate a packet on the session. For example, for a voice application software that generates a voice packet every 20 ms, 40 ms, or 80 ms, T may be set to 20 ms, 40 ms, or 80 ms for the environment in which the voice application software is used.
Specifically, assuming that the mobile stations and the smart antenna correctly recognize the start of the frames (if and when necessary), i.e., having consistent framing structure across the access point and mobiles, the temporal-spatial load equalizer operates as follows:
Specifically, the controller re-adjusts the antenna beam-widths and directions such that the load of non-real-time sessions is equally distributed among them. The downlink may wait for an interval to pass as well. However, on the downlink an optimization may be realized in that the controller may start the shift to the non-real-time cycle as soon as the smart antenna has depleted the real-time services buffers.
On the uplink, the preceding algorithm requires that the mobiles and the smart antenna system correctly identifying the start of the frames. Otherwise, on the uplink, the temporal-spatial load equalizer may not operate properly because users are distributed across the media and have no single reference for the start of a frame, or different cycles within a frame. However, there is no need for strict framing on the downlink because the antenna itself knows the start of a frame or its cycles, and mobiles'receivers pick up the signal when they receive the signal. Two approaches may be used for ensuring consistent framing structure across the access points and the mobiles on the uplink,
Strictly synchronous technologies such as Bluetooth, CDMA) automatically provide consistent framing structure across the access point and mobiles. However, the synchronization on the uplink may create an issue for 802.11b WLAN because the IEEE 802.11b specifications do not provide means of such a strict synchronization between the access point and mobiles. Bluetooth Masters (analogous to APs in WLANs) in fact, do have strict synchronization with Slaves (analogous to mobiles in WLANs), and they do know when each slave will transmit and do know whether each Slave is real-time (e.g., voice) or non-real-time (e.g., data).
There are at least two ways to deal with the 802.11b issue. The first solution is the use of a polling mechanism in the smart antenna system that determines which user has the right to transmit. This approach is analogous to the Point Coordination Function (PCF) specified in the specifications of the 802.11 media access control (MAC) scheme. The advantage of this approach is that it allows contention free voice cycle and potentially increases the number of simultaneous voice sessions per frequency channel. In a nutshell, this addition to 802.11 MAC provides means of four classes of priorities for users packets and an enhanced polling mechanism similar to the PCF that is dubbed as hybrid coordination function (HCF) in this specifications. Preliminary performance evaluation studies show that “the HCF provides means of delivering time-bounded traffic, but requires all stations within the range of the HC (Hybrid Coordinator) to follow its coordination.”
In the absence of strict synchronization, one can use explicit messaging between the access point and the mobiles to provide a consistent framing structure on the uplink beam. In this approach, the access point periodically sends announcement messages to inform the mobiles about the start of a new frame. Having announcement messages, the access point may also opt for using it to also inform the mobiles about the end of the real-time (start of non-real-time) cycle within each frame. This use of announcement messages for informing mobile about the end of the real-time (start of non-real-time) cycle may improve the throughput of the system on the uplink because it reduces the silence between the cycles by τ seconds. A start of frame announcement message is an 802.11 MAC_PDU whose 2 byte Control Field are encoded as follows:
This approach introduces negligible additional signaling overhead, and its START/END control MAC_PDUs do not require acknowledgements by the mobiles because their payloads are null, and the 4 byte long FCS field of the MAC_PDU protects its 2 byte long control field.
Some have proposed an adaptive antenna that measures the aggregate transmission rate of users over its frequency channels and adapts their beam-widths such that the sum of beam-widths of all frequency channels (i.e., the coverage area) remains constant; the beam-width of the channel with highest traffic is narrowed a certain amount (say λ degrees), and the beam-width of the channel with lightest traffic is widen the same amount (i.e., λ degrees); and the beam-widths of other frequency channels remain constant. At least one of the following distinguishes a load equalizing smart antenna from the adaptive antenna scheme as described above in that a load equalizing smart antenna:
Assuming that the system supports only identical voice services and identical best effort data services,
In
In
For simplicity in the temporal-spatial equalizer example of
In a wireless environment, benefits found in reducing the power consumption of each mobile. In order to do so, the load equalizer may be used in conjunction with packet steering approach as known in the art on the uplink beam. Using packet steering, after receiving the request to send (RTS) message from a mobile and granting transmission right to it, the load-equalizing smart antenna directs a point-to-point receive (uplink) beam, i.e., as narrow a beam (e.g., 15°±5°) or as narrow as the antenna array can create, to the location of the mobile. After receiving the packet from the mobile, it may send its acknowledgement (ACK) packet on the downlink beam while restoring the uplink beam to its normal beam-width set by the load equalizer so that it can hear RTS packets of all mobiles assigned to a frequency channel. The smart antenna repeats this process after granting the transmission right to a mobile.
Components of the Temporal-Spatial Load Equalizer
The controller 702 may include a SIP message processor (SMP) 710 that exchanges information with outbound SIP proxy 706 and database update engine 711. Controller 702 may also include a database update engine 711 that exchanges information with resource manager 708, SMP 710, and policy server 709 and stores information in location database 712. Controller 702 may further include a load equalizing (and control) engine 713 that exchanges information with location database 702.
Antenna 701 may include a spatial-temporal de-multiplexer 714 and a spatial-temporal multiplexer 716. Both may be connected to processor 715. The antenna 701 may further include a multi-channel receiver 717 connected to the de-multiplexer 714 and processor 715. The antenna 701 may further include a multi-channel transmitter 719 connected to the multiplexer 714 and processor 715. The multi-channel receiver 717 and multi-channel transmitter 719 may be connected to the antenna array 718.
Processor 715 computes (or captures necessary data for computation of) the direction of arrival (DOA) of users'mobile terminals and passes it (or them) to the load equalizing engine 713. The processor 715 may also receive the beam-widths from the load-equalizing engine of the controller 702 (or controllers 702) and translate them into appropriate weight vector for steering the antenna beams from antenna 701 (or antennas 701). For each network 703, there may be one or more controllers 702 and one or more antennas 701. For each controller 702, there may be one or more antennas 701. Further, each controller 702 may be connected to one or more networks 703.
The controller 702 obtains and processes information regarding the users'locations and their services and turns the information into control actions for steering the antenna 701's beams in accordance with the operator's policies. For instance, the load-equalizing engine 713 may be a policy enforcement point (PEP) whose task is to use users location and services data to maximize the throughput of the smart antenna, i.e., it may realize the temporal-spatial load-equalizing approaches described above. The load-equalizing (or in general, the control) engine 713 represents a dynamic policy enforcement point that can enforce any operator's policies. The controller 702 may include the following:
The SIP Message Processor (SMP) 710 receives a copy of signaling messages from and to users located within the coverage area of the antenna 701, processes/filters them to get (or update) information regarding users and their services and sessions, and forwards the information to the database update engine 711 for updating information in location database 712;
The controller 702 is shown as separate from antenna 701. Alternatively, controller 702 may be integrated with antenna 701. Further, controller 702 may be integrated with outbound SIP proxy 706 in network 703 or somewhere between network 703 and antenna 701 (e.g., integrating the SIP message processor (SMP) 710 with the outbound SIP proxy 706 and the rest of its entities with the antenna 701. In general, the controller 702 may become an integral part of a mobile communication server (MCS).
User Services and Sessions
The following relates to relevant information regarding user'services and sessions and how controller 702 initializes location database 712, obtains relevant data, and updates location database 712.
User Data Object
Each user's mobile terminal may be identified by a user data object in the location database 712 that contains all relevant information on the location of the terminal and ongoing sessions and services originated from or terminated at the user's terminal. A user data object 801 may include one or more of the following attributes as shown in
In general, the triplet comprising the MAC address of user's terminal, its IP address, and the user's SIP URL may be used to identify and update a user data object in the location database 712 of the antenna 701. The MAC address of the terminal identifies it regardless of user services and/or mobility patterns as well as network protocols for supporting mobility. The MAC address may operate as one or more object identifiers in the location database 712 for associating user services and sessions to the location of the mobile terminal. The SIP signaling messages identify a user's session by her/his URL and IP address. The body of the SIP REGISTER message may associate/map user URL and IP address with/to MAC address of her/his terminal. The SIP URL in the session setup messages may associate a session with the user but not always with her/his terminal due to the service mobility. For instance, a user may transfer an ongoing session from her/his mobile to another appliance (e.g., her/his office telephone) that does not necessarily appear in the location database. In this case the IP address within the SIP RE_INVITE (or UPDATE) message may indicate that this session has been transferred to another appliance and can be removed from the user data object representing her/his mobile and its ongoing activities. It allows the database update engine 711 to remove this session from the user data object.
Initialization and Update of the Location Database
An attribute of the user data object that assists in a load equalizing antenna is the location or the azimuthal angle of the location of the user's terminal. The angle may be detected directly or determined from absolute location identification information (GPS, triangulation, or hyperbolic location identification information). An antenna system may use received signals in conjunction with either a position location or Direction of Arrival (DOA) estimation algorithm as known in the art to obtain the azimuthal angle of the user (i.e., terminal) location. Three issues exist with using DOA information in a smart antenna:
To address the first issue, the antenna's beam-widths and their angle may be initialized such that they cover the whole area at the system startup time. The load equalizing (control) engine 713 assigns initial beam-widths to ωk=π/N, 0≦k<N. For instance, a load-equalizing engine starts a three beam antenna with ω0=ω1=ω2=ω3, and starts enforcing the load-equalizing methodology as soon as the initialization interval is expired. The exact value of the initialization interval duration depends on how long it takes the DOA algorithm to determine the locations of all active terminals within the coverage of the system.
The database update engine 711 then initializes users'data objects with at least one of their locations and MAC addresses, while setting the remaining objects to “NULL” until the database update engine 711 receives information on the URLs, IP addresses, and Session IDs and traffic descriptors from the SIP Message Processor (SMP) 710. Database update engine 711 processes the information received from the SIP Message Processor (SMP) 710 to derive and update the remaining attributes of a user data object.
To address the second issue, i.e., allowing users to rejoin the system when they become active again after a period of silence is similar to that of the first question, with a difference that the latter is a periodic process, while the former may be a task that is performed relatively infrequently. In other words, the load equalizing (control) engine 713 of an antenna 701 that has N beams periodically (for instance, every Tp sec) suspends its load-equalizing beam forming, and creates beams of width ωk=π/N, 0≦k<N, to sweep the whole coverage area and instate (or re-instate) all the active users in the location database. As in the initialization task, the length of the suspension time depends on the DOA algorithm. The trade-offs in the choice of the database updating period, Tp, are that it should be 1) much larger than the temporal-spatial frame size T so that the impact of these updates on the performance of the load-equalizer is relatively negligible, and 2) small enough to ensure that chance of locking out an off-hook user is minimal.
Alternatively, another beam may be created (from the same antenna array or another antenna array) to sweep the area to check for all, new, or rejoining mobile terminals.
To address the third issue, the database update engine 711 may modify the database upon reception of new locations, sessions, and services information from either the SIP Message Processor (SMP) 710 or the processor 715 of the antenna 701.
When the database update engine 711 receives data from either the SIP Message Processor (SMP) 710 or the processor 715 of the basic smart antenna, the database update engine 711 processes the information (e.g., calculates the equivalent capacity of the session) as needed, and updates the corresponding user data object, if the object already exists in the database. Barring abnormal conditions, the object may already exist in the database. Otherwise, it creates a new user data object in the database whose identifier is used to access it. The identifier may or may not be the MAC address of the “new” user's terminal.
The database update engine 711 may expunge the stale user data objects if and when needed. For instance, old objects (a few minutes or hours old with no subsequent updates) may no longer be relevant. If the antenna does not receive signal from a user mobile for an extended period of time, it considers her/his user data object stale. In this regard, a user data object may be considered stale, if either the user terminal is broken or it has not renewed its registration for a long period (as specified by the operator's policies) of time.
Obtaining Information on Users Sessions and Services
The controller 702 maintains information regarding a user, his/her terminal MAC and IP addresses, and ongoing sessions and services from SIP signaling messages. With regards to IEEE 802.11 WLANs, a terminal usually registers with the access point first before being able to send information. Thus, the MAC address and location attributes of a user data object usually appear in the location database before the others.
The attributes of a user data object that may be obtained from SIP messages are user SIP URL, IP address and corresponding MAC address of her/his terminal, “Session-IDs” of its ongoing sessions, and their service requirements. Alternatively, these may be derived from other sources. The URL and terminal addresses at the MAC and IP levels can be obtained during SIP registration message flow, while the information regarding ongoing sessions and their services can be derived from the SIP call setup message exchange. The SIP outbound proxy 706 is in the route of signaling messages of mobile originated and mobile terminated sessions of users within the coverage area of the antenna 701. It copies signaling messages to and from users within the coverage area of the antenna 701 to the SIP Message Processor (SMP) 710. The SIP Message Processor (SMP) 710 filters these messages to obtain the required information for passing to the database update engine 711.
The SIP Message Processor (SMP) 710 ensures that the user data objects of the location database contain correct and up to date information regarding the URLs, IP addresses, and MAC addresses of users. The SIP Message Processor (SMP) 710 may perform this and other tasks through one or more of the following:
The following describes the separation of resource and session management in SIP and its implications. As is known in the art, the SIP specifications separates resource management and session management functions. As far as SIP is concerned, resource management is a pre-condition for setting up sessions. Here, the request for setting up a session with a certain amount of resources cannot be completed until resources are reserved, though SIP itself does nothing about getting these resources, and it is up to the end users to obtain necessary resources. However, other protocols may permit the network to take a more active role in reserving the resources.
The message flow for setting up a basic session with a resource preconditions (e.g., a phone call) is shown in
In step 901, a call starts with an INVITE message from the caller containing an SDP (session description Protocol) PDU (Protocol Data Unit) in the body of the INVITE message. The SDP PDU contains within the body of an INVITE message contains the service requirements of a session. The SDP PDU (shown as SDP1) contains the session media information that informs the callee about the type and requirements of the service(s) that the caller terminal can support.
Next in step 902, the callee responds with a 183 session in progress containing an SDP PDU (shown as SDP2) indicating the capabilities of the callee's terminal.
In steps 903 and 904, after receiving the “183” message, the caller and callee exchange PRACK and OK messages. The acknowledgement messages are followed by a reservation period 905 in which the caller and callee reserve resources by their own means and protocols (e.g., RSVP, etc.).
When the reservation process is complete, the caller sends an UPDATE message 906 containing an SDP PDU (shown as SDP3) that describes the media reservations in the caller to callee direction.
The callee sends a “200” OK message 907 that contains an SDP PDU (shown as SDP4) showing media reservation in the callee to caller direction.
The callee sends a ringing message 908 to the caller to alert the caller that the callee is now ringing.
A pair of acknowledgement signals is exchanged as 909 and 910.
Then callee then indicates the call is connected in 911 and the caller acknowledges in 912 and the session starts. Finally, the caller sends an ACK message to the callee, and the session begins in step 913.
The preceding message flow shows that the SIP Message Processor (SMP) 710 processes the INVITE-SDP1901, UPDATE-SDP3906, and 200 OK-SDP4907, and the final ACK 912 messages of the message flow to obtain information regarding a session and its service. The SIP Message Processor (SMP) 710 may or may not ignore all other call set up messages regarding each call that are copied to it. The SIP Message Processor (SMP) 710 may process the INVITE 901, UPDATE 906, 200 OK 907, and ACK 912 messages of each call as follows:
Upon the reception of each INVITE 901 from the outbound SIP proxy 706, SIP Message Processor (SMP) 710 processes the header of the INVITE message to get the Session-ID, caller's SIP URL and IP address, and creates a temporary data object containing one or more of these data.
Upon receiving an UPDATE message with the same Session-ID from the outbound SIP proxy 706, SIP Message Processor (SMP) 710 obtains the media information from the SDP PDU within the UPDATE body. The SDP PDU media identifies the requirements of this session service in the uplink direction. The SIP Message Processor (SMP) 710 adds this session media (i.e., service requirements) on the uplink to the temporary data object whose Session-ID attribute is identical to the Session ID of the UPDATE message.
Upon reception of the subsequent 200 OK 904 with the same Session-ID, the SIP Message Processor (SMP) 710 obtains the media information from the SDP PDU within the body that specifies the service requirements of the session on the downlink. Then, the SIP Message Processor (SMP) 710 includes this downlink session media information into the temporary data object whose Session ID attribute is the same.
Finally, when the SIP Message Processor (SMP) 710 receives a subsequent ACK with the same Session-ID, it forwards this temporary data object to the database update engine 711.
The database update engine 711 processes the forwarded temporary data and updates the corresponding user data object 801, accordingly. The database update engine 711 receives the temporary data object containing one or more of a URL, an IP address, a Session ID, and uplink and downlink media information. The database update engine 711 sends a message to the resource manager that contains the Session ID as well as the uplink and downlink media information of the session. Depending on the resource management scheme of the network, the resource manager either computes the equivalent capacity and sends it to the database update engine 711, or sends the traffic descriptors of the session's uplink and downlink media to the database update engine 711 to compute the equivalent capacity itself. Then, the database update engine 711 uses the MAC address, URL, and IP address to uniquely identify and update the corresponding user data object in the location database.
The control system and technique may further be applied to other arenas as well. Further, one may standardize code points for the Type and Subtype fields of the announcement MAC_PDUs of a WLAN that ensure proper framing on the uplink of a load equalizing smart antenna. The load equalizing smart antenna may be used with other similar antennas and conventional (non-load-balancing) antennas.
Session Admission Control in Packet Networks
The antenna and control system may control load through selective session admission in packet networks. Session (or call) admission control decides whether to accept or reject a request of setting up a new session in accordance with a network's traffic management policies. Since a network supports different classes of services with distinct capacity needs and QoS requirements, an admission control policy generally attempts to establish fair blocking among the service classes, assure that sufficient network resources are available for each admitted session, and maintain an acceptable utilization level for the overall pool of network resources.
The traffic management of packet (ATM or IP) networks has been studied extensively. Numerous admission control, scheduling, flow control, policing and shaping techniques exist and may function with the antenna load-balancing system described herein. Two approaches are described. These approaches are session admission control using either “equivalent capacity” of a session, or the “admission region” of a node in the network.
For admission control using equivalent capacity, the network uses a traffic descriptor of a session and its QoS requirements to estimate the amount of capacity that is required (referred to as equivalent capacity) for supporting the session. Various analytical expressions for derivation of the equivalent capacity have been proposed. The equivalent capacity of a session may be chosen such that, for a given buffer size of X packet per session at the access interface, the packet loss ratio of the session on the access link does not exceed a specified value β agreed upon in the service layer agreement of the session. The primary QoS constraint on the performance of real-time services is that packet delay should not exceed an upper bound T. Thus, in order to satisfy it, the maximum buffer size per a real-time session should be X≦0.5┌T/μ┐, where μ is the mean service (access plus transmission) time of a real-time packet on the frequency channel.
If the peak rate of the traffic on the i-th session is Rp(i), its mean rate Ra(i), and its mean burst length δi−1. Assuming that the traffic activity on the i-th session conforms to a two-state Markov chain model, and using the results of Anick et al. (D. Anick, D. Mitra, and M. M. Sondhi, “Stochastic Theory of Data Handling System with Multiple Sources”, Bell System Technical Journal (BSTJ), Vol. 61, No. 8, October 1982) with a simple fluid approximation of the session traffic, Guerin, et al. (R. Guerin, H. Ahmadi, and M. Naghshineh, “Equivalent Capacity and Its Application to Bandwidth Allocation in High-Speed Networks”, IEEE Journal on Selected Areas of Communications, Vol. 9, No. 7, September 1991) arrives at the following expression for equivalent capacity, C1, of the i-th session:
where α=−Ln(β) and ρi=Ra(i)/Rp(i). It is worth noting that as ρi→1, e.g., a voice telephony session, then Ci→Rp(i), i.e., the equivalent capacity of a constant bit rate session equals its peak rate.
The network admits the request for setting up a session if the free capacity on the access channel (link/medium) exceeds the equivalent capacity of the requested session, i.e., ΣjCj≦C, where C is the channel capacity. It is worth noting on a multiple access share medium channel such as an 802.11 WLANs, C represents the maximum throughput of the media under “nominal” operation of its MAC protocol.
For an admission control using admission region approach, the network uses the admission region “maps/curves” as guidelines to decide whether to accept or reject session set-up requests in packet networks. These maps/curves may be stored in the network elements such as switching nodes, smart antenna controllers, or elsewhere or not stored in the network yet access from the network. An admission region is a map/curve whose points represent the number of simultaneous sessions of different services classes that can be supported by the channel or node.
The admission region maps are usually obtained off-line, through measurement, simulation, or analysis, assuming a predetermined number of service classes in the network in the network. This approach may be effective for small networks with limited number of service classes.
For example, in case of an antenna system supporting only voice and best effort data services, one needs to measure (or derive) the admission region of each frequency channel as a function of its beam-width. The admission region is a curve that shows the number of simultaneous voice sessions that a system can support versus the volume of data traffic present in the system. One benefit of using an admission control using an admission region approach is that the admission region reflects the nominal operation of the MAC protocol.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5790070 | Natarajan et al. | Aug 1998 | A |
| 5815116 | Dunbridge et al. | Sep 1998 | A |
| 5936577 | Shoki et al. | Aug 1999 | A |
| 6104930 | Ward et al. | Aug 2000 | A |
| 6181955 | Dartois | Jan 2001 | B1 |
| 6229486 | Krile | May 2001 | B1 |
| 6236849 | Reudink et al. | May 2001 | B1 |
| 6301238 | Hagerman et al. | Oct 2001 | B1 |
| 6347220 | Tanaka et al. | Feb 2002 | B1 |
| 6768913 | Molnar et al. | Jul 2004 | B1 |
| 6804521 | Tong et al. | Oct 2004 | B2 |
| 6937863 | Gordon et al. | Aug 2005 | B1 |
| 20020034967 | Taniguchi et al. | Mar 2002 | A1 |
| 20020089956 | Haugli et al. | Jul 2002 | A1 |
| 20020115474 | Yoshino et al. | Aug 2002 | A1 |
| 20020146983 | Scherzer et al. | Oct 2002 | A1 |
| 20030017853 | Kanamaluru et al. | Jan 2003 | A1 |
| 20030032429 | Macridis et al. | Feb 2003 | A1 |
| 20030050051 | Vilander | Mar 2003 | A1 |
| 20030118015 | Gunnarsson et al. | Jun 2003 | A1 |
| 20030181163 | Ofuji et al. | Sep 2003 | A1 |
| 20040077379 | Smith et al. | Apr 2004 | A1 |
| 20040095901 | Rajkotia | May 2004 | A1 |
| 20040233872 | Lobinger et al. | Nov 2004 | A1 |
| 20050197060 | Hedinger et al. | Sep 2005 | A1 |
| Entry |
|---|
| J. Rosenberg et al., “SIP:Session Initiation Protocol”, pp. 1-240, dated Jun. 2002. |
| Tsern-Huei Lee et al., “Design of a Real-Time Call Admission Controller for ATM Networks”, IEEE/ACM Transactions on Networking, vol. 4, No. 5, pp. 758-765, Oct. 1996. |
| M. Decina et al., “Bandwidth Assignment and Virtual Call Blocking in ATM Networks”, IEEE, pp. 881-888, 1990. |
| ArrayComm, “IntelliCell®: A Fully Adaptive Approach to Smart Antennas”, pp. 1-18, dated 2002. |
| Internet printout: http://www.vivato.net/prod—tech—technology.html—“Vivato Switches are “Changing the Physics” of Wi-Fi”, dated Apr. 20, 2004. |
| Lehne et al., “An Overview of Smart Antenna Technology for Mobile Communications Systems”, IEEE Communications Surveys, Fourth Quarter 1999, vol. 2, No. 4, pp. 2-13. |
| Joseph C. Liberti, Jr., et al., “Introduction to Smart Antennas: Spatial Processing for Wireless Systems”, Smart Antennas for Wireless Communications, pp. 81-116. |
| Brian D.O. Anderson et al., “Filtering, Linear Systems, and Estimation”, Optimal Filtering, pp. 9-61, 1979. |
| Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, ANSI/IEEE Std. 802.11, 1999 Edition. |
| IEEE P802.11 Wireless LANs, Normative Text for Tge Consensus Proposal, pp. 1-25, Sep. 2002. |
| Stefan Mangold et al., “IEEE 802.11e Wireless LAN for Quality of Service”, pgs. |
| Dr. Richard Roy, Presentation, ArrayComm, “Smart Antenna Technology—Past, Present, and Future”, ITS ART Symposium, Sep. 9-11, 1998. |
| G. Camarillo et al., “Integration of Resource Management and SIP”, pp. 1-24, Oct. 2002. |
| R. Braden et al., “Resource ReSerVation Protocol—(RSVP)—Version 1 Functional Specification”, p. 1- 112, Sep. 1997. |
| The ATM Forum Technical Committee, Traffic Management Specification Version 4.0, Apr. 1996. |
| The ATM Forum Technical Committee, Addendum to Traffic Management V4.0 for ABR parameter negotiation, pp. 1-3, Jan. 1997. |
| Guest Editorial, “Congestion Control in High-Speed Packet Switched Networks”, IEEE Journal on Selected Areas in Communications, vol. 9, No. 7, pp. 965-967, Sep. 1991. |
| Marek Wemik et al., “Traffic Management for B-ISDN Services”, IEEE Network, pp. 10-19, Sep. 1992. |
| Roch Guérin et al., “Equivalent Capacity and Its Application to Bandwidth Allocation in High-Speed Networks”, IEEE Journal on Selected Areas in Communications, vol. 9, No. 7, pp. 968-981, Sep. 1991. |
| D. Anick et al., “Stochastic Theory of a Data-Handling System with Multiple Sources”, The Bell System Technical Journal, vol. 61, No. 8, pp. 1871-1895, Oct. 1982. |
| Israel Cideon et al., “Bandwidth Management and Congestion Control in planet”, IEEE Communications Magazine, pp. 54-64, Oct. 1991. |
| F. P. Kelly, “Effective bandwidth at multi-class queues”, Queueing Systems 9, pp. 5-16, 1991. |
| Anwar I. Elwalid et al., “Effective Bandwidth of General Markovian Traffic Sources and Admission Control of High Speed Networks”, IEEE, pp. 256-265, 1993. |
| G. Gallassi et al., “Resource Management and Dimensioning in ATM Networks”, IEEE Network Magazine, pp. 8-17, 1990. |
| G. Gallassi et al., “Bandwidth Assignment in Prioritized ATM Networks”, IEEE, pp. 852-856, 1990. |
| Jay Hyman et al., “Joint Scheduling and Admission Control for ATS-based Switching Nodes”, pp. 223-234, 1992. |
| L. Kleinrock, “Queueing Systems, vol. 2: Computer Applications”—Computer Time-Sharing and Multiaccess Systems, John-Wiley & Sons, pp. 156-269, 1976. |
| R. Sedgewick, “Algorithms in C, 3rd Edition, Part 1-4: Fundamentals, Data Structures, Sorting and Searching”—“Quicksort” pp. 303-333, Addison-Wesley, Aug. 2001. |
| Number | Date | Country | |
|---|---|---|---|
| 20060009232 A1 | Jan 2006 | US |
