WO2011124235A1 - Method and system for handling queues in communication networks, corresponding computer program product - Google Patents
Method and system for handling queues in communication networks, corresponding computer program product Download PDFInfo
- Publication number
- WO2011124235A1 WO2011124235A1 PCT/EP2010/002160 EP2010002160W WO2011124235A1 WO 2011124235 A1 WO2011124235 A1 WO 2011124235A1 EP 2010002160 W EP2010002160 W EP 2010002160W WO 2011124235 A1 WO2011124235 A1 WO 2011124235A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- point
- nodeb
- information
- queue length
- queue
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/50—Queue scheduling
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W8/00—Network data management
- H04W8/02—Processing of mobility data, e.g. registration information at HLR [Home Location Register] or VLR [Visitor Location Register]; Transfer of mobility data, e.g. between HLR, VLR or external networks
- H04W8/04—Registration at HLR or HSS [Home Subscriber Server]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/25—Flow control; Congestion control with rate being modified by the source upon detecting a change of network conditions
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/30—Flow control; Congestion control in combination with information about buffer occupancy at either end or at transit nodes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/35—Flow control; Congestion control by embedding flow control information in regular packets, e.g. piggybacking
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/0278—Traffic management, e.g. flow control or congestion control using buffer status reports
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/52—Allocation or scheduling criteria for wireless resources based on load
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/20—Control channels or signalling for resource management
- H04W72/21—Control channels or signalling for resource management in the uplink direction of a wireless link, i.e. towards the network
Definitions
- the disclosure relates to techniques for handling queues in communication networks.
- the disclosure was devised with attention paid to its possible application in scheduling real-time flows in uplink packet access cellular networks, such as UMTS High- Speed Uplink Packet Access (HSUPA) and Long Term Evolution (LTE) .
- HSUPA High- Speed Uplink Packet Access
- LTE Long Term Evolution
- Cellular communication networks e.g. 3G+ cellular networks, such as HSUPA and LTE (Long Term Evolution of UMTS)
- 3G+ cellular networks such as HSUPA and LTE (Long Term Evolution of UMTS)
- HSUPA and LTE Long Term Evolution of UMTS
- Real-time services have a bounded end-to-end delay requirement, which, to be properly enforced, leads to the delay at each scheduling hop to be bounded as well.
- the end-to-end path of each packet includes two end (access) segments and one core network.
- Core networks are generally overprovisioned and/or make use of standardized QoS- enabling architectures, such as Differentiated Services (DiffServ) and/or Multiprotocol Label Switching-based Traffic Engineering (MPLS-TE) , which are well-tuned and normally effective in providing QoS assurance; in that context, access segments plays a key role in enabling reliable real-time services.
- DiffServ Differentiated Services
- MPLS-TE Multiprotocol Label Switching-based Traffic Engineering
- Document EP-A-1 885 090 discloses a scheduling procedure to be implemented in MAC for IEEE 802.16d/e (WiMAX) or similar PMP (Point to Multipoint) telecommunications networks.
- the procedure involves processing at every radio frame only the connections previously planned to be serviced in that frame and in determining for every serviced connection the most suitable frame for next scheduling.
- An array of tables is used to store the connection identifiers of the connections to be serviced at the current and next radio frames up to a planned time horizon.
- the connections inserted in the current frame tables are processed, the bandwidth is allocated in variable length blocks, and the frame of next servicing is evaluated according to the executed allocation and to the traffic rate parameters.
- the bandwidth allocation as well as the calculation of the next scheduling frame is performed based on the scheduling type of every service connection.
- an estimation is performed of the input queue length of a Terminal Station (TS) . This is updated at the reception of every bandwidth request by adding the value of incremental requests or by setting the new value in case of aggregate requests; at the reception of an uplink PDU the TS queue estimate is decremented by the number of bytes actually used, and the queue estimate is reset when a burst completely filled in by padding is received, as a burst not used for traffic nor for management data is interpreted as an aggregate request set to 0.
- TS Terminal Station
- the scheduling procedure for telecommunications networks discussed in the foregoing exploits a queue estimation framework to perform uplink scheduling by resorting to a scheduling procedure that processes for every radio frame only the connections previously planned to be serviced in that frame .
- the object of the invention is thus to provide an improved solution enabling i.a. a more refined queue estimation, with the possibility of proactive scheduling.
- the invention also relates to a corresponding system (e.g. a node in network) and to corresponding computer program product, loadable in the memory of at least one computer and including software code portions for performing the steps of the method of the invention when the product is run on a computer.
- a corresponding system e.g. a node in network
- computer program product loadable in the memory of at least one computer and including software code portions for performing the steps of the method of the invention when the product is run on a computer.
- reference to such a computer program product is intended to be equivalent to reference to a computer- readable medium containing instructions for controlling a computer system to coordinate the performance of the method of the invention.
- Reference to "at least one computer” is evidently intended to highlight the possibility for the present invention to be implemented in a distributed/ modular fashion.
- Various embodiments provide a queue estimation procedure that can be exploited as a framework by the uplink scheduler in order to fully exploit the channel conditions experienced by the user.
- Various embodiments provide for estimating both the size of the queue and the PDU generation time, and not merely the queue length.
- the resulting output is a complete data structure, and not just a number.
- Various embodiments are based on reconstructing the queues of each UE at the NodeB, so as to know the size and generation time of their packets and enable a scheduler to use such information. While various embodiments are amenable to all types of traffic, various embodiments are adapted for handling periodic traffic (e.g., VoIP) .
- periodic traffic e.g., VoIP
- Various embodiments consider the uplink access segment of a cellular network, such as e.g. HSUPA.
- scheduling is coordinated by a centralized entity, called NodeB.
- UEs User Equipments
- TTI Transmission Time Interval
- SG Serving Grants
- the SGs may be computed according to several performance criteria, e.g. so as to limit the received power at the NodeB, so that the latter is able to decode simultaneous transmissions from several UEs .
- Data Packets which may be referred to as Protocol Data Units (PDUs) , are buffered in the UE queues length and are transmitted across the radio interface .
- PDUs Protocol Data Units
- the NodeB obtains the backlog state of the UE queues from the Scheduling Information (SI) , which are either piggybacked in the PDUs sent by the UE (if any) or in a stand-alone transmission (otherwise) , whereby the SI are quantized according to a table, so that, as already noted in the foregoing, reconstructing therefrom the exact backlog state may not be always possible.
- SI Scheduling Information
- a "clever" scheduling solution computes the SG based on the SI reported by the UEs, thus dispensing with the risk of issuing overlength SGs and undesirably wasting resources and reducing the number of UEs that can be served in a TTI .
- Various embodiments take into account the fact that, when real-time traffic is considered, in addition to knowing how many PDUs are included in each queue and the relevant size, it may also be interesting to know the deadline for transmitting each of them, which cannot be inferred from the SI alone.
- a NodeB can reconstruct a sufficiently accurate estimate of the backlog state of UE queues taking the SI as an input, thus providing a reconstructed estimate (which may be termed a Virtual Queue or VQ) , which is represented by a list of couples ⁇ [lo,hi], d ⁇ , where lo, hi are the lower and upper bounds for the number of PDUs in the UE queue generated by time d.
- VQ Virtual Queue
- Various embodiments may co-operate with any scheduling procedure that takes into account the above information (e.g., those used for real-time scheduling in the downlink direction, where queues are physically placed at the NodeB to be adapted for working with HSUPA via minor modifications) .
- the Virtual Queue or VQ procedure does not make assumptions as to the actual traffic generation profile, thus being amenable to any kind of traffic, including non real-time traffic (e.g., HTTP), where a scheduler may harvest a smaller benefit by knowing the size and generation time of all the queued packets.
- non real-time traffic e.g., HTTP
- voice flows have a periodic packet generation behavior: they alternate "on" periods (talkspurts) , when they generate fixed-length packets with a constant interarrival time (e.g., 20s for the AMR codec) , and - if they have Voice Activity Detection (VAD) , as it normally happens - "off” or silence periods, when they either do not generate packets at all, or generate smaller packets.
- talkspurts when they generate fixed-length packets with a constant interarrival time (e.g., 20s for the AMR codec)
- VAD Voice Activity Detection
- the voice codec is made known at the RNC at the setup of a VoIP flow, for example in the Source Statistic Descriptor field of the RAB Assignment message sent to the RNC (Radio Network Controller) when the PDP Context is activated.
- the RNC can therefore provide the NodeB with the above information, and the latter can be used to identify the flow characteristics (i.e., the flow period and packet size, whether it has VAD or not, whether it generates packets during off periods, etc.).
- the VQ procedure can thus be refined to better match the flow characteristics, resulting in a reduced error in the reconstructed VQ.
- such signaling delay can be avoided by predicting the backlog status of the UE queue at the time a possible SG will actually be used, i.e. in the subsequent TTI. In that way, SG can be issued proactively, i.e. based on the predicted status of the UE queue, before the SI is actually conveyed to the NodeB. This is achieved by computing the SG on a projection of the state of the VQ at the time when the SG will actually be exploited by the related UE.
- proactive scheduling can coexist with standard (reactive) scheduling of non-periodic or poorly predictable uplink flows, and can be turned on and off at will on the same flow, depending on whether a reliable estimate of the packet generation instants is available or not.
- Virtual Queue techniques can also be applied to other technologies.
- LTE Long Term Evolution of UMTS
- scheduling of uplink resources is done by eNodeB.
- the latter assigns certain time/frequency resources to the UEs and informs UEs about the transmission formats to use.
- scheduling decisions may be based on QoS parameters, UE buffer status, uplink channel quality measurements.
- the UE buffer status reported to the eNodeB is a coarsened version of the real UE buffer occupancy, and a quantization table, similar to the one employed in HSPA, is used for encoding the UE buffers occupancies .
- the possibility of reconstructing the UEs buffers status at the NodeB level in HSPA can be easily ported, with minor adjustments, into the LTE eNodeB MAC layer, in order to enable deadline-based scheduler to be executed also in the uplink side of the LTE architecture.
- virtual queueing can be applied to any technology that satisfies the following requirements :
- the status of uplink buffers is reported by conveying to the NodeB either an exact or an approximate buffer length, measured in number of PDUs. Reporting can be either periodical or non periodical;
- the receiver decides scheduling grants for the senders.
- an improved estimate of the queue length may be obtained by periodically increasing the queue length by a fixed amount and subtracting the number of detected information units transmitted from the second point (e.g. the User Equipment or UE) to the first point (e.g. the base station), in case the arrival process for the information units at the second point is known to be periodic; the estimated queue length is compared to the quantized queue length information reported by the second point (e.g. the UE) , and any mismatch in the above comparison is used to infer the onset and the termination of periodic packet generation at the second point.
- the second point e.g. the User Equipment or UE
- the first point e.g. the base station
- FIG. 1 is representative of SI and SG timing in an embodiment
- FIG. 2 is a block diagram illustrative of the scheduling framework in an embodiment
- the UMTS radio access network includes three major network elements: the Radio Network Controller (RNC) , the Base Station (NodeB) and the User Equipment (UE) .
- RNC Radio Network Controller
- NodeB Base Station
- UE User Equipment
- a RNC controls one or more NodeBs, the latter being in charge of a single cell where UEs roam.
- the NodeB coordinates scheduling decisions using TTIs of 2 or 10ms.
- the Protocol Data Units or PDUs are physically stored at the UEs.
- the NodeB obtains the backlog state of the UE queues from the Scheduling Information (SI) , which includes the total backlog on the UE and the percentage of the latter accounted to the highest priority flow (i.e., the one which will be selected by the internal UE scheduler) .
- SI Scheduling Information
- Both are quantized, using five and four bit indexes respectively, according to two non- linear tables (see the 3 rd Generation Partnership Project website already captioned in the foregoing) so that reconstructing the exact backlog state may not be possible.
- the UE always reports an overestimated backlog to the NodeB.
- the NodeB allocates data transmission power to the various UEs by assigning them scheduling grants (SGs) . Depending on the received SG, the UE selects the appropriate coding, which also determines the employed TBS.
- SGs Schestrial Grant
- TBSs Transmission Block Sizes or Sets
- E-DCH Transport Format Combination (ETFC) table which is known to both the NodeB and the UEs .
- a null SG called the zero grant, prevents the UE from transmitting any data.
- UE User Equipment
- i has one real-time flow, which is internally scheduled at the highest priority.
- the QoS class of the flow is known to the NodeB from the setup negotiation.
- Some other information, hereinafter called flow information, is normally available at least at the RNC (which can, in turn, communicate them to the NodeB) .
- this information may be assumed to include the voice codec employed, one out of a finite choice of known possibilities (which is what normally happens) .
- Table 1 A standard SI quantization table (Table 1) is reported below, both in bytes (left) and in 40 -byte PDUs (right) . In Table 1 herein, those SI values which cannot be reported in this case are shown in grey.
- the table reports in its left hand side a dimension in bytes, whereas the queues actually contain a number of fixed- length PDUs, whose length is selected by the NodeB.
- the PDU length can be set arbitrarily.
- a common value is 40 bytes, which is such that a voice packet is entirely contained into a PDU.
- the number of 40 -byte PDUs associated to each SI value is shown on the right side of Table 1. For instance, the UE reports a value of 18 if the queue is buffering 24 to 31 PDUs.
- the quantization table is divided in two zones: a non ambiguous zone, consisting in the set of SI that allows one to infer an exact backlog at the UE (i.e., 0 ⁇ S/ ⁇ 12 f or 40-byte PDUs) and an ambiguous one, where the quantization intervals are larger than the PDU size.
- a non ambiguous zone consisting in the set of SI that allows one to infer an exact backlog at the UE (i.e., 0 ⁇ S/ ⁇ 12 f or 40-byte PDUs)
- an ambiguous one where the quantization intervals are larger than the PDU size.
- the width of the ambiguous region depends on the PDU size. The larger the PDU size, the smaller the ambiguous region, and, within the latter, the width of each quantization interval
- the PDUs will be assumed to be 40 bytes long. This may be regarded as a "pejorative" assumption in terms of quantization. However, various embodiments do not rely on the PDU size, although their performance - and effectiveness - may be related thereto.
- a UE application generates packets, so that, at some time instants, a number of PDUs is inserted in the (FIFO) UE queue.
- the UE receives a SG from the NodeB, it transmits one MAC-E PDU, which includes zero or more PDUs and/or SI, in a single TTI .
- Those PDUs may belong to different packets, as original packets are reassembled at the NodeB. This means that the scheduling module in the latter may be made aware of packet boundaries when it receives a set of PDUs. However, no such assumption is made here.
- Packets queued at the UE can be removed either by scheduling decisions taken at the NodeB, or dropped by the UE after a known dropping timeout. Without loss of generality, one may assume that one packet is generated in a TTI. In fact, the UE is not required to transmit packets atomically, and it can mix PDU of several packets (however maintaining the FIFO ordering) in the same transmission. Furthermore, it may not be possible to pin down events with a smaller resolution than the TTI, from a real-time scheduling point of view packets of a flow generated in the same TTI would have the same deadline nonetheless.
- times will be normalized to the TTI length, so that the sequence of TTIs is a sequence of natural numbers .
- the SI issued at a given TTI T at the UE reports the state of its queue up to T. However, this SI arrives at the NodeB at time T+l. Assuming that the NodeB is able to compute a scheduling decision in zero time, the NodeB might be able to keep into account that SI when issuing the subsequent SG. The latter in turn arrives at the UE at time T+2. Therefore, in the very best case, the signaling delay will equal 2 TTI.
- Figure 1 herein is exemplary of SI and SG timing in communication between a NodeB and a User Equipment or UE.
- a 10ms TTI deployment such a signaling delay is indeed attainable, since the SG is actually 2ms long, and they are repeated five times in the 10ms period.
- the NodeB might take, e.g., 2ms for computing the SGs for the UEs in the cell, leave the first 2ms instance blank, and actually repeat the correct SG four times.
- the UEs would therefore be able to decode the correct SG simply by skipping the first instance.
- the LTE system allows a similar mechanism, with enhanced flexibility.
- the RRC Radio Resource Control
- the non-scheduled mode can be used to serve a MAC-d flow of a UE. In this case, the latter is served periodically (for up to a configurable maximum number of periods) , and is given a fixed grant on a given set of TTIs within a period.
- a non-scheduled flow can transmit up to a pre-defined number of bits, whether its queue is empty or not, whether the related PDUs are urgent or not, whether the channel conditions for the UE are optimal or largely suboptimal. Furthermore, queue length and delays are not taken into account, as SI are not considered. This makes such a mechanism inefficient when used on a large scale (i.e., for many users simultaneously), or for mobile users with varying channels. In HSUPA, additional mechanisms would also be required to keep the intra-cell interference generated by non- scheduled users under control.
- VQ in all its versions, included those specialized for CBR (Constant Bit Rate) service
- CBR Constant Bit Rate
- VoIP applications represent an important class of real-time traffic in today's HSPA systems. As already anticipated, a specialized VQ procedure will be presented for periodic traffic.
- the traffic profiles of VoIP applications can be classified into four categories:
- VAD Voice Activity Detection
- VBR Very Bit Rate
- This exemplary disclosure describes specialized VQ procedures for the first two categories; extensions for the third one are straightforward, especially if the scenario is limited to a communication where the Quasi-CBR has constant size packets at constant intervals, behaving basically like a CBR on/off.
- packet generation periods are normally larger than the TTI (whichever the version, 2ms or 10ms) .
- voice streams using the AMR codec are CBR (with on/off alternation due to Voice Activity Detection), with frame generation periods of 20ms (i.e. 2 or 10 TTI, depending on the HSUPA version) .
- the periods considered are not necessarily integer multiples of a TTI.
- single "on” and “off” periods are normally much larger than the inter packet generation time (and, consequently, of the TTI) .
- the average length of these periods is in the order of 100-1000 TTIs, whatever the voice model adopted and the TTI length.
- quasi-CBR sources may appear to the scheduler as
- the GSM AMR codec generates 32 bytes of payload during talkspurts, and either 0 or 5 bytes of payload during silence periods. If the PDU length is above 72 bytes, then the SI reports just one PDU in all cases.
- the Virtual Queuing block collects the sequence of SI and transmitted PDUs from the UE, and uses this information to reconstruct the state of the UE queue in the most accurate possible way, using affordable computations.
- a basic procedure which does not make any assumption on the traffic profile, will be described in the following. In some cases, we might know more about the flow, and exploit this knowledge to improve the VQ estimate.
- a "flow profile" may be defined as a set of information items such as: whether the traffic is periodic, and whether the generated packets have a constant length. The above information can be collected at flow setup, through means that are known to the person skilled in the art and do not need not be described here.
- the NodeB is configured with a table which associates codecs with the following information: i) a flow type (e.g., CBR, or CBR on/off, etc.), ii) a flow period, and iii) a packet length. If a flow is known to be CBR or CBR on/off, this information can be used both to specialize the VQ procedure and to enable proactive SG assignment. As for the first issue, knowing its packet length and period allows one to overcome the uncertainties in the VQ estimate that arise due to the SI quantization. Furthermore, as it allows one to predict the size and generation instant of packets, it enables the NodeB to assign suitably large SGs proactively to periodic flows.
- a flow type e.g., CBR, or CBR on/off, etc.
- Figure 2 is representative of an exemplary scheduling framework including a Virtual Queuing arrangement VQ having SI plus TX'd PDUs as inputs together with information as to flow type, period, size(s), and so on developed by a flow profile module FP starting from flow information as provided e.g. by a codec .
- VQ Virtual Queueing
- Various embodiments pursue the goal of enabling the NodeB to reconstruct the exact backlog state on the UE, which may not be always possible due to the quantized nature of the SI.
- the estimation process depends on the PDU length, and, in general, the shorter the PDU size, the more difficult the accurate estimations of their number (see Table 1) .
- a VQ is a FIFO queue of items ⁇ [lo, hi] , d ⁇ , where d is an estimated packet generation time, and lo, hi are the lower and upper bounds for the overall VQ backlog including all packets generated until time d. How those time instants d are computed will be described in the following.
- the length associated to the head of the VQ still describes the head-of-line packet, which in fact enables deadline-based scheduling.
- the VQ estimation works as follows. Every time a set of k PDUs is transmitted by the UE to the NodeB, the number k is subtracted from the upper and lower bounds of every entry in the VQ, and when the upper bound of the head-of-line entry reaches zero the latter is removed from the front .
- the new information is used to either improve the length estimate for the tail of the VQ, or to detect the generation of a new packet at the UE.
- No other packet is assumed to be generated at the UE after time 0 for a while, i being a generic instant.
- o 3 ⁇ 4lL 'O ⁇ ' are the most accurate bounds on the length of the queue available at time 1 given the above information. Now, in the absence of a new packet arrival, O 3 ⁇ 4L O" are increasing and decreasing respectively, thus narrowing down the uncertainty for .
- both 0 L 'O 3 ⁇ 4 " should be recomputed and the corresponding VQ entry updated.
- ⁇ may increase if another packet is generated at the UE at some time ⁇ , then might possibly take place for some n ⁇ k f because ⁇ ' may increase .
- the UE transmits four PDUs.
- VQ includes •v entries lL J (a superscript is used for the entries in the VQ and a subscript for time instants) : every time a set of ⁇ PDUs are transmitted from the UE to the NodeB, the following actions take lace:
- the head-of-line entry has to be removed. This implies that one may be considering more PDUs than necessary as being generated at a given time instant, which may be avoided by using additional information. If, e.g., the NodeB sees packet boundaries and one can be reasonably sure that only one packet is generated at a TTI, this last part of the procedure can be refined according to known methods .
- the last entry in the VQ (i.e., the N th ) , is instead updated through (3) . If updating each entry of the VQ at every transmission is too costly, a tradeoff between accuracy and speed can be achieved as follows :
- each VQ entry is augmented with a counter " .
- an improved estimate of the queue length may be obtained by periodically increasing the queue length by a fixed amount and subtracting the number of detected information units transmitted from the UE to the base station.
- the estimated queue length may be compared to the quantized queue length information reported by the UE, and any mismatch in the above comparison is used to infer the onset and the termination of periodic packet generation at the UE.
- the VQ procedure described in the foregoing can however be employed with any kind of traffic, as it does not rely on additional information on the flow profile (e.g., whether the traffic is periodic or not) .
- the VQ procedure can be specialized .
- CBR flows can be considered as a subcase of CBR on/off flows, with an arbitrarily long "on" period.
- the quasi-CBR case can be readily derived through straightforward modifications starting from the CBR on/off case.
- the size of a packet is constant, on and off periods are one order of magnitude larger than the packet generation period.
- the VQ procedure can be specialized as follows.
- the packet period is equal to x TTI, where x is an integer (this hypothesis is made solely for the sake of simplicity, and is in no way mandatory) . Then: - within a talkspurt, every x TTI insert a constant length packet in the VQ;
- the TTI when a packet is generated, from which all the packet generation instants can be estimated, can be detected, for instance, if one receives two consecutive SI which reveal a packet generation. For instance, assuming a
- a condition used to detect the generation of a packet may be an increase in the reported SI . This means that one may implicitly assume that, during a silence period, the UE queue can be drained sufficiently, so that there is no ambiguity in the detection of the first packet. Such assumptions are reasonable, since the duration of a silence period is considerably larger than the packet period and the ambiguous region starts when there are several (e.g., more than 5) packets in the queue. Similarly, the onset of a silence period is detected through mismatch between the SI and the VQ length. Again, this condition might be detected with some delay if the UE queue is in the ambiguous region when this happens. However, late detection of a silence period is not particularly harmful, since the VQ is rapidly reset to the correct length, and virtual packets entered by mistake have little, if any, chance of being seen by the scheduler.
- the signaling delay associated to SI reporting and SG scheduling amounts to at least two TTIs, assuming that at time T* +l the NodeB can issue a SG which takes into account the SI sent by the UE at time T* .
- the latter is unavoidable if reactive SG scheduling is used, even if an UE queue is always emptied right after each non-zero SI report.
- the TTI duration in HSUPA can be as high as 10ms, 20ms of added delay in the uplink direction are not negligible, especially with voice applications .
- a delay can be removed by employing a proactive SG assignment scheme, by exploiting the VQ procedure and the prediction of the time instants at which packets are generated.
- SG can be scheduled based on the presumed backlog state of the UE at the time when the SG will actually be used. For instance, if the NodeB knows that voice packets are generated each 20 ms (i.e., every other TTI) starting from time 0, it can schedule a SG large enough to hold two packets at time 2, and periodically schedule another SG for one packet, as shown in Figure 4, which is representative of a high level example of proactive scheduling.
- proactive scheduling would yield a benefit in terms of delay for each packet in a talkspurt except the first one.
- the beginning of a talkspurt can only be detected by looking at the SI (i.e., reactively) , so that the first packet actually has a higher delay (3.5 TTI on average) than the rest.
- the onset of a silence period can only be detected reactively, i.e. through a mismatch between the VQ and the SI, an SG will be wasted at the end of each talkspurt (unless the UE has lower priority traffic to send) .
- Proactive scheduling may be enforced once the packet generation pattern of the flow is known. For that reason, each flow may be associated to a proactive Boolean variable, which is initially false and is set to true whenever the NodeB has achieved enough information on the packet generation pattern. It may then become false again if packet generation becomes too jittery (possibly because of operating system overhead at the sending application or due to clock skews) or unknown (e.g., at the beginning of a silence period for a CBR on/off voice flow) .
- proactive and reactive grant scheduling can be mixed for the various flows at a given time.
- H-ARQ process has its turn. In that case, the NodeB will simply skips the UE for allocating grants.
- Proactive scheduling can be employed with CBR and CBR on/off flows, provided that the correct VQ procedure is used for each flow.
- Various embodiments enable a scheduler to perform scheduling of real-time uplink flows by reconstructing, through a Virtual Queueing technique, the status of the uplink buffers.
- the outcome of such a technique is that the scheduler possesses a virtualized version of the uplink flow buffer, where virtual packets are associated to their estimated size and generation time.
- Various embodiments will thus enable a scheduler to make real-time decisions (e.g., schedule PDUs by deadline), which could not be possible otherwise.
- Virtual Queueing techniques considered herein can be applied to any system (e.g., cellular ones) where, e.g.: - packets are segmented into a number of fixed- length PDUs, whose length is known at both the sender and the receiver, such a length not having to be the same for all senders ;
- the receiver decides scheduling grants for the senders
- HSUPA High Speed Uplink Packet Access
- LTE Long Term Evolution
- Various embodiments concern a standard Virtual Queueing technique, which does not make any assumptions on the packet generation process, thus being suitable for any kind of real-time traffic (whether CBR or not) .
- Various embodiments concern a specialized VQ version for CBR (or CBR on/off, or Quasi-CBR) traffic, which capitalizes on knowing the packet size and period of a connection.
- CBR CBR on/off, or Quasi-CBR
- a proactive version of the scheduling process can be adopted at the scheduler.
- the scheduler computes its scheduling grants based not only on the (state) information arrived at previous reporting intervals, but on the predicted state of the queue at the moment the grant will be actually used.
- VQ virtual resource
- This disclosure demonstrates the effectiveness of various embodiments of VQ used in conjunction with e.g. a HY-CART scheduler, originally designed for HSDPA connections .
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Databases & Information Systems (AREA)
- Mobile Radio Communication Systems (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
Description
Claims
Priority Applications (6)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP10715685.3A EP2556714B1 (en) | 2010-04-06 | 2010-04-06 | Method and node for handling queues in communication networks, and corresponding computer program product |
JP2013503003A JP5576553B2 (en) | 2010-04-06 | 2010-04-06 | Method and system for handling queues in a communication network and corresponding computer program product |
CN201080067050.1A CN102918911B (en) | 2010-04-06 | 2010-04-06 | The method and system of the queue in process communication network, corresponding computer program |
KR1020127028903A KR101648835B1 (en) | 2010-04-06 | 2010-04-06 | Method and system for handling queues in communication networks, corresponding computer program product |
US13/639,373 US8767760B2 (en) | 2010-04-06 | 2010-04-06 | Method and system for handling queues in communication networks, corresponding computer program product |
PCT/EP2010/002160 WO2011124235A1 (en) | 2010-04-06 | 2010-04-06 | Method and system for handling queues in communication networks, corresponding computer program product |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/EP2010/002160 WO2011124235A1 (en) | 2010-04-06 | 2010-04-06 | Method and system for handling queues in communication networks, corresponding computer program product |
Publications (2)
Publication Number | Publication Date |
---|---|
WO2011124235A1 true WO2011124235A1 (en) | 2011-10-13 |
WO2011124235A8 WO2011124235A8 (en) | 2011-12-15 |
Family
ID=43127136
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/EP2010/002160 WO2011124235A1 (en) | 2010-04-06 | 2010-04-06 | Method and system for handling queues in communication networks, corresponding computer program product |
Country Status (6)
Country | Link |
---|---|
US (1) | US8767760B2 (en) |
EP (1) | EP2556714B1 (en) |
JP (1) | JP5576553B2 (en) |
KR (1) | KR101648835B1 (en) |
CN (1) | CN102918911B (en) |
WO (1) | WO2011124235A1 (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011139058A2 (en) * | 2010-05-04 | 2011-11-10 | Samsung Electronics Co., Ltd. | Method of control indication in multi-input multi-output communication systems |
GB201111106D0 (en) * | 2011-06-30 | 2011-08-10 | Xelerated Ab | Method, network device, computer program and computer program product for communication queue state |
US9407508B2 (en) * | 2013-09-04 | 2016-08-02 | Tube, Inc. | Client-side inference of wireless network states |
US11716737B2 (en) * | 2018-05-08 | 2023-08-01 | Commscope Technologies Llc | Gratuitous PUSCH grants during LTE RRC connection and NAS attach procedures |
CN112042254A (en) | 2018-05-08 | 2020-12-04 | 康普技术有限责任公司 | Proactive PUSCH grant to prevent rate throttling |
CN109996289A (en) * | 2019-04-03 | 2019-07-09 | 成都中科微信息技术研究院有限公司 | A kind of method for distributing uplink resource in smart grid |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1885090A1 (en) | 2006-08-01 | 2008-02-06 | Siemens S.p.A. | Time horizon scheduling method for transmitting traffic data bursts with variable length in PMP telecommunications networks |
WO2009080081A1 (en) * | 2007-12-21 | 2009-07-02 | Telecom Italia S.P.A. | A scheduling method and system for communication |
EP2117186A1 (en) * | 2006-12-31 | 2009-11-11 | ZTE Corporation | Scheduling method and system for hsupa |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1604687A (en) * | 2003-08-16 | 2005-04-06 | 三星电子株式会社 | Method and apparatus for assigning scheduling for uplink packet transmission |
JP5007329B2 (en) * | 2006-03-29 | 2012-08-22 | リサーチ イン モーション リミテッド | Communication mode control method, mobile communication system, and mobile communication terminal |
US20110103395A1 (en) * | 2009-11-03 | 2011-05-05 | Qualcomm Incorporated | Computing the burst size for a high speed packet data networks with multiple queues |
US8923270B2 (en) * | 2011-10-04 | 2014-12-30 | The Chinese University Of Hong Kong | Method for link buffer size and queue length estimation for bandwidth-varying mobile data networks |
-
2010
- 2010-04-06 JP JP2013503003A patent/JP5576553B2/en active Active
- 2010-04-06 EP EP10715685.3A patent/EP2556714B1/en active Active
- 2010-04-06 CN CN201080067050.1A patent/CN102918911B/en active Active
- 2010-04-06 US US13/639,373 patent/US8767760B2/en active Active
- 2010-04-06 KR KR1020127028903A patent/KR101648835B1/en active IP Right Grant
- 2010-04-06 WO PCT/EP2010/002160 patent/WO2011124235A1/en active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1885090A1 (en) | 2006-08-01 | 2008-02-06 | Siemens S.p.A. | Time horizon scheduling method for transmitting traffic data bursts with variable length in PMP telecommunications networks |
EP2117186A1 (en) * | 2006-12-31 | 2009-11-11 | ZTE Corporation | Scheduling method and system for hsupa |
WO2009080081A1 (en) * | 2007-12-21 | 2009-07-02 | Telecom Italia S.P.A. | A scheduling method and system for communication |
Non-Patent Citations (4)
Title |
---|
B. WANG ET AL.: "Performance of VOIP on HSDPA", PROC. OF IEEE VTC SPRING, 5 June 2005 (2005-06-05) |
M. ANDREOZZI ET AL.: "Flexible Scheduling for Real-Time Services in High-Speed Packet Access Cellular Networks", EUROPEAN WIRELESS 2009, 17 May 2009 (2009-05-17) |
MATTEO MARIA ANDREOZZI ET AL: "Flexible scheduling for Real-Time services in High-Speed Packet Access cellular networks", WIRELESS CONFERENCE, 2009. EW 2009. EUROPEAN, IEEE, PISCATAWAY, NJ, USA, 17 May 2009 (2009-05-17), pages 189 - 193, XP031583298, ISBN: 978-1-4244-5935-3 * |
QINLONG QIU ET AL: "Hierarchy Virtual Queue Based Flow Control in LTE/SAE", FUTURE NETWORKS, 2010. ICFN '10. SECOND INTERNATIONAL CONFERENCE ON, IEEE, PISCATAWAY, NJ, USA, 22 January 2010 (2010-01-22), pages 78 - 82, XP031648880, ISBN: 978-0-7695-3940-9 * |
Also Published As
Publication number | Publication date |
---|---|
WO2011124235A8 (en) | 2011-12-15 |
CN102918911A (en) | 2013-02-06 |
JP2013528977A (en) | 2013-07-11 |
CN102918911B (en) | 2015-12-16 |
EP2556714B1 (en) | 2014-01-08 |
US8767760B2 (en) | 2014-07-01 |
KR20130045860A (en) | 2013-05-06 |
JP5576553B2 (en) | 2014-08-20 |
US20130028220A1 (en) | 2013-01-31 |
KR101648835B1 (en) | 2016-08-17 |
EP2556714A1 (en) | 2013-02-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6879561B1 (en) | Method and system for wireless packet scheduling with per packet QoS support and link adaptation | |
CA2656859C (en) | Compressed delay packet transmission scheduling | |
EP2474192B1 (en) | Methods and arrangements for allocating scheduling request resources in a wireless communication system | |
US9019851B2 (en) | Network node and method related to buffer estimation | |
EP2103054B1 (en) | Signature sequences and methods for time-frequency selective channel | |
RU2510598C2 (en) | Method and device in wireless communication system | |
EP2556714B1 (en) | Method and node for handling queues in communication networks, and corresponding computer program product | |
EP2926611B1 (en) | Transmitting radio node and method therein for scheduling service data flows | |
KR20100023034A (en) | Buffer status reporting | |
EP1376948A1 (en) | Quality of service scheduling for packet switched data services | |
US20090103438A1 (en) | Grant Based Adaptive Media Access Control Scheduling | |
Irazabal et al. | Preventing RLC buffer sojourn delays in 5G | |
US8130713B2 (en) | System and method for credit-based channel transmission scheduling (CBCTS) | |
TWI513347B (en) | A method for communicating in a network | |
Afrin et al. | A delay sensitive LTE uplink packet scheduler for M2M traffic | |
US8355403B2 (en) | Stale data removal using latency count in a WiMAX scheduler | |
CN101848494B (en) | Dispatching method of real-time business and device thereof | |
CN111510953B (en) | Data processing method, terminal and storage medium | |
EP3790203A1 (en) | Device and method for application-requirement aware medium access control | |
KR100934990B1 (en) | METHOD AND APPARATUS FOR A REAL TIME SCHEDULING A WiBro MOBILE TERMINAL ADAPTED TO SCHEDULING OF A WiBro RADIO ACCESS STATION IN WiBro/Mobile WiMAX | |
CN116367322A (en) | Wireless resource scheduling method and device and working machine | |
CN118140521A (en) | RAN enhancement taking into account CBS behavior in TSCs | |
CN118200243A (en) | RB resource scheduling method and device | |
JP2005094761A (en) | Differentiated scheduling method in downlink communication between rnc and node-bs in network | |
CN111385883A (en) | Data transmission method and device, storage medium and terminal |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201080067050.1 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 10715685 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13639373 Country of ref document: US Ref document number: 2010715685 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2013503003 Country of ref document: JP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 20127028903 Country of ref document: KR Kind code of ref document: A |