WO2009141012A1

WO2009141012A1 - Packet latency estimation

Info

Publication number: WO2009141012A1
Application number: PCT/EP2008/056367
Authority: WO
Inventors: Roland Carlsson; Aare MÄLLO
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2008-05-23
Filing date: 2008-05-23
Publication date: 2009-11-26
Anticipated expiration: 2010-11-23
Also published as: US8488485B2; EP2291955A1; US20110075588A1; ATE536025T1; EP2291955B1

Abstract

Intermediary node (14) adapted for receiving a sequence of a packets from a server (10) the intermediary node comprising a time estimator (103) adapted for performing the following steps - resolving the sequence number (k) and time of arrival to the intermediary node for at least a plurality (q) of incoming packets (21); - establishing the frame period of the incoming packets (T)(22); - establishing (23) a sequence of normalized packet arrival times (e_k) as corresponding to the sensed frame period (T); - calculating (24) the relative arrival time rtr_{ec, k} of the plurality (q) of incoming packets in relation to the normalized packet arrival times; - creating (25) a cumulative density function (CDF) for a given sequence of packets; - establishing (26) the threshold value for the relative arrival time rt_PL yielding the predefined packet loss (PL) based on the cumulative density function (CDF).

Description

PACKET LATENCY ESTIMATION

Technical field

This invention pertains to inelastic services form a server that transmits data in a frame- wise fashion (e.g. VoIP) at a fixed frequency, at least during given periods of time, over one or more nodes, in which least one node experiences a time varying channel, e.g. radio. In particular, the invention relates to a scheduler and a method therefore in a wireless base station.

Background

Some widely used Internet services are inelastic such as VoIP (Voice over Internet Protocol), that is, data packets are generated frame-wise at a predominantly fixed fre- quency. Such services often require a small roundtrip time (RTT) in order to work well.

In mobile networks (e.g. IEEE 802.16 and 3rd Generation Partnership Project (3GPP)), a scheduler in the base station distributes the transmission capacity "fairly" between different data transmission flows or at least in such a manner that given needs of given occur- ring user entities can be served in accordance with agreed service level requirements. At the same time, the scheduler typically aims to utilize the transmission capacity in a way so that the "total transmissions" in the network can be maximized. This maximization of total transmissions typically involves delaying individual transmissions to suitable transmissions instances. Hence, scheduling involves that the available bandwidth is utilized in an optimized way such that a majority of users of the system are served according to agreed quality of service requirements.

A low latency requirement for a given traffic flow, takes up a high "transmission capacity", since the flexibility for scheduling concurrent traffic flows is diminished. For wireless systems, where channel conditions typically vary substantially over time, e.g. due to the changing location for mobile stations, a large allowable delay, i.e. the time from a packet is received from a higher order equipment until the packet is transmitted to the mobile station, is preferred. A VoIP service is often routed over several nodes where some nodes may use wireless transmissions. In order to ensure reasonable round trip times for e.g. VoIP services, the delay for each individual VoIP packet is typically set to a predefined maximum allowable value in each node.

The low round trip time (RTT) requirements for VoIP services imply that transmission capacity must be dedicated for such wireless transmissions. Since a VoIP session may be routed over several nodes, each node must comply with a predefined allowable maximum latency in order to guarantee a low total RTT.

In many prior art systems, latency requirements imply a negative toll on the total capacity of individual nodes and of the system in which the nodes are comprised.

Summary

It is a first object of the invention to increase the capacity in systems in which periodic transmissions are offered over at least a time varying channel.

This object has been solved by providing a intermediary node (14) adapted for receiving a sequence of a packets from a server (10) and re-transmitting the sequence of packets over a time variant channel to a receiving entity (15), wherein a given predefined latency L_n pertain to the intermediary node and wherein a predefined packet loss PL pertains to a sequence of packets transmitted from the server over a path of a series of arbitrary nodes () to the intermediary node (14) and further on to the receiving entity (15). The intermediary node comprises a time estimator (103) adapted for performing the following steps

- resolving the sequence number (k) and time of arrival to the intermediary node for at least a plurality (q) of incoming packets; - establishing the frame period of the incoming packets (T);

- establishing a sequence of normalized packet arrival times (e_k) as corresponding to the sensed frame period (T);

- calculating the relative arrival time rtr_{ec, k} of the plurality (q) of incoming packets in relation to the normalized packet arrival times; - creating a cumulative density function (CDF) for a given sequence of packets; - establishing the threshold value for the relative arrival time rt_PL yielding the predefined packet loss (PL) based on the cumulative density function (CDF).

The above abject has also been accomplished by a method for receiving a sequence of a packets from a server (10) over an intermediary node (14) and re-transmitting the sequence of packets over a time variant channel to a receiving entity (15), wherein a given predefined latency L_n pertaining to the intermediary node and wherein a predefined packet loss PL pertains to a sequence of packets transmitted from the server over a path of a series of arbitrary nodes (1 1 , 12, 13, 16) to the intermediary node (14) and further on to the receiving entity (15); comprising the following steps

- resolving the sequence number (k) and time of arrival to the intermediary node for at least a plurality (q) of incoming packets (21);

- establishing the frame period of the incoming packets (T)(22);

- establishing (23) a sequence of normalized packet arrival times (e_k) as corresponding to the sensed frame period (T);

- calculating (24) the relative arrival time rt_{rec, k} of the plurality (q) of incoming packets in relation to the normalized packet arrival times;

- creating (25) a cumulative density function (CDF) for a given sequence of packets;

- establishing (26) the threshold value for the relative arrival time rt_PL yielding the prede- fined packet loss (PL) based on the cumulative density function (CDF).

According to the invention, it is observed that even if a first node manages to transmit faster than it's allowed maximum latency, the saved delay time can not be used in the intermediary node in order to minimize the needed transmission capacity. The reason is that the absolute reception deadline in the final receiver (e.g. the mobile receiving entity) is not known by the intermediary node.

According to the invention, it is moreover observed that the channel capacity, say between a base station and mobile station (UE), varies over time, especially when the UE is moving. When several UEs are connected to the cell, the channel capacity is unique for each UE connection. Hence, an enhanced capacity will arise for the channel at different times for the different connections.

As described above, the predefined maximum latency is typically held low in the network in order to guarantee the RTT requirement for VoIP. However, if the scheduler strives to secure a given low latency, it implies that the scheduler must transmit data to a UE irre- spective of whether the momentary channel capacity is above or below average. The scheduler may transmit the data on a poor channel requiring excessive transmission capacity.

According to the invention, it is found that it would be preferable to 'wait with the transmission' until the channel capacity to the UE is good. Such handling would lead to an increased scheduling gain. If the "allowable delay" is large, the scheduler can 'wait' to transmit until the channel conditions are sufficiently good. Hence, the required transmission capacity can be reduced if the allowable delay can be increased. By striving to maximize the allowed time delay for the wireless node, the needed transmission resources can be minimized without the total RTT requirement being exceeded.

Further advantages of the invention will appear from the following detailed description of the invention.

The invention is applicable for e.g. inelastic services that produce data in a frame wise fashion (e.g. VoIP).

Brief description of the drawings

Fig. 1 shows transmission from a server over a number of nodes and a wireless node,

fig. 2 shows an exemplary table relating to incoming packets;

fig. 3 shows an exemplary probability density function (PDF) over relative reception times in a given node,

fig. 4 shows a cumulative density function (CDF) for relative reception times of packets,

fig. 5 is a routine according to a first embodiment of the invention,

fig. 6 shows a base station according to a first preferred embodiment of the invention, and

fig. 7 shows a base station according to a second preferred embodiment of the in- vention.

Detailed description

First preferred embodiment of the invention

A data service provided via a telecom system is normally transmitted over a number of nodes before the data reach its final destination.

Figure 1 shows a communication network in which a communication route is established from a server over a number of transmission nodes, 11-13 (1..N-1), through a intermediary node 14 (N), and further to a user entity (15) over a time variant channel. The server transmits packets according to a service transmission at a generally fixed frame rate T. Due to delays in individual nodes in the network, packets are received at a time varying rate, and some packets may even be lost. The pathway may change during the trans- mission such that the route changes over time such that other nodes, e.g. node 16, are included in the path or replaces previously used nodes, but it is assumed that the route generally remains constant at least for a period of time being larger than the frame rate T. The intermediary node may utilize the time variant channel, such as to transmit at op- timized times.

For instance, the server may constitute a server on the Internet, routers may constitute the nodes, and a wireless base station, connecting to the user entity over the air, may constitute the intermediary node. The intermediary node may comprise a scheduler, by which packets to a plurality of users are scheduled for transmission. In particular, if the end users are mobile, the radio channel may experience large variations, which, according to the invention, are advantageously dealt with by scheduling the transmission of packets at times where the channel conditions are above a predetermined quality threshold.

For a VoIP (voice over IP) service, voice packets are decoded at a pace corresponding to the pace at which the voice packets were encoded. I.e. there is a fixed time distance between transmissions of packets; i.e. packets are transmitted at a predominantly fixed frequency (corresponding to the frame period), whereby the receiver "plays back" the voice packets with a fixed delay D.

The data latency for a node is typically specified to be less than a predetermined constant value. Normally, each node transmits the data with a margin to the latency requirement. In such a network, frames transmitted periodically at a frame rate e.g. T= 2, may arrive at times t, as exemplary indicated in the table in fig. 2 (rounded values). The relative arrival time, rt, relative to a normalized arrival time, e, given by the frame rate T are indicated.

According to the invention, an absolute transmission deadline per packet in at least one node in a given network may be estimated. Thereby, a longer delay can be allowed for the scheduling of each individual packet in a given system compared to if no estimation is performed. According to the invention, a higher scheduling gain may be obtained. A high scheduling gain results in a higher total data throughput, which will benefit all user entities in a given cell associated with the intermediary node. The basis for calculating the absolute transmission deadline shall be explained in the following. In order to ensure that the maximum acceptable delay is D, from server 10 to user entity 15, a predefined maximum latency is specified for each node L_n, where n is the node number so that

JV

(I) ∑L_n ≤ D w=l

and whereby N is a given number of nodes for the transmission.

For VoIP services typically, a number of packet losses are acceptable without considerable deterioration of the voice quality. This predetermined value of the acceptable loss rate is denoted PL.

Hence, for a node n, it would appear that at least a ratio of 1-PL of the received packets are received in time to be played back at the voice decoder provided that the latency in node n will be less or equal to the predefined maximum latency L_n and that all nodes 1 to N-1 fulfil their requirements as to the given predefined maximum latencies = L_n.

Each received packet is time stamped with the reception time t_rec,k where k is the packet sequence number. Let t_rec,i (i-β. the reception time for the first packet) be the reference time for the subsequent packet arrivals. The relative packet arrival time for each packet is calculated as

C) rt_{rec k} = t_recΛ - T - {k - l)- t_{rec i}

where T is the frame rate used by the voice coder and where rt_rec,i = 0, c.f the example of fig. 2.

The received relative arrival time for k individual packets rt_rec,k can then be presented in a probability density function (PDF) as shown in fig. 3. Integrating the PDF, a cumulative density function (CDF) can be found as shown in fig. 4.

Since it can be assumed that at least a ratio of 1-PL packets are received in time at the voice decoder as given by the predefined acceptable packet loss, it can be expressed alternatively that at least the packets having a value rt for the relative arrival time being less than (or equal) to a threshold value for the relative arrival time rt_PL will be received in time in order to be played back by the voice decoder:

(III) CDF(rt_PL) = l - PL

In other words, a threshold for the relative arrival time rt_PL providing the predetermined packet loss PL can be established according to the function

(IV) rt_PL = CDF^~] (l - PL)

The earliest absolute transmission deadline (EATD) can then be calculated as:

(V) EATD_k = t_{rec k} + rt_PL - rt_recΛ + L_n

This value may be delivered to the scheduler.

Alternatively, it can be expressed that the maximum latency for packet k in node n can be increased from the given maximum predefined latency requirement pertaining to all packets for a given node L_n to an extended maximum latency requirement for packet k;

(VI) L_n^ L_n + MAX{Q,rt_PL -rt_{rec k})

without increasing the packet loss at the voice decoder of the user entity.

The extended maximum latency requirement may be provided to the scheduler.

In fig. 6, a base station node corresponding to node 14 in fig. 1 is shown. The base station 14 comprises a time estimator 103, a queue buffer 104, a transmitter 105 for trans- mitting packets over an air interface to user entity UE 15 and a scheduler 101 , which performs the scheduling of packets which is momentarily stored in buffer 104. A receiver 106 is also provided for receiving up-link traffic from the user entity. According to one preferred embodiment the time estimator 103 is performing the estimation of the cumulative density function and is providing the extended maximum latency requirement or the earliest absolute transmission deadline (EATD) to the scheduler 101.

It is noted that the actual absolute transmission deadline can be set even later as defined by the applied scheduling policy, the prevalent channel conditions, transmission demand and / or the given capacity in the intermediary node.

According to the first embodiment of the invention, packet losses are ignored and pack- ets are consequently assigned sequence numbers in the order they appear in the base station 14. The frame rate T can be sensed either from an average or given as a set system parameter depending on the given service.

In conclusion, there is provided:

A intermediary node (14) adapted for receiving a sequence of a packets from a server (10) and re-transmitting the sequence of packets over a time variant channel to a receiving entity (15), wherein a given predefined latency L_n pertain to the intermediary node and wherein a predefined packet loss PL pertains to a sequence of packets transmitted from the server over a path of a series of arbitrary nodes () to the intermediary node (14) and further on to the receiving entity (15).

The intermediary node comprises a time estimator (103) adapted for performing the following steps - resolving the sequence number (k) and time of arrival to the intermediary node for at least a plurality (q) of incoming packets - step 21 ;

- establishing the frame period of the incoming packets (T) - step 22;

- establishing a sequence of normalized packet arrival times (e_k) as corresponding to the sensed frame period (T) - step 23; - calculating the relative arrival time rt_{rec, k} of the plurality (q) of incoming packets in relation to the normalized packet arrival times step - 24;

- creating a cumulative density function (CDF) for a given sequence of packets - step 25;

- establishing the threshold value for the relative arrival time rt_PL yielding the predefined packet loss (PL) based on the cumulative density function (CDF) - step 26;

Advantageously, for each subsequently received packet, the time estimator is - establishing in step 27 an absolute transmission deadline (EATD_k) from the time of arrival of the received packet (t_rec,_k); the established threshold value for the relative arrival time (rtp_L); the relative arrival time (rt_rec,k) and the predefined latency value (L_n) for the re- transmission node.

Alternatively, for each subsequently received packet the time estimator is

-establishing in step 27 an extended maximum latency (L_{n k}) from the predefined latency value (L_n) for the intermediary node and a maximum value of the difference between the threshold value for the relative arrival time rt_PL, and the relative arrival time rt_rec, _k for the given packet.

Advantageously, a sequence - or moving window - of individual packets corresponding to a plurality of last received packets is used for establishing the CDF. Thereby, the latest data available for the given communication path is reflected. Hence after step 27, the routine is going to step 21.

There is moreover provided a method for receiving a sequence of a packets from a server (10) over an intermediary node (14) and re-transmitting the sequence of packets over a time variant channel to a receiving entity (15), wherein a given predefined latency L_n pertaining to the intermediary node and wherein a predefined packet loss PL pertains to a sequence of packets transmitted from the server over a path of a series of arbitrary nodes (11 , 12, 13, 16) to the intermediary node (14) and further on to the receiving entity (15); comprising the following steps

- establishing the frame period of the incoming packets (T)(22);

- establishing (26) the threshold value for the relative arrival time rt_PL yielding the prede- fined packet loss (PL) based on the cumulative density function (CDF). Further embodiments of the invention

According to a further aspect of the invention a prevalent packet loss in any of the previous nodes 1 to N -1 are identified. If packet k is lost, or if packet k-1 and k are concatenated into a new larger packet, in any of the nodes in the path from server to receiving entity, the loss will result in a subsequent packet rt_rec,kWill be calculated as the real rtrec,k+i ⁺ T, i.e. all findings of the relative times rt after packet k-1 will be assigned an in- creased, and hence erroneous, relative arrival time rt corresponding to the delay by frame time T.

In fig. 7, a second embodiment of a base station is provided, the second bases station differs over the fig. 6 base station in that a packet sniffer / filtering unit 102 is employed in the base station for detecting a possible frame number / associated with the packet data stream if provided.

According to one further embodiment of the invention, all packets are advantageously identified by a transmission number k (usually provided for packet data) such that it is possible to compensate the relative packet arrival time for lost or concatenated packets, cf. expression II.

An alternative more sophisticated handling where packet loss is affecting the L_{n k} expression may be implemented, such that packets with rt_{rec k} larger than rt_PL may be treated different from other packets, e.g. by causing the time estimator to restart the moving window or by associating an incremented transmission number with the belatedly received packet.

Claims

1. Intermediary node (14) adapted for receiving a sequence of a packets from a server (10) and re-transmitting the sequence of packets over a time variant chan- nel to a receiving entity (15), wherein a given predefined latency L_n pertaining to the intermediary node and wherein a predefined packet loss PL pertains to a sequence of packets transmitted from the server over a path of a series of arbitrary nodes (11, 12, 13, 16) to the intermediary node (14) and further on to the receiving entity (15);

the intermediary node comprising a time estimator (103) adapted for performing the following steps

- establishing the frame period of the incoming packets (T)(22);

- calculating (24) the relative arrival time rt_rec, _k of the plurality (q) of incoming pack- ets in relation to the normalized packet arrival times;

- establishing (26) the threshold value for the relative arrival time rt_PL yielding the predefined packet loss (PL) based on the cumulative density function (CDF);

2. Intermediary node according to claim 1 , wherein for each subsequently received packet

- establishing an absolute transmission deadline (EATD_k) from the time of arrival of the received packet (tea); the established threshold value for the relative arrival time (rtpj; the relative arrival time (rUα) and the predefined latency value (L_n) for the intermediary node.

3. Intermediary node according to claim 1 , wherein for each subsequently received packet

-establishing an extended maximum latency (L_n,k) from the predefined latency value (L_n) for the intermediary node and a maximum value of the difference between the threshold value for the relative arrival time rt_PL, and the relative arrival time rtrec, k for the given packet.

4. Intermediary node according to claim 2 moreover comprising a scheduler (101), wherein the absolute transmission deadline (EATD) is provided to the scheduler.

5. Intermediary node according to claim 3 moreover comprising a scheduler (101), wherein the extended maximum latency (L_{n k}) is provided to the scheduler.

6. Intermediary node according to any of claims 1-5, wherein the threshold for the relative arrival time is defined according to

rt_Pl = CDF^'1 {{ -PL)

where PL is the predefined packet loss and CDF is the cumulative density function.

7. Intermediary node according to claim 2 or 4, wherein the earliest absolute transmission deadline (EATD) is calculated as:

EATD_k = t_{rec k} + rt_PL - rt_recΛ + L_n

where t_rec,_k is the time of arrival of the received packet; rt_PL is the established threshold value for the relative arrival time; rt_rec,k is the relative arrival time and L_n is the predefined latency value.

8. Intermediary node according to claim 3 or 5, wherein the extended maximum latency requirement for packet k; L_n,k is calculated according to

L_{n k} = L_n +M4X(0,rt_PL ~rt_{rec k})

where L_n is the given maximum predefined latency requirement pertaining to all packets.

9. Intermediary node according to any preceding claim, wherein in a packet sniffer or filtering unit (102) is provided for detecting a possible frame number associated with the packet.

10. Intermediary node according to any preceding claim, wherein a sequence - or moving window - of individual packets corresponding to a plurality of last received packets (24 - 28) is used for establishing the cumulative density function (CDF).

11. Method for receiving a sequence of a packets from a server (10) over an intermediary node (14) and re-transmitting the sequence of packets over a time variant channel to a receiving entity (15), wherein a given predefined latency L_n pertaining to the intermediary node and wherein a predefined packet loss PL pertains to a sequence of packets transmitted from the server over a path of a series of arbitrary nodes (11 , 12, 13, 16) to the intermediary node (14) and further on to the receiving entity (15); comprising the following steps

- establishing the frame period of the incoming packets (T)(22);

- calculating (24) the relative arrival time rt_{rec, k} of the plurality (q) of incoming pack- ets in relation to the normalized packet arrival times;

12. Method claim 11 , wherein for each subsequently received packet

- establishing an absolute transmission deadline (EATD_k) from the time of arrival of the received packet (tac_k), the established threshold value for the relative arrival time (rtpι_), the relative arrival time (rt_rec,k) and the predefined latency value (L_n) for the intermediary node.

13. Method according to claim 11 , wherein for each subsequently received packet

-establishing an extended maximum latency (L_n,k) from the predefined latency value (L_n) for the intermediary node and a maximum value of the difference be- tween the threshold value for the relative arrival time rt_PL, and the relative arrival time rt_{rec, k} for the given packet.

14. Method according to any of claims 11-13, wherein the threshold for the relative ar- rival time is defined according to

rt_PL = CDF ^ (l - PL)

where PL is the predefined packet loss and CDF is the cumulative density func- tion.

15. Method according to claim 12, wherein the earliest absolute transmission deadline, EATD, is calculated as:

EATD₁ = t_recJc + rt_PL - rt_recJl + L_n

where tre_C,k is the time of arrival of the received packet; rtpL is the established threshold value for the relative arrival time; rt_rec,_k is the relative arrival time and L_n is the predefined latency value.

16. Method according to claim 13, wherein the extended maximum latency requirement for packet k, L_{n k}, is calculated according to

L_{n k} = L_n +MAX{Q,rt_PL -rt_rec,)

17. Method according to any preceding claim 11 - 16, wherein detecting a possible frame number associated with the packet.

18. Method according to any preceding claim 11 - 17, wherein a sequence - or moving window - of individual packets corresponding to a plurality of last received packets (24 - 28) is used for establishing the cumulative density function (CDF).