EP1454463A1

EP1454463A1 - Method and system for minimizing jitter in a packet switched network

Info

Publication number: EP1454463A1
Application number: EP02791163A
Authority: EP
Inventors: John Tillberg
Original assignee: Telia AB
Current assignee: Telia Co AB
Priority date: 2001-12-04
Filing date: 2002-12-03
Publication date: 2004-09-08
Also published as: SE0104053L; NO20042192L; SE518550C2; AU2002365781A1; WO2003049386A1

Abstract

The invention relates to method and system at packet switching information transmission. The information is transmitted in form ofpackets from a source (K) via a number of routers (R) to a reciever (D). Each packet which arrives at/to a router (R) is checked whether it is received within an indicated time-frame or not. Packets which are arriving out of said time-frame are allocated a flag which indicates whether the packets have been received too early or too late. The packets are aftere that forwarded to next/subsequent router which at examination of the packet observes that said flag is set. Depending on the character of the flag, the current packet is forwarded prioritized or delayed. In connection with that the packets are forwarded said flag is deselected/cancelled/removed. The receiver (D) in this way will receive packets in a sequence with fewer errors than what is normally the case.

Description

Method and system for minimizing jitter in a packet switched network.

TECHNICAL FIELD

The present invention relates to information transmission at which especially packet switching transmission is referred to.

TECHNICAL PROBLEM

The computer networks of today must be equipped with methods to manage to carry services with strict demands on delays (for instance Voice over IP (VoIP)) and jitter (for instance Streaming) . The Quality of Service (QoS) -methods of today solves this by utilizing in the first place dynamic resource allocation (for instance RSVP) and traffic differentiation (for instance Differentiated Services (DiffServ) ) together with sophisticated network planning. A limitation at these methods is that they cannot adapt themselves to the momentary fluctuations of the data communications. This can result in jitter. There is in other words no way to, as regards time, compensate packets which have been delayed. At packet switching there consequently is a need route the packets in the transmission/switching in a way that they will reach the receiver within certain intervals of time in relation to each other. When this is not done, interruptions in the transmission may occur, or packets be received in incorrect order. In the cases packets are prioritized, there further is a risk that the packets will reach the receiver in wrong relative order. Further problem is to restore the packets as regards time and in relation to each other correct order and right time interval. Within the known technology corrections of this kind are not previously known. PRIOR ART

By the American document US6167030 is evident that it is known to, in a transmitter node, decide priority for cells depending of the filling degree of the buffer. The priority can be changed in the following/subsequent nodes, but only as a consequence of that previous cells in the same message have been given higher priority. Further the document EPA473330 describes how the priority of packets in a queue can be upgraded after a waiting time in queue. It consequently is previously known to change the priority of packets depending on the filling degree in a buffer in the transmitter node. Further the priority can be changed further on in the network, however not as a consequence of delays. It is further known to increase the priority of delayed packets in a switch, which information is not forwarded to the next node.

Priority of certain packets can result in that the packets at reception at the receiver is received in incorrect order. Consideration has been taken to this in certain cases at a receiver which restores the packets in their correct order. However, there are systems in operation which are not capable of restoring the packets in their correct order at incorrect reception

THE SOLUTION

The present invention relates to method and system at information transmission between a source (K) and a receiver (D) via a number of links (R) . Packets which are received in a first link (R) , are examined with regard to a time average (W) . Packets which have time averages out of set frames are allocated a priority. At reception of said packet in the next link, the packet is time compensated depending on the degree of priority which has been set in the previous link. The variance of the average delay of the packets is in this way reduced. In a further development of the invention, the packets in the first link are equipped with a flag which indicates that said packets shall have special treatment. The flag indicates, by different signaling, whether the packets shall be given priority or be delayed in the second link. After that the second link removes the set marking for the packet in question, after which it is treated in normal way in the subsequent link. To decide whether the packets have been delayed or received too early, an upper and a lower value of normal delay is indicated, which values also indicate how said priority shall be set. In a further development of the invention, the receiver (D) respective the second link (R) are allowed to restore the packets in their correct order in the cases the packets are received in incorrect order. For this purpose the links and/or the receivers include means for identification of the correct order of the packets. The system includes functions of similar kind as the method.

ADVANTAGES

The invention is in principle usable in all computer networks where jitter shall be minimized. Particularly is the invention usable for IP-based cellular Radio Access (RAN) for UMTS. A RAN can cover geographically large areas and include a large number of nodes. Up to 100% of the traffic in these networks can consist of real time traffic. Often leased lines are used to the base stations, which is expensive in comparison with the transmission costs in the core network. Because the number of base stations is expected to be large, the costs for said connections will be very considerable for the operators. An over- dimensioning of the connections, with the aim to reduce delays and jitter is no attractive alternative. RAN is further expected to support a plurality of types of QoS for a large number of sessions. The efficiency demands on QoS- mechanisms will be very high in this type of network. The present invention consequently has the advantage that packets in packet switching transmissions can be compensated with regard to the variance of the average delay. By the packets being given different priorities routers (R) in the system can analyze and correct forwarding of the packets. It consequently gives instructions for how the packets shall be handled/treated in the distribution chain to avoid/minimize the variance of the average delay between the packets .

DESCRIPTIONS OF DRAWINGS

Figure 1 shows the principle of the invention. K = Source; R = Router; D = Destination; W = Delay;

E (W) = Average delay; A = Delayed packet; B = Compensation Figure 2 shows the invention in flow charts.

PREFERRED EMBODIMENT The invention is a method to reduce jitter in IT-networks by compensating packets which have been delayed due to congestions in the network. Each router in the network checks continuously the size of the queues in its own buffers, and the time average of these over a certain period of time (not too long) . By doing this, a router itself can discover if incoming packets are exceptionally delayed. If that is the case, the packets are marked to indicate this to the subsequent router. This is detected by the subsequent router which then places these packets in a queue with higher priority, and cancels the priority of the packets again. This results in that the effect of the momentary delay is reduced, which implies that the variance of the delay (jitter) en-to-end will be lower. However, it should be mentioned that the method can result in that certain packets arrives at destinations in incorrect order. This implies that the method is not suitable for TCP- traffic, but only for UDP.

In Figure 2 is shown how occasional congestions are registered at router 2. Router 2 then marks the packets to indicate abnormal delay. Router 3 detects this and prioritizes these packets. This results in that they at router 3 experience a shorter than usual. Router 3 consequently compensates for a share of the abnormal delay that the packets experienced at router 2. This means that the total delay from source to destination becomes - from a statistical point of view - more even.

An example of how the method might be realized is by means of DiffServ. For example should two Assured Forwarding

(AF) -classes be possible to use for this; one for "normal" UPD-traffic (AF2), and one with higher priority which is used exclusively for delayed traffic (AF1) . All AF2-traffic which has been abnormally delayed is re-marked to AF1- traffic. All traffic which is received as AF1 is prioritized before AF2 and is re-marked back to AF2 again. This example might be used at, for instance, streaming of audio/video. One possibility is software programming of DiffServ-capable routers to implement this.

The method according to the above differs from existing QoS-mechanisms as far as the network has ability to in a simple way dynamically be able to compensate packets which have already been delayed without the need to know any specific information abut each individual flow.

The method can in principle be used in all data communication networks which intend to minimize jitter.

Particularly interesting to use is the method for IP-based cellular radio access networks (RAN) for UMTS. A RAN can cover geographically large areas and include a large number of nodes. Up to a hundred per cent of the traffic in these networks may be constituted of real time traffic. Often leased lines are used to the stations, which is very expensive in comparison with the transmission costs in the core network. Because the number of base stations is expected to be very large, the costs for these connections will be very considerable for the operators. An over dimensioning of these connections, with the aim to keep delays and jitter down, consequently is no alternative. At the same time a RAN is expected to be able to support a large number of different types of QoS for a large number of sessions. The efficiency demands on QoS-mechanisms consequently will be very high in this type of networks, so the above described method should be an interesting alternative .

The invention consequently on the one hand is a method for reduction of jitter at packet switching information transmission, and on the other a system in which said method is utilized. The principle of the invention has been described above. It is, however, obvious that the packets which are prioritized can reach the destination too early. The invention in one development describes a solution also to this problem. The routers consequently do not only checks if packets are arriving late, but also if they are arriving too early. The treatment/handling of early arrived packets in principle agrees with what has been described above, with the difference that said packets are put in a specific queue where the packets in a subsequent router are delayed. To distinguish packets which shall be delayed, accelerated or have a normal treatment, respective packet in the router which detects errors, if any, is given a flag which is identified by subsequent router which after that treats the packets according to the indication of the flag. By the packets being compensated individually, these can arrive at subsequent router, respective destination, in disorder. If no compensation for this condition is stored in the routers, the destination has to attend to that the packets are treated/handled in correct order. This implies that the destination must include methods for restoring the packets to their original order.

Figure 2 shows in form of flows the functions of the invention. A packet (A) is received by the router (B) which examines/checks if the packet has flag set (C) . If it is not set, is after that examined if the packets have arrived delayed (D) . If the packet has not been delayed, is examined if the packet has arrived too early (E) in relation to the set time frames. If also this check is negative, the packet is forwarded in normal way (F), after which the subsequent router examines the packet in similar way when it is received.

In the cases the flag is set, the process changes to (1) where is examined if the packet is delayed (G) . If the packet is delayed, it is placed in a queue for more rapid treatment than normal (H) . After that, the flag is deselected/cancelled (I) and the packet is forwarded (F) in the chain. In the cases the packet is not delayed (G) , the packet is placed in a queue for delay (K) , after which said flag is deselected/cancelled (I) and the packet is forwarded in the chain.

In the cases it is found that a packet has been delayed (D) , the process changes to loop 2 where flag is set (L) and the packets shall be prioritized. After that, the packet is forwarded (F) .

In the cases it is found that the packets have arrived too early (E) , the process changes to phase 3 where the flag is set that the packets shall be delayed (M) , after which said packet is forwarded in the chain. The frames which are set for the times within the packets shall be received are, as the expert well knows, depending on the behavior of the system.

The systems include a source (K) from which information is transmitted in the system via a chain of routers (R) to eventually be received by a receiver (D) . Each router identifies the packets which have been received and compare their arrival, from a time aspect, in relation to other packets. At the comparison is found whether the packets have arrived in right time, too early or too late. Packets which arrive within set time frames are forwarded without measure. Packets which arrive too late are allocated a flag which indicates that the packet in question is late. In the subsequent router, this detects this condition and places said packet in a queue to accelerate distribution of said packet. After that, said flag is deselected/cancelled and the packet is after that regarded as normal at the treatment in the subsequent router. In the case that a packet arrives to the first router too early, the packet is allocated a different flag which indicates that it has been received early. When said packet has been received in the second router, this detects this condition and places the packet in a queue which delays the packet before it is forwarded in the chain. In connection with the forwarding said flag is deselected/cancelled.

Systems of this kind run the risk of creating disorder between the packets since they by the procedure which has been described can land up in disorder. To bridge this problem the receiver must have means to restore the packets in their original order. This is done by the packets having identificators which the receiver (D) can utilize. Identificators for this purpose are since earlier known within packet switching information. It can at that be referred to that packets need not necessarily be distributed the same route, which is earlier known. The packets can go different routes, at which the receiver has to arrange the packets in their correct order before they are utilized.

Other possibilities to overcome the trouble with the order between the time arrival of the packets are to introduce checks/controls in the routers (R) . It will be obvious that the routers in this respect have the same function as the receiver (D) .

The invention relates to method and systems at packet switching information transmission. The information is transmitted in form of packets form a source (K) via a number of routers (R) to a receiver (D) . Each packet which arrives at/to a router (R) is checked whether it is received within a given time-frame or not. Packets which arrive out of said time-frame are allocated a flag, which indicates whether the packets have been received too early or too late. The packet is after that forwarded to subsequent router which at examination of the packet notices that said flag is set. Depending on the character of the flag, the packet in question is forwarded prioritized or delayed. In connection with that the packets are forwarded, said flag is deselected/cancelled. The receiver (D), in this way will receive packets in a sequence with fewer incorrections than what is normal.

Claims

PATENT CLAIMS

1. Method at packet switching information transmission, where the information is transmitted from a source (K) to a receiver (D) via a number of links (R) , c h a r a c t e r i z e d in, that packets received in a first link (R) are examined with regard to a time average, that packets which have abnormal queu waiting period are allocated a priority, at which the packets in a second link are forwarded with a time compensation corresponding to the priority which has been allocated to the packets, at which the variance of the average delay for the packets is reduced.

2. Method as claimed in patent claim 1, c h a r a c t e r i z e d in, that the packets, at observed abnormal queu waiting periods in the first link are equipped with a flag indicating that marked packets shall have special treatment.

3. Method as claimed in patent claim 2, c h a r a c t e r i z e d in, that the flag in a first case indicates that the packets are delayed to the first link, that the flag in a second case indicates that the packets arrive earlier than expected at/to the first link.

4. Method as claimed in patent claim 3, c h a r a c t e r i z e d in, that the flag in the first respective second case are indicated differently, which is in a clear-cut way registered by the second link.

5. Method as claimed in patent claim 4, c h a r a c t e r i z e d in, that the packets are received in the second link, where said flag is sensed/registered, after which current packets will have special treatment and are time compensated depending on the first or the second case.

6. Method as claimed in patent claim 5, c h a r a c t e r i z e d in, that the packets in the first case are given priority in relation to other packets .

7. Method as claimed in patent claim 5, c h a r a c t e r i z e d in, that the packets in the second case are given a delay in relation to the other packets .

8. Method as claimed in any of the patent claims 2-7, c h a r a c t e r i z e d in, that the second link removes said flag when the packets in question have been time compensated.

9. Method as claimed in any of the previous patent claims, c h a r a c t e r i z e d in, that too early received packets, respective too late received packets, are indicated as an average delay.

10. Method as claimed in any of the previous patent claims, c h a r a c t e r i z e d in, that the average delay for the packets from the source (A) to the receiver (B) is allowed to vary within a predetermined margin, which is regarded as normal.

11. Method as claimed in any of the previous patent claims, c h a r a c t e r i z e d in, that the receiver (D) when necessary restores the order between the packets which have been received in incorrect order.

12. Method as claimed in any of the patent claims 1-10, c h a r a c t e r i z e d in, that the second link (R) when necessary restores the order between the packets which have been received in incorrect order.

13. Method as claimed in any of the patent claims 11 or 12, c h a r a c t e r i z e d in, that the receiver (D) respective the links (R) include means for identification of the relative order of the packets.

14. System at packet switching information transmission, including a source (K) from which information is transmitted in packets to a receiver (D) via a number of links (R) , c h a r a c t e r i z e d in, that a first link (R) is arranged to, from a time aspect, check the arrival of the packets in relation to a time average, at which the packets which are found to be received out of the indicated time average are allocated a mark, that a second link (R) is arranged to give a specific treatment to the packets which have been allocated said mark, at which the second link is arranged to on the one hand, from a time aspect, correct the forwarding of the packets, and on the other cancel/remove the marking.

15. System as claimed in patent claim 14, c h a r a c t e r i z e d in, that the marking is arranged to indicate a first marking in the cases the packets have arrived at/to the first link (R) too late, respective a second marking in the cases the packets have arrived too early.

16. System as claimed in patent claim 15, c h a r a c t e r i z e d in, that the second link (R) is arranged to give priority to the packets which have been allocated the first marking.

17. System as claimed in patent claim 15, c h a r a c t e r i z e d in, that the second link (R) is arranged to delay packets which have been allocated the second marking.

18. System as claimed in any of the patent claims 14-17, c h a r a c t e r i z e d in, that the second link (R) , as the receiver (D) , is arranged to check the relative order of the packets, and to restore the packets which have been received in incorrect order, to their correct order.

19. System as claimed in any of the patent claims 14-18, c h a r a c t e r i z e d in, that the second link (R) , is arranged to calculate an average delay E (W) at the establishing of the arrival, from a time aspect, of the packets, and that limits for allowed average delay are specified.

20. System as claimed in patent claim 18, c h a r a c t e r i z e d in, that the second link (R) , respective the receiver (D), are allocated means for determining of the relative order of the packets.