CA2256341C

CA2256341C - Process for optimizing the utilization of connecting links in systems which transmit data in data packets

Info

Publication number: CA2256341C
Application number: CA002256341A
Authority: CA
Inventors: Uwe Briem
Original assignee: Siemens AG
Current assignee: Siemens AG
Priority date: 1996-05-21
Filing date: 1997-05-12
Publication date: 2003-02-11
Anticipated expiration: 2017-05-12
Also published as: ATE216176T1; EP0900513A1; CA2256341A1; WO1997044985A1; DE59706973D1; AU2950797A; ES2175407T3; JPH11514176A; DK0900513T3; DE19620428A1; AU711972B2; JP3260383B2; EP0900513B1

Abstract

The weighted fair queueing scheduling method has been developed in the prior art for the transmission of data packets. This method ensures only a lower limit on the transmission rate of the data packets. In order to be able to achieve an upper limit on the transmission rate as well, a further scheduling method may be used prior to this, in the method according to the invention.

Description

96 P 1499 P - 1 _ Fi ~ i . ',=---; ~ ~'~ "~ ~ , Description ~~~ ...,:.;:L~., .;.. ~;
Method for optimization of the utilization of connecting sections in systems in which information is transmitted in data packets.
The invention relates to a method as claimed in the precharacterizing clause of patent claim 1.
In modern packet switching systems, information is transmitted in data packets . One example - of this is ATM cells . These have a header part and an information part. The header part is used to store connection infor-mation, and the information part to store the wanted data to be transmitted. As a rule, the actual transmission takes place via connecting sections between the trans-mitter and receiver. In this case, there may be a requirement to utilize the connecting sections in such a manner that a plurality of transmitting devices transmit the cell streams originating from these devices via the same connecting section.
In order to allow the transmission of the respec tive cell streams to be carried out in accordance with the requirements of the individual cell streams, a so called WEIGHTED FAIR QUEUEING SCHEDULING method has become generally accepted in the prior art. The corres ponding relationships are described, for example, in the document "Virtual Spacing for Flexible Traffic Control", J.W. Roberts, International Journal of Communication Systems, Vol. 7, 307-318 (1994). In this case, the individual cell streams are assigned different weighting factors, which are used to control the actual trans-mission process on the individual connecting sections.
Reference should be made to Figure 3 to assist under-standing.

By way of example, this shows cell streams 1 ...
n. The n cell streams are passed from a transmitting device DEMUR in the direction of one or more receivers.
In practice, only one common connecting section is used in this case . The n cell streams are furthermore assigned weighting factors r1 .., rn. To assist understanding, it is assumed that it is intended to pass only two cell streams, namely the cell streams 1, 2, via a connecting section. The connecting section is furthermore intended to have a maximum transmission capacity of.150 Mbit/s.
The two cell streams 1 and 2 are assigned weightings r1 =
2 and r2 - 1. This results in the cell stream 1 being transmitted at a transmission rate of 100 Mbit/s, and the cell stream 2 at only 50 Mbit/s, if cells for both cell streams are present for transmission. If only one of the two cell streams has cells to transmit, this cell stream is assigned the total transmission capacity of 15o Mbit/s.
Figure 2 shows how the theoretical considerations addressed above are implemented in practice in the prior art. This shows how data packets, or ATM cells, are dealt with using the weighted fair queueing scheduling algorithm. In this case, incoming cells are supplied to the input device EE, are passed on to the demultiplexing device DEMUR and are stored there with the aid of a demultiplexing function, which is implemented here, and with the assistance of an identifier QID in a logic queue. The identifier QID is in this case contained in the cell header of each cell.
At the same time, control data which are deter-mined in the input device EE are for this purpose supplied to a scheduler device S. A scheduling algorithm which is known per se is executed in this device.

This may be, for example, the weighted fair queueing scheduling algorithm or any other algorithm. This algorithm determines, for example, the sequence in which or the time at which it is intended to read the cells which are stored in the buffer stores P1...Pn. The result of the assessment of the control data by this algorithm is supplied to the output device AE. The cells stored in the buffer stores P1...Pn are now read, on the basis of the result of the assessment, by the algorithm which is being executed in the scheduling device S. Furthermore, an acknowledgement signal is fed back to the input device EE. After this and when a new cell with an identifier QID
arrives in the input device EE and when an acknowledge-ment 'selected QID' is present, the input device EE uses the buffer filling level for QID - i as well as the scheduling method to decide whether the message "SCHEDULE
QID" is generated. This message indicates to the scheduler device S that it should carry out initial planning for the next transmission time for this identi fier QID, in some way.
A problematic feature of such a procedure is that, although the weighted fair queueing scheduling algorithm guarantees minimum cell rates, a maximum cell rate limiting cannot be carried out here. However, this is a major factor since, in practice, both minimum and maximum cell rates often have to be complied with - for example in the case of ABR (available bit rate) traffic.
The invention is based on the object of indicat ing a way in which the weighted fair queueing scheduling algorithm can be modified in such a way that optimized transmission is ensured here as well.

The object is achieved on the basis of the features specified in the precharacterizing clause of patent claim 1, by means of the features in the characterizing part.
An advantageous feature of the invention is that a two-stage scheduling method may be carried out depending on an identifier which is contained in the packet header. In this case, the result of the first stage is used as an input signal for the second stage. This results in particular in the capability to control both a lower limit and an upper limit of the cell rate. In particular, this method is not limited to the use of a specific algorithm.
In accordance with this invention, there is provided a method for optimization of the utilization of connecting sections in systems in which information is transmitted in data packets, having a scheduling method (S2) by means of which connection parameters, which are representative of lower transmission rates of the data packets, are guaranteed during the transmission process, and having a queue identifier (QID) which is stored in the packet header, characterized in that a further scheduling method (S1) may precede the scheduling method (Sz) depending on the queue identifier (QID), by means of which further scheduling method (S1) the connection parameters which are representative of upper transmission rates of the data packets are limited during the transmission process.
Further refinements of the invention are provided in the dependent claims.
Claim 2 provides that the connection parameters are limited during the transmission process, by means of the first stage of the two-stage scheduling method. This is 4a intended, in particular, to control limiting of the cell rate. This results in the cells not being transmitted at higher cell rates during the transmission process.
Claim 3 provides that the second stage of the two-s stage scheduling method is the weighted fair queuing scheduling algorithm. This is linked to the advantage that a proven method can be used. A further advantage of this is that this algorithm guarantees lower limiting of the cell rate.
Claim 4 provides for an input device to contain a table which contains the current filling levels of the buffer stores. This is linked to the advantage that a current map of these filling levels is stored here at all times.

Claim 5 provides that, depending on the control data obtained from the scheduler device, the output device takes cells from at least one of the buffer stores and acknowledges this process to the input device. As a result of the feedback, the reading process has a direct influence on the first stage of the two-stage method. The two stages of the two-stage scheduling method thus do not operate independently of one another. The way in which the first stage operates is influenced by the way in which the second stage operates. The identifier or the packet length may be used, for example, as feedback parameters.
Claim 6 provides that the identifier is entered while the connection is being set up.
Claim 7 provides that the data packets are ATM
cells. The invention can thus be applied in particular to ATM networks.
In the figures:
Figure 1 shows the method according to the invention, Figure 2 shows the practical application of the prior art, Figure 3 shows theoretical considerations relating to the prior art.
Figure 1 shows the method according to the invention. In this case, it is assumed that the infor-mation is transmitted in ATM cells, using an asynchronous transfer method (ATM).

The cells are supplied to the input device EE in a cell stream. The routing information is stored in the header part of each cell. Furthermore, an identifier QID
has been stored here while the connection is being set up. This identifier is a cell stream identifier which is entered in the cell header on a connection-specific basis or for a group of connections. As a rule, the identifier QID is assigned simple numerical values. In the present exemplary embodiment, the identifier QID is intended to have the values 1...N. Originating from the input device EE, the cells themselves are supplied to the demultiplex ing device DEMUR where they are written into buffer stores P1...Pn, designed as a logic queue, with the aid of a demultiplexing function, which is implemented here, and with the assistance of the identifier QID.
The input device EE furthermore contains a table T as to which of the connections require the connection parameters to be limited during the transmission process .
In the present exemplary embodiment, it is assumed that limiting of the cell rate is controlled in the sense of limiting the connection parameters. In order to verify the connection, the identifier QID is taken from each of the incoming cells and is compared with the entries in the table T.
If it is not intended to limit the cell rate for a connection, corresponding control data are supplied via the connecting section B to the scheduler device S2, bypassing the scheduler device S1. There, the control data are used in a scheduling algorithm which is known per se. In the present exemplary embodiment, this is intended to be weighted fair queueing scheduling method, which has already been described in the introduction.
Such an algorithm results in a lower cell rate being guaranteed, in the sense of guaranteeing the connection parameters of the cells during the transmission process.

According to the present exemplary embodiment, the cell rate for one of the connections is limited, for example for the connection with the number 8 (QID=8?. In this case, the control data are supplied via the connect-s ing section A to the scheduler device S1. Here, an algorithm starts to be executed, which controls an upper limit on the cell rate. This is done by a function implemented here using the identifier QID for initial planning of the control data supplied from the input device EE, such that the individual cells do not exceed a predetermined rate. At the time at which the scheduler device S1 would read a cell, it produces, however, a control signal itself for initial planning of the read time, corresponding to the general scheduling algorithm being executed in the scheduler device S2. No initial planning of the next event takes place in the scheduler device S1 for the same identifier QID. Thus, stimulated by the scheduler device S1, the scheduler device S2 plans the sequence for the indicated identifier QID, corres-ponding to the scheduling algorithm being executed here.
The cells initially planned by the scheduler device S1 may thus experience an additional delay. The peak bit rate set in the scheduler device S1 may thus be different from that used to read the cells.
By way of example, the weighted fair queueing scheduling algorithm is intended to be used in the scheduler device SZ in the present exemplary embodiment, although other algorithms can also be used. The method according to the invention is not limited to the use of a specific algorithm.
The result of the evaluation of the algorithm being executed in the scheduler device Sz is passed to the output device AE.

g _ Whenever it is intended to read the next cell from a buffer store Pl...Pn with a specific identifier QID, this is indicated to the output device AE. This reads the first cell with the indicated identifier QID from the buffer store P1...Pn in question, and reports this together with the corresponding identifier QID to the input device EE. The latter then checks whether a further cell with this QID is stored in the buffer store. If this is the case, a corresponding signal (SCHEDULE QID) is sent to the scheduler device S1. If this is not the case, no further action takes place in the sense of initial planning (reading) in the scheduler device S1 for this identifier QID.
This method means that an event for an identifier QID can be initially planned only in the scheduler device S1 or S2, but not at the same time in both devices.
Furthermore, the two function blocks S1 and Sz are not linked to a specific implementation. This two-stage algorithm is thus used to determine the sequence in which and the time at which it is intended to read the cells which are stored in the buffer stores P1...Pn.
Finally, it should also be mentioned that the above exemplary embodiment has been described using the example of ATM cells. However, the invention is not just limited to this. The method according to the invention can likewise be used for the transmission of information in data packets as such. However, in this case it is necessary to ensure that the packet length is added to the control data.

Claims

1. A method for optimization of the utilization of connecting sections in systems in which information is transmitted in data packets, having a scheduling method (S2) by means of which connection parameters, which are representative of lower trans-mission rates of the data packets, are guaranteed during the transmission process, and having a queue identifier (QID) which is stored in the packet header, characterized in that a further scheduling method (S1) may precede the scheduling method (S2) depending on the queue identifier (QID), by means of which further scheduling method (S1) the connection parameters which are representative of upper transmission rates of the data packets are limited during the transmission process.

2. The method as claimed in claim 1, characterized in that the scheduling method (S2) is a weighted fair queueing scheduling algorithm.

3. The method as claimed in claim 1 or 2, characterized in that an input device (EE) contains a table (T) which contains the current filling levels of buffer stores (P1...Pn).

4. The method as claimed in one of the preceding claims, characterized in that, depending on the control data which are obtained from the scheduling method (S2), an output device (AE) takes data packets from at least one of the buffer stores (P1...Pn) and acknowledges this process to the input device (EE).

5. The method as claimed in one of the preceding claims, characterized in that the queue identifier (QID) is entered while the connection is being set up.

6. The method as claimed in one of the preceding claims, characterized in that the data packets are ATM cells.