Patent Application
20100124262
The present invention relates to telecommunications in general, and, more particularly, to adaptive traffic partitioning.
When information is transmitted from one node (e.g., a computer, a telecommunications terminal, a switch, a router, etc.) to another node via a communications channel, the performance of the channel can vary unpredictably due to channel loss, jitter, noise, and so forth. A well-known technique for monitoring the performance of a communications channel, and how the performance varies over time, is to inject additional traffic, known as probe traffic, into the transmission. A disadvantage of this technique, however, is the associated overhead and reduced throughput that results from the introduction of additional traffic into the communications channel.
The present invention pertains to situations in which there is a plurality of communications channels available for a source node to transmit information to a receiving node. In particular, the present invention provides a technique for adaptively and intelligently partitioning traffic among the communications channels, without injecting probe traffic into any of the channels. The present invention is therefore capable of adjusting the partitioning of traffic over time in order to maintain maximum performance, without any of the disadvantages associated with the introduction of probe traffic.
In accordance with the illustrative embodiment of the present invention, a source node transmits traffic to a receiving node via K communications channels in accordance with partition coefficients α1, . . . , αK, and the source node receives information regarding the quality of service (QoS) provided by each of the communications channels. The source node adjusts the values of the partition coefficients α1, . . . , αK adaptively based on:
In accordance with the illustrative embodiment, the channel QoS information is expressed as values of a quality-of-service metric (e.g., throughput, jitter, error rate, re-transmission frequency, some combination of two or more such metrics, etc.) for each of the channels. In addition, in the illustrative embodiment the measure of channel independence is a real-valued coefficient where a value of zero indicates that two channels are completely independent—i.e., that traffic and changes in QoS in one of the channels does not affect traffic or QoS in the other channel. Finally, in the illustrative embodiment partition coefficients α1, . . . , αK are real numbers in [0, 1] that indicate the fraction of traffic that is transmitted over each of the K communications channels
The illustrative embodiment comprises: receiving values q1, . . . , qK of a quality-of-service metric for K communications channels, wherein K is an integer greater than one, and wherein the values q1, . . . , qK are obtained without injecting any probe traffic into any of the K communications channels; and determining values for K coefficients α1, . . . , αK that quantify how subsequent traffic is to be partitioned among the K communications channels; wherein the determination of the coefficients α1, . . . , αK is based on: (i) the values q1, . . . , qK, (ii) one or more previous values of the coefficients α1, . . . , αK, and (iii) a measure of channel independence for one or more pairs of the K communications channels.
Communications channels 102-1 through 102-K are conduits for the transmission of information from source node 101 to receiving node 103, as is well-known in the art. In accordance with the illustrative embodiment, communications channels 102-1 through 102-K are bidirectional, and thus enable receiving node 103 to transmit information to source node 101 as well. As will be appreciated by those skilled in the art, however, in some embodiments of the present invention communications channels 102-1 through 102-K might instead be unidirectional (i.e., enabling transmission in only one direction from source node 101 to receiving node 103), and it will be clear to those skilled in the art, after reading this disclosure, how to adapt the method of
Source node 101 is one of a computer, a telecommunications terminal, a switch, a router, etc. that is capable of transmitting and receiving traffic via communications channels 102-1 through 102-K, and of performing tasks 310, 360, and 370, described below and with respect to
Receiving node 103 is one of a computer, a telecommunications terminal, a switch, a router, etc. that is capable of transmitting and receiving traffic via communications channels 102-1 through 102-K, and of performing tasks 320 through 350, described below and with respect to
Receiver 201 is capable of receiving signals via communications channels 102-1 through 102-K and of forwarding information encoded in these signals to processor 202, in well-known fashion.
Processor 202 is a general-purpose processor that is capable of receiving information from receiver 201, of reading data from and writing data into memory 203, of executing instructions stored in memory 203, and of forwarding information to transmitter 204, in well-known fashion. As will be appreciated by those skilled in the art, in some alternative embodiments of the present invention processor 202 might be a special-purpose processor, rather than a general-purpose processor.
Memory 203 is capable of storing data and executable instructions, in well-known fashion, and might be any combination of random-access memory (RAM), flash memory, disk drive, etc. In accordance with the illustrative embodiment, memory 203 stores executable instructions corresponding to tasks 310, 360, and 370, described below and with respect to
Transmitter 204 is capable of receiving information from processor 202, and of transmitting signals that encode this information via communications channels 102-1 through 102-K, in well-known fashion.
At task 310, partition coefficients α1, . . . , αK are initialized. As will be appreciated by those skilled in the art, in some embodiments of the present invention the coefficients might be initialized with equal values (e.g., 1/K, etc.), while in some other embodiments of the present invention the coefficients might be initialized based on information such as historical quality of service (QoS) data for the channels, parameter specifications of the channels (e.g., bandwidth, etc.), etc.
At task 320, QoS metric values q1, . . . , qK are estimated for communications channels 102-1 through 102-K, in well-known fashion. As will be appreciated by those skilled in the art, in some embodiments of the present invention in which traffic is composed of Voice over Internet Protocol (VoIP) packets, the delay, loss, and jitter of the packets might be estimated, and an R-factor QoS metric that combines these parameters into a single value might then be computed. Alternatively, in some other embodiments of the present invention, other traffic characteristics such as latency or measured bandwidth might be estimated, and another type of QoS metric based on these characteristics might then be computed.
Task 330 determines whether F(q1), . . . , F(qK), where F is a function, differ from their previous values by more than a threshold 6. As will be appreciated by those skilled in the art, in some embodiments F might simply be the identity function, while in some other embodiments F might be another type of function; moreover, the difference between new and previous values of F(q1), . . . , F(qK) might be computed in a variety of ways, such as the sum of the differences of the individual F(qi), the sum of the squares of the differences of the individual F(qi), the minimum difference over all of the F(qi), the maximum difference over all of the F(qi), and so forth. As will further be appreciated by those skilled in the art, the selection of an appropriate threshold value δ is a design decision that is typically best determined in each particular implementation.
If the new values differ from the old values by more than threshold δ, then execution continues at task 350, otherwise execution proceeds to task 340.
Task 340 checks whether at least Δ seconds have elapsed since the last time values q1, . . . , qK were transmitted to source node 101 at task 350—or, if task 350 has not yet been executed, since the method of
At task 350, values q1, . . . , qK are transmitted to source node 101, in well-known fashion. In accordance with the illustrative embodiment, values q1, . . . , qK are transmitted to source node 101 via communications channels 102-1 through 102-K. However, as will be appreciated by those skilled in the art, in some other embodiments of the present invention values q1, . . . , qK might be transmitted in an alternative fashion (e.g., via some other communications channel or network not depicted in the drawings, etc.), and it will be clear to those skilled in the art, after reading this disclosure, how to make and use such embodiments of the present invention.
At task 360, new values for partition coefficients α1, . . . , αK are determined based on:
In accordance with illustrative embodiment, previous values of α1, . . . , αK are considered at task 360 in order to smooth out changes of α1, . . . , αK over time. As will be appreciated by those skilled in the art, there are a variety of techniques known in the art for achieving such smoothing, such as a moving average, a maximum incremental change, and so forth, and the parameters associated with such techniques (e.g., the size of the moving average window, the value of the maximum incremental change, etc.) can be set to control the degree of smoothing (or alternatively, how aggressively the values of α1, . . . , αK change over time). Similarly, previous values of q1, . . . , qK are also considered at task 360 to smooth out changes over time.
In accordance with illustrative embodiment, the measure of channel independence is based on at least one of:
In particular, in the illustrative embodiment the measure of channel independence is a real number ρ, where ρ=0 indicates that two channels are completely independent, which means that:
As will be appreciated by those skilled in the art, the measure of channel independence might be either non-symmetric or symmetric, where the former case corresponds to an ordered pair of channels, and the latter case corresponds to an un-ordered pair of channels. Moreover, as will be appreciated by those skilled in the art, some other embodiments of the present invention might employ a different measure of channel independence than that of the illustrative embodiment, and it will be clear to those skilled in the art, after reading this disclosure, how to make and use such embodiments.
As will further be appreciated by those skilled in the art, there are a variety of ways in which the particular values of partition coefficients α1, . . . , αK might be set. For example, in some embodiments values α1, . . . , αK might be set proportional to the QoS values q1, . . . , qK (subject, possibly, to the smoothing influences of prior values), while in some other embodiments of the present invention the values of α1, . . . , αK might be based on a ranking of values q1, . . . , qK, irrespective of the relative magnitudes of q1, . . . , qK. For example, suppose K=3, q1=2 (the particular units are immaterial), q2=5, and q3=3. In the former case, the partition coefficients would be set as α1=0.2, α2=0.5, and α3=0.3, while in the latter case, the partition coefficients might be set as, say, α1=0.3, α2=0.37, and α3=0.33.
As will appreciated by those skilled in the art, there are yet other ways that the values of partition coefficients α1, . . . , αK might be set. For example, in some embodiments of the present invention, a geometric partition might be employed so that a disproportionately large fraction of traffic is transmitted over the channel with the best quality of service. Furthermore, in some embodiments of the present invention, values α1, . . . , αK might be set so that a very small fraction, or even zero, traffic is transmitted over one or more particular channels. In the case that zero traffic is transmitted over a particular channel (i.e., αi=0 for some 1≦i≦N), it might be advantageous to periodically set αi>0 so that the quality of service of the channel is periodically measured.
At task 370, source node 101 transmits traffic via communications channels 102-1 through 102-K in accordance with partition coefficients α1, . . . , αK (i.e., the traffic is distributed among the channels such that communications channel 102-i receives fraction αi of the traffic, for all i∈{1, . . . , K}).
As will be appreciated by those skilled in the art, although in the illustrative embodiment tasks 310, 360, and 370 are performed by source node 101, in some other embodiments of the present invention some or all of these tasks might be performed by some other entity, such as a third node that is capable of communicating with source node 101 and receiving node 103. Similarly, although in the illustrative embodiment tasks 320 through 350 are performed by receiving node 103, in some other embodiments of the present invention some or all of these tasks might be performed by some other entity—perhaps the same as the entity that performs tasks 310, 360, and 370, or perhaps yet another entity (e.g., a fourth node, etc.). In any case, it will be clear to those skilled in the art, after reading this disclosure, how to make and use such alternative embodiments of the present invention.
It is to be understood that the disclosure teaches just one example of the illustrative embodiment and that many variations of the invention can easily be devised by those skilled in the art after reading this disclosure and that the scope of the present invention is to be determined by the following claims.
