Method and device for efficient bandwidth management

A method comprising the steps of:

• in a packet network, adjusting an average of a predetermined parameter associated with bandwidth control to have a value within a predetermined stable region;
• and
• via said adjusting step, dynamically affecting allocated bandwidth for parent classes of traffic;
• wherein the predetermined parameter is related to a number of borrowing attempts during a predetermined observation window time.

A method comprising the steps of:

• in a packet network adjusting an average of a predetermined parameter associated with bandwidth control to have a value within a predetermined stable region;
• and
• via said adjusting step, dynamically affecting allocated bandwidth for parent classes of traffic;
• further comprising:
• measuring the predetermined parameter as a number of borrowing attempts during a predetermined observation window time.

A method comprising the steps of:

• in a packet network, adjusting an average of a predetermined parameter associated with bandwidth control to have a value within a predetermined stable region;
• and
• via said adjusting step, dynamically affecting allocated bandwidth for parent classes of traffic;
• further comprising:
• calculating the average of the predetermined parameter from a number of borrowing attempts using exponential smoothing.

A method comprising the steps of:

• in a packet network, adjusting an average of a predetermined parameter associated with bandwidth control to have a value within a predetermined stable region;
• and
• via said adjusting step, dynamically affecting allocated bandwidth for parent classes of traffic;
• further comprising:
• calculating the average of the predetermined parameter from a number of borrowing attempts according to the formula, Ai—avg(1−α)*Ai—avg+αAi,
• wherein Ai—avg represents the average of the number of borrowing attempts, Ai is a number of borrowing attempts during a most recent measurement window, a value of α is preselected as a negative power of two, and Ai—avg is updated every observation window, a pre-determined parameter in seconds.

A method comprising the steps of:

• in a packet network, adjusting an average of a predetermined parameter associated with bandwidth control to have a value within a predetermined stable region;
• and
• via said adjusting step, dynamically affecting allocated bandwidth for parent classes of traffic;
• wherein the predetermined parameter is related to a number of borrowing attempts and the predetermined stable region comprises a range bounded according to:
• If Ai—avg>Thr(Ai)lower, Bi=maximum(Bi−ωidown, Min(Bi))
• Else If Ai—avg>Thr(Ai)lower, Bi=minimum(Bi+ωiup, Max (Bi))
• Where Ai is a number of borrowing attempts during a most recent measurement window and
• Ai—avg is the average of the number of borrowing attempts/controlled state;
• upper and lower thresholds for the predetermined stable region are preset at predetermined values:
• Thr(Ai)lower is a lower threshold for borrowing attempts, where Thr(Ai)lower>0;
• Thr(Ai)upper is an upper threshold for borrowing attempts, where Thr(Ai)upper>Thr(Ai)lower;
• increment and decrement units ωidown and ωiup, which denote an update granularity on allocated bandwidth Bi, are preset at predetermined values;
• Max(Bi) is a maximum value of allocated bandwidth;
• Min(Bi) is a minimum value of allocated bandwidth;
• and
• exponential smoothing technique is used as follows,
• Ai—avg(1−α)*Ai—avg+α*Ai,
• wherein a value of α is preselected as a negative power of two and Ai—avg is updated every observation window, a pre-determined parameter in seconds.

A method comprising the steps of:

• in a packet network, adjusting an average of a predetermined parameter associated with bandwidth control to have a value within a predetermined stable region;
• and
• via said adjusting step, dynamically affecting allocated bandwidth for parent classes of traffic;
• further comprising:
• calculating the average of the predetermined parameter as a queue length according to the formula,
• Qi—avg(1−α)*Qi—avg+α*Qi,
• wherein Qi—avg represents the average of the queue length, Qi is an instnataneous measurement of queue length, a value of α is preselected as a negative power of two, and Qi—avg is updated every observation window, a pre-determined parameter in seconds.

A method comprising the steps of:

• in a packet network, adjusting an average of a predetermined parameter associated with bandwidth control to have a value within a predetermined stable region;
• and
• via said adjusting step, dynamically affecting allocated bandwidth for parent classes of traffic;
• wherein the predetermined parameter is related to a queue length and the predetermined stable region comprises a region bounded according to:
• If Qi—avg<Thr(Qi)lower, Bi=maximum(Bi−ωidown, Min(Bi))
• Else If Qi—avg>Thr(Qi)upper, Bi=minimum(Bi+ωiup, Max(Bi)),
• Qi is an instantaneous measurement of queue length;
• Qi—avg is a calculated average value for an average queue length;
• upper and lower thresholds are preset at predetermined values:
• Thr(Qi)lower is a lower threshold for queue length, where Thr(Qi)lower>0;
• Thr(Qi)upper is an upper threshold for queue length where Thr(Qi)upper>Thr(Qi)lower;
• increment and decrement units ωidown and ωiup, which denote an update granularity on allocated bandwidth Bi, are preset at predetermined values;
• Max (Bi) is a maximum value of allocated bandwidth;
• Min (Bi) is a minimum value of allocated bandwidth;
• and
• a queue size for class i, which is also an upper bound for a queue length Qi, is related to an upper bound on a delay jitter as:
• Delay_jitter—i=maxQi/Bi
• wherein, if a jitter upper bound Delay_jitter_i is preselected, then a linear relationship exists between maxQi and Bi, an allocated bandwidth.