A simulation study of the Adaptive RIO (A-RIO) queue management algorithm

Building a controlled delay assured forwarding class in differentiated services networks

“…For an assured rate of 'x%', the committed information rate (CIR) also known as the 'green rate' of each source is calculated as follows: greenRate = x * bottleneckBandwidth/(100 * numSources). The values of other parameters such as peak information rate (PIR), committed burst size (CBS) and peak burst size (PBS) were chosen in accordance with those in [14], where P IR = 2 * CIR, CBS = 1.5 * CIR and P BS = 1.5 * P IR.…”

“…DRED-MP and ARIO-D have a similar objective as that of ARIO but both employ staggered thresholds to achieve their objective. These approaches may however suffer from problems arising out of 3 One set of values for the RED parameters do not suffice for all network conditions using staggered thresholds which have been documented in [14]. Additionally, DRED-MP has several tunable parameters (L, Lg, Ly, Lr, averaging weight β and control gain α) and no guidelines to tune these have been provided.…”

“…Even though this mechanism succeeds in achieving relative differentiation among the several precedences within a queue, it is plagued by the parameter sensitivity problem 3 [13] [7]. Several schemes such as ARIO [14], DRED-MP [1], ARIO-D and ARIO-L [12] STAVQ [3], Fuzzy logic based AQM [4], have been proposed to overcome the shortcomings of RIO. Most of these schemes are in some way (directly or indirectly) a variant of RIO with support for adaptive parameter tuning.…”

“…These predicted values then feed into the packet drop probability computation function. The performance of PAQMAN-DS has been evaluated against the RIO [5] and ARIO [14] schemes through ns-2 simulations. Results from these simulations show that irrespective of the traffic load, PAQMAN-DS tightly regulates the delay below the control target unlike RIO and ARIO which fail to achieve this.…”

A simulation study of a new rate‐and‐queue‐based active queue management algorithm

“…The probability of dropping a packet from a particular connection is roughly proportional to that connection's share of the bandwidth through the gateway. In other words, there are no precise rules for tuning those parameters (Orozco et al , 2005) and most published results point at the difficulty of finding a robust RED configuration (Bonald et al , 2000; Firoiu and Borden, 2000). Lin and Moms (1997) observed that RED allows unfair bandwidth sharing when a mixture of different traffic shares a link.…”