Error performance of atm networks

Abstract

In this paper, we describe the calculation of the performance parameters of a point-to-point ATM connection transported on an STS-3c SONET (SDH STM-1) interface on the basis of the physical layer model derived from experimental data. An estimate for Cell Error Ratio (CER) was presented Wil] based on the objectives specified in the draft Recommendation G.82X,21 for Constant Bit Rate (CBR) channels. The general question of error performance in ATM layer is revisited here and parameters defined in draft Recommendation 1.3513131 including CER are calculated. 'Mese parameters are Cell Error Ratio (CER), Cell Loss Ratio (CLR), and Severely Errored Cell Block Ratio (SECBR). Statistics of consecutive errored and lost cells are likewise calculated for each run of the simulation. The physical layer performance parameters such as Bit Error Ratio (BER), Errored Second Ratio (ESR), and Severely Errored Second Ratio (SESR) are also calculated for comparison with corresponding objective values for CBR channels. In order to estimate the values of these parameters, an accurate characterization of the underlying physical layer is necessary. This characterization should be based on the behavior of the real channels in digital networks and satisfy the objective values indicated in the standards. [31 The bursty nature of errors in the physical layer is described by the Mdden Markov Model (HMM) [41 151 [61 171181191 the parameters of which are estimated using experimental data. The importance of using this model is demonstrated by the results presented in this paper: the ATM performance for channels with independent errors differs significantly from the ATM performance for channels with bursts of errors and equal bit error ratios. The performance parameters are evaluated both analytically and using computer simulation. The simulation program and analytical expressions that we have developed can be used to investigate the critical values of the parameters of the model in a top-down fashion. Simulation results can also be used to determine bounds on the objective values for the performance parameters of the ATM layer. Extensions of the simulation provide insight into the requirements of the physical layer, the ATM layer, and the layers using the services of the ATM layer, to achieve desired quality of service for applications using the ATM network. In our calculations, losses due to policing and congestion control are not included. Cell loss calculated here is only due to errored bits in the ATM cell headers. Severely Errored Cell Blocks are excluded ftom calculation of CER and CLR as required by 1.35B. Block size is assumed to be equal to an STS-3c SONET frame. Simulation results show that CER values are several orders of magnitude higher for the simulation runs using channels with independent error bits than for the channels with bursts of errors. The scattered error bits contribute to the total count of errored cells in the case of channels with independent error bits. Cell loss, on the other hand, is much higher for the the channels with bursts of errors, because the probability of uncorrected errors in the cell header is much higher in this case. It is also shown that the density of bit errors in the burst state and the average length of each state has a strong effect on the interpretation of the error performance parameters. For instance excessive error densities in the short burst state can render. CER meaningless due to the effect of cell loss as well as the SECBR.