Point Process Approaches to the Modeling. and Analysis of Self-Similar Trac { Center for Telecommunications Research

Size: px
Start display at page:

Download "Point Process Approaches to the Modeling. and Analysis of Self-Similar Trac { Center for Telecommunications Research"

Transcription

1 Proc. IEEE INFOCOM '96, San Francisco, CA, March 996. Point Process Approaches to the Modeling and Analysis of Self-Similar Trac { Part I: Model Construction Bong K. Ryu Steven B. Lowen Department of Electrical Engineering and Center for Telecommunications Research Columbia University, New York, NY 27 fryu,steve6g@ctr.columbia.edu Abstract We propose four fractal point processes (FPPs) as novel approaches to modeling and analyzing various types of self-similar trac: the fractal renewal process (FRP), the superposition of several fractal renewal processes (Sup-FRP), the fractal-shot-noise-driven Poisson process (FSNDP), and the fractal-binomialnoise-driven Poisson process (FBNDP). These models fall into two classes depending on their construction. Study of these models provides a thorough understanding of how self-similarity arises in computer network trac. We nd that (i) all these models are (second-order) self-similar in nature; (ii) the Hurst parameter alone does not fully capture the burstiness of a typical self-similar process; (iii) the heavy-tailed property is not a necessary condition to yield selfsimilarity; and (iv) these models permit parsimonious modeling (using only 2{5 parameters) and fast simulation. Simulation veries that these models exhibit fractal behavior over a wide range of time scales. Introduction It is expected that future high-speed networks will support a wide variety of services that exhibit extremely diverse trac characteristics. Characterizing the statistical behavior of a trac source is crucial to the proper design of high-speed networks, ensuring that they provide the pre-negotiated quality of service (QOS) to users while achieving high network utilization. Traditional trac models, most of which assume Markov characteristics, have been used extensively; in many cases they prove adequate for evaluating network performance (see, for example, [3, 5] and references therein). However, traditional Markov-based models often prove inadequate to the task of eciently characterizing the diverse trac now encountered over current teletrac networks [, 7, 8, 2, 22]. Recent analyses of high-quality trac measurements have revealed the prevalence of self-similarity (or long-range dependence) in data [7] and compressed video []. Despite the signicant attention self-similarity has received recently, a clear picture of its nature and importance has not yet emerged [3, 4, 6, 7, 7, 9]. Different manifestations and instances of self-similarity yield dierent results. This appears to stem from the nature of self-similar processes themselves. This study focuses on explaining the dynamics of individual arrivals giving rise to (second-order) selfsimilarity, using a point processes formulation. Thus, the fractal point processes (FPPs) introduced in this paper reveals how self-similarity arises from individual packet arrivals in the packet-switching networks. They also enjoy the properties of parsimonious modeling and fast simulation, and have direct applications in modeling Ethernet trac using TCP/IP (data [8] and video [2]) as well as long-range dependent VBR video trac over ATM [9]. We begin by introducing the family of fractal point processes (FPPs), whose increments processes are (second-order) self-similar. These models naturally fall into two classes, one based on renewal processes, and the other on doubly stochastic Poisson processes (DSPPs). Next, we construct and analyze four FPP models: the fractal renewal process (FRP), the superposition of several fractal renewal processes (Sup- FRP), the fractal-shot-noise-driven Poisson process (FSNDP), and the fractal-binomial-noise-driven Poisson process (FBNDP). Finally, we present simulation results for all FPPs, including comparison with theory. 2 Fractal Point Processes (FPPs) 2. Denition We call a point process fractal when a number of the relevant statistics exhibit scaling with related scaling exponents, indicating that the represented phenomenon contains clusters of points over all (or a relatively large set of) time or length scales [4, 5]. Such scaling leads mathematically to power-law dependencies in the scaled quantities [, 4]. Thus fractals and power-law forms of their statistics are closely related. Each statistic which scales will therefore provide an exponent; for a (mono-) fractal process all are simply related, yielding a single fractal exponent for the process. For a general point process, fractal scaling in one statistic does not necessarily imply fractal scaling in other statistics; if scaling exists in only one statistic, then we do not call this process fractal [, 4].

2 All the FPP models we describe in this paper are stationary, and we therefore consider only stationary point processes in the following analysis. 2.2 Statistical measures Important second-order statistics for an FPP are the power spectral density (PSD), coincidence rate (CR), index of dispersion for counts () [also called the Fano factor], and count-based covariance function (COV). We employ the PSD, S(!), derived directly from the point process itself, and not from the sequence of arrivals recorded in adjacent counting windows of equal duration. The CR measures the correlation between pairs of arrivals (events) with a specied time delay between them, regardless of intervening arrivals, and is related to the autocorrelation function used with continuous processes [2]. It is dened as PrfE(; ) and E(; + )g G() lim ; ()! 2 where E(s; t) denotes the occurrence of at least one event of the point process in the half-open interval [s; t). A particularly useful statistic is the F (T ), which is dened as the variance of the number of arrivals in a specied time window of width T divided by the mean number of arrivals. Finally, the COV C(k; T ) is dened as the covariance between the number of arrivals in two counting windows of counting time T and separation kt. If we dene X k N(kT )? N[(k? )T ], where N(t) represents the number of arrivals up to time t, then C(k; T ) Cov(X n ; X n+k ): (2) These four second-order statistics (, PSD, COV, CR) may be obtained from each other by means of the following relations [4, 2] Z T F (T ) = (T )? (T? jj) G()? 2 d?t Z S(!) =? G()e?j! d (3) C(k; T ) = Z T?T (T? jj) G(kT + )? 2 d; valid for any regular point process, where is the expected rate. 2.3 Fractal nature of FPPs For fractal point processes with fractal exponent < <, the second-order statistics will have special forms, and in the case of a purely fractal process all of the following relationships hold [4, 2]: G()= 2 = + (jj= )? + ()= S(!)= = + (j!j=! )? + (!=2) F (T ) = + (T=T ) (4) + (T=T ) C(k; T ) = T k = ; (T=T ) r 2 (k + )=2 k > ; with! T = cos(=2)?( + 2)? T = ( + )=2; (5) where R (x) is the Dirac delta function,?(x) e?t t x? dt is the gamma function, and r 2 (f(k)) f(k + )? 2f(k) + f(k? ) is the second central difference operator. The second line of (4) implies =f noise, and the last line of (4) also implies that the autocorrelation function () r(k; T ) is given by r(k; T ) C(k; T )=C(; T ) T = T + T 2 r2 (k + ) (k > ); (6) making the process X = fx n g long-range dependent with g(t ) T =(T + T ) and H = ( + )=2 [2]. For very small T, i.e. T T, the prefactor g(t ) approaches zero, so that the LRD property becomes negligible at this time scale. For large values of T, on the other hand, g(t ) approaches unity, increasing the degree of the long-range dependence. Thus (6) shows how the time scale aects the LRD property of a process X; the Hurst parameter H alone is insucient for characterizing the long-range dependence of an FPP. Broadly speaking, the time T marks the lower limit for signicant scaling behavior in the and. For this reason, we call this parameter the fractal onset time. We note that a purely fractal process, one which exhibits scaling over all time and frequency ranges as in (4), presents mathematical diculties; these processes have innite power, for example. This diculty is obviated in practice by restricting the validity of (4) to a nite range of times and frequencies, which must be the case for all experimental data. In this case, if any one of the relationships in (4) holds over the relevant time and frequency ranges then the other three must also, with the parameters again given by (5). Thus for FPPs all the second-order statistics exhibit power-law scaling with closely related exponents. This means that to second order, for < <, and over the specied ranges, three parameters suce to specify an FPP: the mean rate, the fractal exponent, and the fractal onset time T. [Equivalently, either w and can be specied instead of T, since any one species the other two via (5).] 2.4 Self-similar nature of FPPs As a further illustration of the fractal qualities of FPPs, we show that they are indeed (second-order) self similar for fractal exponents in the range < <. Let X n (m) m P? km n=(k?)m+ X n (concatenating adjacent counting intervals). Then we have C (m) (k; T ) = m?2 C(k; mt ), yielding r (m) (k; T ) = r(k; mt ): (7) Combining the results from (6) and (7) results in r (m) (k; T ) = (mt ) T + (mt ) 2 r2 (jkj + ): (8)

3 As m!, the second term of the denominator of (8) will dominate, resulting in lim m! r(m) (k; T ) = 2 r2 (jkj + ): Comparing this with (6), it is easy to see that X has a non-degenerate correlation structure and therefore is asymptotically (second-order) self-similar. In addition, since C (m) (; T ) = Var(X (m) ), we have i Var(X (m) ) = T hm? + (T=T ) m?(?) ; so that the variance of X (m) varies as m?(?) for large m. Thus the process X has the slowly-decaying variance property, another mathematically equivalent manifestation of self-similarity [7]. Therefore, the increments process X constructed from all FPPs with < < are asymptotically second-order self-similar with Hurst parameter H = ( + )=2. 3 Two FPP Construction Methods FPPs are characterized by power-law behavior in their second-order statistics, as shown in (4). Analytical relations exist to generate all these statistics from each other, as given by (3); thus mathematically all are equivalent. We employ the coincidence rate in the following, since this proves easier analytically; we reiterate the form for a FPP given in the rst line of (4): G() = 2 [ + (jj= )? ] + (): (9) 3. Renewal point process method A renewal point process by denition has i.i.d. interarrival times; thus the interarrival time pdf completely species the process. If this distribution is heavy tailed, then the coincidence rate G() will also decay in a power law form, as given by (9). This yields the fractal renewal point process (see Sec. 4); the superposition of a number of independent and identical realizations of this process also has a coincidence rate of the same form, and therefore also belongs to the FPP family of processes (see Sec. 5). Since the behavior of the tail of the interarrival time pdf determines the power-law form of the CR and hence the fractal nature of the FPP, the behavior of the interarrival time pdf at short times is arbitrary. Thus a wide variety of renewal-based FPPs exist for a given average rate and power-law exponent. 3.2 Doubly stochastic Poisson process (DSPP) method The doubly stochastic Poisson point process (DSPP) method derives from the similarity between the CR of a DSPP and the autocorrelation function of its rate. To show this, let I(t) denote the stationary stochastic rate of a DSPP, and R I () denote the autocorrelation function of this rate. Then, for 6= we have PrfE(; ) and E(; + )g G() lim ;! 2 E [PrfE(; ) and E(; + )j Ig] = lim! 2 I(t)I(t + ) = lim E! = E[I(t)I(t + )] R I (): 2 Thus if a stationary continuous-time stochastic process I(t) with an autocorrelation function R I () having the form of (9) serves as the rate for a DSPP, the result will be a FPP. In the following we consider two such continuous-time stochastic processes: Fractal Binomial Noise (FBN) and Fractal Shot Noise (FSN). FBN consists of the superposition of several independent and identical fractal ON/OFF processes whose sojourn times are heavy-tailed (see Sec. 6). FSN is a type of shot noise [6] where the linear lter assumes a power-law decaying form (see Sec. 7). In the following we present several FPP models constructed by the above approaches which may be employed to generate data mimicking self-similar trac. For each model we rst identify the relevant parameters, then analyze how these parameters determine the three fundamental quantities:, H, and T. In particular, we determine whether all three of these quantities can be independently specied. Finally, we consider the remaining parameters (those in excess of the three needed to specify, H, and T ), if any, and their eects on the character of the FPP. Such eects involve statistics over short time scales and clustering property over all (or relatively large range of) time scales, a salient feature of FPPs 4 Fractal Renewal Process (FRP) Perhaps the simplest FPP is the standard fractal renewal process [3]. For the standard FRP, the times between adjacent arrivals are independent random variables drawn from the same pdf. Fig. -(a) provides a schematic representation of this point process. In particular, the pdf decays as a power law p 2 (t) = kt?(+) for A < t < B, () otherwise, where A and B are cuto parameters, is a fractal exponent ( < < 2), and k is a normalizing constant determined by the requirement R p 2(t) dt =. For < < this process is fully fractal: the power spectral density, coincidence rate, index of dispersion for counts, and even the interarrival time pdf all exhibit power-law scaling as in (4) over time scales lying between A and B, and with related power-law exponents completely determined by =. For < < 2 the PSD,, CR, and (but not the interarrival time pdf) still exhibit power-law scaling of the form of (4), but with an associated exponent given by = 2?. For gamma 2 the process no longer has fractal second-order statistics. Thus is limited to a range between zero and unity, and for each value of there corresponds two values of. In practice, however, the range < < 2 proves

4 R (a) Fractal Renwal Process (b) Fractal ON/OFF Process Figure : The standard FRP and alternating FRP (fractal ON/OFF process) Models. superior. For < < the resulting simulations are extremely bursty and do not mimic practical traf- c data, and sample statistics do not reliably follow the analytic forms of (4) except for prohibitively large simulation lengths. Another advantage of employing > is that it renders the outer cuto B unnecessary; setting B! leads to a positive mean rate, in contrast to the < case. Eliminating the outer cuto also yields better power-law behavior in the PSD and, and simplies simulation [3]. The probability density function then assumes the simpler form for t A, p 2 (t) = A t?(+) () for t > A. Choosing in the range < < 2 proves far superior to < < for the same desired value of, but the form of the interarrival pdf in () can be further improved. In particular, the resulting F (T ) has a dip near T = T, caused by the abrupt cuto in the interarrival time pdf that still remains; furthermore, the power spectral density exhibits excessive oscillations for the same reason. Improvement results from employing a smoother pdf [4] A p 2 (t) =? e?t=a for t A, e? A t?(+) (2) for t > A, which is continuous for all t. The practical FRP model as presented in (2) has only two parameters: and A; thus this model cannot fully specify the set of fundamental quantities, H, and T. Thus the FRP model only proves useful for simulating data with T of the order of unity. Using the relation = 2? we obtain H = ( + )=2; = + (? )? e?? A? ; T = 2??2 (? )? (2? )(3? )e? [ + (? )e ] 2 A : (3) 5 Superposition of Fractal Renewal Processes (Sup-FRP) A dierent fractal point process results from the superposition of a number of independent and identical FRPs. Although the resulting FPP (Sup-FRP) is no longer renewal, the marginal distribution of the interarrival times is still heavy-tailed [2]. Furthermore, the corresponding PSD, CR,, and retain their scaling behavior, albeit over a somewhat reduced range of times and frequencies. With each component FRP described by the interarrival time pdf given in (2), three parameters characterize the Sup-FRP: and A from the individual FRPs, and M, the number of FRPs superposed. In this case, the three fundamental quantities become [3] H = ( + )=2 = M + (? )? e?? A? ; T = 2??2 (? )? (2? )(3? )e? [ + (? )e ] 2 A : (4) with = 2?. The only dierence between (4) and (3) is a factor of M in the expression for ; the quantities H and T remain unchanged. The Sup- FRP model indeed has three parameters, although M can take only positive integer values. Thus arbitrary values of, H, and T cannot be achieved exactly. In practice, we specify and H, and adjust M to approximate T as closely as possible. The integer M is of the order T ; for most trac data this quantity greatly exceeds unity, so that the nite resolution of M does not pose a signicant problem. 6 Fractal-Binomial-Noise-Driven Poisson Process (FBNDP) The fractal renewal process (FRP) described previously is recast as a real-valued process which alternates between two values, such as zero and R (> ). This alternating fractal renewal process (AFRP) would then start at a value of zero (\OFF"), and then switch to a value of R (\ON") at a time corresponding to the rst event in the FRP. At the second such event, the AFRP would switch back to zero, and would proceed to switch back and forth at every successive event of the FRP. Thus for the AFRP, all ON/OFF periods are i.i.d. with the same heavy-tailed distribution as in the FRP. Figure -(b) illustrates such a process. As with the Sup-FRP, M independent and identical fractal ON/OFF processes may be added together, yielding fractal binomial noise (FBN) with the same fractal exponent as the single fractal ON/OFF process [3]. Let I(t) denote the resulting FBN process. The autocorrelation function of I(t) is merely a scaled version of the of the individual AFRP processes, which in turn follow the decaying power-law form of (9) [3]. Thus I(t) can serve as a stationary stochastic rate function for a Poisson process, resulting in the fractal-binomial-noise-driven Poisson point process (FBNDP). The construction of the FBNDP is schematized in Fig. 2. Since I(t) is a binomial process, and the ON and OFF periods have identical mean values, we have PrfI(t) = nrg = 2?M M! n!(m? n)! (5) for n = ; ; 2; : : :M. The FBNDP has four free parameters (A,, R, and M) which determine, H and T as follows, where we

5 Fractal-Binomial-Noise-Driven Poisson Process (FBNDP) R R Poisson Generator N(t) I(t) Figure 2: The FBNDP Model. FBN homogeneous, but rather reects the variations of the fractal-shot-noise driving process. Thus the two forms of randomness inherent in the DSPP are, in the particular case of the FSNDP, two separate Poisson processes, linked by a power-law-decaying linear lter. As a result of this lter, the FSNDP exhibits fractal behavior of the form of (4) for time scales in the range A T B. The FBNDP and the FSNDP thus both belong to the fractal DSPP family, comprising a realvalued fractal rate function driving a Poisson point process; they dier only in how the fractal rate functions are constructed. The FSNDP model does not involve the heavy-tailed property in the same manner that the FRP, Sup-FRP, and FBNDP models do, since the interarrival time statistics of both Poisson processes in the FSNDP model have exponential tails. use the same density as in (2) for the ON and OFF periods of the individual AFRP processes [2]: H = ( + )=2; = RM=2; T = ( + )(2? )? R? A? [(? )e 2? + ]: (6) Thus, H, and T can all be specied, and in fact an extra parameter exists, implying that dierent FB- NDPs can be constructed with the same, H, and T. For example, decreasing M (while increasing R to keep the overall rate constant) will increase the probability that the rate becomes zero [see (5)], during which no arrivals can occur. Since the OFF period is also heavy-tailed, the resulting FBNDP exhibits a high degree of clustering, especially for the limiting case M =. For this value of M, heavy-tailed periods of arrivals will alternate with heavy-tailed inter-burst quiescent periods. Similarly, increasing the value of M reduces the degree of clustering. 7 Fractal-Shot-Noise-Driven Poisson Process (FSNDP) The fractal-shot-noise-driven Poisson point process (FSNDP) [2] is a special case of the doubly stochastic Poisson process (DSPP). For the FSNDP, the rate of the inhomogeneous Poisson process is fractal shot noise [], which is itself a ltered version of a different, homogeneous Poisson point process. Figure 3 schematically illustrates the FSNDP as a two-stage stochastic process. The rst stage is a homogeneous Poisson process (HPP) with constant rate. Its output M(t) becomes the input to a linear lter with a power-law decaying impulse response function h(t) = c=t?=2 for A < t < B, otherwise, (7) with, A, and B dened as following (), and c a positive amplitude constant. This lter produces fractal shot noise I(t) at its output, which then becomes the stochastic rate for the last stage, a second Poisson point process. The resulting process N(t) is not Rate (µ) Poisson Generator Linear Filter h(t) Fractal-Shot-Noise-Driven Poisson Process (FSNDP) N(t) M(t) I(t) Poisson Generator t Homogeneous Poisson Process Fractal Shot Noise FSNDP Events Figure 3: The FSNDP Model. The FSNDP model has ve parameters: A, B,, k, and, which determine H,, and T as follows [2]: H = ( + )=2 = 2? cb =2 T ( + )?(? =2) = 2?( + =2)?(? ) c? B =2 : t t t (8) Since A does not appear in (8), in many applications it can be set to zero, which results in a rate function I(t) with innite variance. This proves not be problematic, however, since an integrated version of the shot noise (which necessarily has a nite variance), is more directly involved in FSNDP statistics. Four parameters remain, so that as with the FBNDP a free parameter exists which determines the clustering characteristics of the process; we focus on the product B. For small values of this quantity, successive impulse response functions rarely overlap, leading to large gaps between impulse response functions where FSNDP arrivals cannot occur. Thus the process appears more clustered. 8 Simulation of FPPs We performed simulations on the FRP, Sup-FRP, FSNDP, and FBNDP models; for the FBNDP we employed two sets of parameters. For each FPP model, we set = arrivals/sec and H = :9; we also set

6 ( x ) 2 4 ( x ) 2 4 ( x ) 2 4 ( x ) 2 4 ( x ) 2 4 FRP Sup-FRP with M = FSNDP with B = ^3 sec FSNDP with B = ^5 sec FBNDP with M = FBNDP with M = ( x ) 2 ( x ) 2 ( x ) 2 ( x ) 2 ( x ) 2 FRP Sup-FRP with M = FSNDP with B = ^3 sec FSNDP with B = ^5 sec FBNDP with M = FBNDP with M = ( x ) ( x ) Figure 4: Sample path comparison of FPPs. Parameters as in Table except for the values of B in FSNDP simulations. Left: counting time T = : sec, Right: T = sec. Model A (msec) B (sec) M R FRP.48.8 Sup-FRP FSNDP?4 7.8 FBNDP # FBNDP # Table : Parameter values of the resulting FPP model for each of the ve FPPs simulated from the specication = (points/sec), H = :9, and T = msec. For all simulations, the simulation length L is 5 3 sec, yielding an expected number of arrivals of 5 6. T = msec, or as close as possible to that value. For the FRP, setting and H as above results in T = :2 msec. Since the parameter M in the Sup-FRP can take only integer values, the closest value in this case is T = 9:6 msec, achieved with M = 8. For the FSNDP, we set A =?7 sec and B = 7 sec to minimize the eects of fractal cutos on the estimated fractal exponent [2], resulting = 9:6?5 and c = 6:6 3 from (8). The value of B in particular must exceed the simulation length; this value provides a comfortable margin of 2, leading to only small deviations from ideal fractal behavior [2]. Finally, for the FBNDP we simulated two cases: M = 2 (FBNDP #) and M = (FBNDP #2). Table presents the parameter values derived from the simulation specication for each of the ve models simulated. Fig. 4 shows the sample path behavior of these processes at two dierent time scales. For the FSNDP we use B = 3 sec and B = 5 sec to make shot-noise uctuations more visible. Notice the sharp dierence in burst structure among models. This further illustrates that the Hurst parameter alone does not capture the burstiness of LRD trac. Model b (points/sec) b H CPU (min) FRP Sup-FRP FSNDP FBNDP # FBNDP # Table 2: Estimated parameters from the FPP simulations, and the averaged time required to perform each run. 95% condence intervals are included for the rate estimate. For each of the ve FPPs simulations listed in Table we performed independent runs. Fig. 5 presents the resulting estimated F(T b ) and br(k; T ) for each FPP model; means and 95% condence intervals are shown. We chose a counting time of T = msec (= T ) for the estimated br(k; T ), so that power-law behavior would be readily apparent at this time scale. The analytical values F (T ) and r(k; T ) are also included, obtained from (4) and (6). Table 2 lists the three estimated fundamental parameters corresponding to these ve FPP simulations. The

7 estimated fractal exponent b was obtained by a leastsquares t to a doubly logarithmic plot of the, log[ F(T b )? ] vs. log(t ), between counting times of T = : msec and T = sec, with two samples per decade. We obtained the estimated Hurst exponent by the simple relation H b = (b + )=2. The simulation results support the notion that the four models indeed generate fractal point processes with the desired rate of arrivals and the Hurst parameter. Table 2 shows that the estimated rate follows the desired value of = closely and with little variation for all ve simulated FPP models. The estimated Hurst parameter H b also closely follows the original value.9. However, Fig. 5 illustrates that some deviation between simulated and analytical results occurs in the and. The dierence in the is systematic with all simulated points falling below the analytical curve, and roughly paralleling it. The simulated FPP results from all ve plots also differ from the analytically predicted results, as shown in Table 2, although they all follow the form of (4). This bias appears to be intrinsic to the fractal nature of the processes [4], varying among the FPP models and even between the two versions of the FBNDP simulated. See [2] for detailed analysis of Fig. 5 and the bias. Despite the various biases and variances, all FPP models simulated indeed generate stochastic point processes which are fractal, and thus exhibit power-law behavior. Each model has its own distinctive characteristics, and may prove best in simulating a particular fractal trac source; all are useful. Finally, we consider the eciency of these ve FPP simulation methods. The last column of Table 2 shows the average CPU times for each FPP simulation of 5 million arrivals on an SGI Challenger XL running IRIX 5.3. Three of the simulations require about two minutes, or 24 sec per event. The FBNDP # requires the longest time to simulate, for an extremely large number of power-law computations are required due to the small value of A. Thus, FPP models can generate large sets of fractal data eciently on modern workstations. 9 Concluding Remarks We have introduced four fractal point processes (FPPs) which provide an improved understanding of self-similar trac observed in current computer networks: the Fractal Renewal Process (FRP), Superposition of i.i.d. FRPs (Sup-FRP), the Fractal- Shot-Noise-Driven Poisson Process (FSNDP), and the Fractal-Binomial-Noise-Driven Poisson point Process (FBNDP). Some of these models appear to exhibit greater variability than earlier models [8, 23, 24]. We have shown that these models fall into two classes, renewal-based and DSPP-based, depending on their construction, and how the heavy-tailed property need not relate to the fractal nature of these models. Perhaps most importantly, the Hurst parameter H alone does not suce in describing the burstiness and clustering in a typical self-similar process, despite assertions to the contrary [7]. The FPP models presented in this paper cover a Analysis Simulation (a) FRP Model Lag k (Time Unit T =. sec) (b) Sup-FRP Model Lag k (Time Unit T =. sec) (c) FSNDP Model.... (d) FBNDP Model # (M = 2).... (e) FBNDP Model #2 (M = ) Lag k (Time Unit T =. sec) Lag k (Time Unit T =. sec) Lag k (Time Unit T =. sec) Figure 5: Results of simulations of ve FPP models. Left: doubly logarithmic plot of the estimated bf(t ) vs. counting time T. Right: plot of the estimated br(k; T ) vs. lag k. The counting time T is : sec. Mean values obtained from the simulations are indicated by a dot, with 95% condence intervals delineated by error bars. Analytic values are included for comparison.

8 broad spectrum of applications ranging from Ethernet trac modeling [2] to analysis of ATM networks [9]. Two issues are being considered relevant to these models: (i) generating unbiased synthetic fractal data; and (ii) queueing analysis. Both of which will be treated in future publications. With these issues fully studied and resolved, the FPP models will prove more useful in understanding and analyzing various types of selfsimilar trac in current and future high-speed networks. Acknowledgement Part of this work was completed while the rst author was on leave with AT&T Bell Laboratories, Murray Hill, New Jersey. He is grateful to Anwar Elwalid and Debasis Mitra for their generosity and support. References [] J. Beran, R. Sherman, M. S. Taqqu, and W. Willinger. Long-range dependence in variable-bit-rate video trac. IEEE Trans. Comm., 43:566{579, 995. [2] D. R. Cox and P. A. W. Lewis. The Statistical Analysis of Series of Events. Methuen, London, 966. [3] A. Elwalid, D. Heyman, T. V. Lakshman, D. Mitra, and A. Weiss. Fundamental bounds and approximations for ATM multiplexers with applications to video teleconferencing. IEEE JSAC, 3:4{6, 995. [4] A. Erramilli, O. Narayan, and W. Willinger. Experimental queueing analysis with long-range dependent packet trac. Submitted to IEEE/ACM Trans. Net., 994. [5] V. S. Frost and B. Melamed. Trac modeling for telecommunications networks. IEEE Comm. Mag., 32(3):7{8, March 994. [6] D. P. Heyman and T. V. Lakshman. What are the implications of long-range dependence for VBRvideo trac engineering? Preprint, 995. [7] W. E. Leland et al. On the self-similar nature of Ethernet trac (extended version). IEEE/ACM Trans. Net., 2:{5, 994. [8] N. Likhanov, B. Tsybakov, and N. D. Georganas. Analysis of an atm buer with self-similar (\fractal") input trac. In Proc. IEEE INFOCOM '95, 995. [9] S. B. Lowen. Fractal renewal processes as a model of charge transport in amorphous semiconductors. Phys. Rev. B, 46:86{89, 992. [] S. B. Lowen. Fractal Stochastic Processes. PhD thesis, Columbia University, New York City, 992. [] S. B. Lowen and M. C. Teich. Power-law shot noise. IEEE Trans. Inf. Th., 36:32{38, 99. [2] S. B. Lowen and M. C. Teich. Doubly stochastic Poisson point process driven by fractal shot noise. Phy. Rev. A, 43:492{425, 99. [3] S. B. Lowen and M. C. Teich. Fractal renewal processes generate /f noise. Phy. Rev. E, 47:992{, 993. [4] S. B. Lowen and M. C. Teich. Estimation and simulation of fractal stochastic point processes. Fractals, 3:83{2, 995. [5] B. B. Mandelbrot. The Fractal Geometry of Nature. W. H. Freeman, 982. [6] A. Papoulis. Probability, Random Variables, and Stochastic Processes. McGraw-Hill, New York, third edition, 99. [7] M. Parulekar and A. M. Makowski. Buer over- ow probabilities for a multiplexer with selfsimilar input. In Proc. IEEE INFOCOM '96, 996. [8] B. K. Ryu. Implications of self-simialrity for providing QOS guarantees end-to-end in high-speed networks: A framework of Application Level Trac Modeling. In Proc. Int. Zurich Symp. '96, 996. [9] B. K. Ryu and A. Elwalid. The importance of Long-Range Dependence of VBR video trac in ATM trac engineering: Myths and reality. Technical Memorandum, AT&T Bell Laboratories, 995. [2] B. K. Ryu and S. B. Lowen. Modeling, analysis, and generation of self-similar trac with the Fractal-Shot-Noise-Driven Poisson process. In IASTED Proc. Int. Conf. Modeling and Simulation '95, 995. [2] B. K. Ryu and S. B. Lowen. Point process approaches to the modeling and analysis of selfsimilar trac: Part I - model construction. Technical Report CU/CTR/TR , Ctr. Tel. Res. Columbia University, 995. URL: [22] B. K. Ryu and H. E. Meadows. Performance analysis and trac behavior of Xphone videoconferencing application on an Ethernet. In Proc. Third Int. Conf. Comp. Comm. Net., 994. [23] M. S. Taqqu and J. B. Levy. Using renewal processes to generate long-range dependence and high variability. In E. Eberlein and M. S. Taqqu, editors, Dependence in Probability and Statistics, volume, pages 73{89. Birkhauser, Boston, 986. [24] D. Veitch. Novel models of broadband trac. In Proc. GLOBECOM '93, 993.

The Importance of Long-Range Dependence of VBR Video Trac in. Bong K. Ryu

The Importance of Long-Range Dependence of VBR Video Trac in. Bong K. Ryu The Importance of Long-Range Dependence of VBR Video Trac in ATM Trac Engineering: Myths and Realities Bong K. Ryu Center for Telecommunications Research, Columbia University, New York, NY 10027 ryu@ctr.columbia.edu

More information

Multiplicative Multifractal Modeling of. Long-Range-Dependent (LRD) Trac in. Computer Communications Networks. Jianbo Gao and Izhak Rubin

Multiplicative Multifractal Modeling of. Long-Range-Dependent (LRD) Trac in. Computer Communications Networks. Jianbo Gao and Izhak Rubin Multiplicative Multifractal Modeling of Long-Range-Dependent (LRD) Trac in Computer Communications Networks Jianbo Gao and Izhak Rubin Electrical Engineering Department, University of California, Los Angeles

More information

In Proceedings of the Tenth International Conference on on Parallel and Distributed Computing Systems (PDCS-97), pages , October 1997

In Proceedings of the Tenth International Conference on on Parallel and Distributed Computing Systems (PDCS-97), pages , October 1997 In Proceedings of the Tenth International Conference on on Parallel and Distributed Computing Systems (PDCS-97), pages 322-327, October 1997 Consequences of Ignoring Self-Similar Data Trac in Telecommunications

More information

ESTIMATING SCALING EXPONENTS IN AUDITORY-NERVE SPIKE TRAINS USING FRACTAL MODELS INCORPORATING REFRACTORINESS

ESTIMATING SCALING EXPONENTS IN AUDITORY-NERVE SPIKE TRAINS USING FRACTAL MODELS INCORPORATING REFRACTORINESS ESTIMATING SCALING EXPONENTS IN AUDITORY-NERVE SPIKE TRAINS USING FRACTAL MODELS INCORPORATING REFRACTORINESS S.B. LOWEN Department of Electrical and Computer Engineering 44 Cummington Street Boston University,

More information

Packet Size

Packet Size Long Range Dependence in vbns ATM Cell Level Trac Ronn Ritke y and Mario Gerla UCLA { Computer Science Department, 405 Hilgard Ave., Los Angeles, CA 90024 ritke@cs.ucla.edu, gerla@cs.ucla.edu Abstract

More information

IP Packet Level vbns Trac. fjbgao, vwani,

IP Packet Level vbns Trac.   fjbgao, vwani, IP Packet Level vbns Trac Analysis and Modeling Jianbo Gao a,vwani P. Roychowdhury a, Ronn Ritke b, and Izhak Rubin a a Electrical Engineering Department, University of California, Los Angeles, Los Angeles,

More information

Network Traffic Characteristic

Network Traffic Characteristic Network Traffic Characteristic Hojun Lee hlee02@purros.poly.edu 5/24/2002 EL938-Project 1 Outline Motivation What is self-similarity? Behavior of Ethernet traffic Behavior of WAN traffic Behavior of WWW

More information

Exploring regularities and self-similarity in Internet traffic

Exploring regularities and self-similarity in Internet traffic Exploring regularities and self-similarity in Internet traffic FRANCESCO PALMIERI and UGO FIORE Centro Servizi Didattico Scientifico Università degli studi di Napoli Federico II Complesso Universitario

More information

A source model for ISDN packet data traffic *

A source model for ISDN packet data traffic * 1 A source model for ISDN packet data traffic * Kavitha Chandra and Charles Thompson Center for Advanced Computation University of Massachusetts Lowell, Lowell MA 01854 * Proceedings of the 28th Annual

More information

Capturing Network Traffic Dynamics Small Scales. Rolf Riedi

Capturing Network Traffic Dynamics Small Scales. Rolf Riedi Capturing Network Traffic Dynamics Small Scales Rolf Riedi Dept of Statistics Stochastic Systems and Modelling in Networking and Finance Part II Dependable Adaptive Systems and Mathematical Modeling Kaiserslautern,

More information

Asymptotic Delay Distribution and Burst Size Impact on a Network Node Driven by Self-similar Traffic

Asymptotic Delay Distribution and Burst Size Impact on a Network Node Driven by Self-similar Traffic Èíôîðìàöèîííûå ïðîöåññû, Òîì 5, 1, 2005, ñòð. 4046. c 2004 D'Apice, Manzo. INFORMATION THEORY AND INFORMATION PROCESSING Asymptotic Delay Distribution and Burst Size Impact on a Network Node Driven by

More information

Characterization and Modeling of Long-Range Dependent Telecommunication Traffic

Characterization and Modeling of Long-Range Dependent Telecommunication Traffic -- -- Characterization and Modeling of Long-Range Dependent Telecommunication Traffic Sponsor: Sprint Yong-Qing Lu David W. Petr Victor Frost Technical Report TISL-10230-4 Telecommunications and Information

More information

Estimation of the long Memory parameter using an Infinite Source Poisson model applied to transmission rate measurements

Estimation of the long Memory parameter using an Infinite Source Poisson model applied to transmission rate measurements of the long Memory parameter using an Infinite Source Poisson model applied to transmission rate measurements François Roueff Ecole Nat. Sup. des Télécommunications 46 rue Barrault, 75634 Paris cedex 13,

More information

Mice and Elephants Visualization of Internet

Mice and Elephants Visualization of Internet Mice and Elephants Visualization of Internet Traffic J. S. Marron, Felix Hernandez-Campos 2 and F. D. Smith 2 School of Operations Research and Industrial Engineering, Cornell University, Ithaca, NY, 4853,

More information

a + d < 1 with a = 0.03, d = 0.045, Nc = 40, and N = Nb a = 0.3, d = , Nc = 40, and N = 100.

a + d < 1 with a = 0.03, d = 0.045, Nc = 40, and N = Nb a = 0.3, d = , Nc = 40, and N = 100. On the Relevance of Time Scales in Performance Oriented Traffic Characterizations M. Montgomery and G. de Veciana Department of Electrical and Computer Engineering University of Texas at Austin Austin,

More information

On the Limitations of the Variance-Time Test for Inference of Long-Range Dependence

On the Limitations of the Variance-Time Test for Inference of Long-Range Dependence On the Limitations of the Variance-Time Test for Inference of Long-Range Dependence Marwan Krunz Department of Electrical & Computer Engineering University of Arizona Tucson, AZ 85721 krunz@ece.arizona.edu

More information

Source Traffic Modeling Using Pareto Traffic Generator

Source Traffic Modeling Using Pareto Traffic Generator Journal of Computer Networks, 207, Vol. 4, No., -9 Available online at http://pubs.sciepub.com/jcn/4//2 Science and Education Publishing DOI:0.269/jcn-4--2 Source Traffic odeling Using Pareto Traffic Generator

More information

CAC investigation for video and data

CAC investigation for video and data CAC investigation for video and data E.Aarstad a, S.Blaabjerg b, F.Cerdan c, S.Peeters d and K.Spaey d a Telenor Research & Development, P.O. Box 8, N-7 Kjeller, Norway,egil.aarstad@fou.telenor.no b Tele

More information

Resource Allocation for Video Streaming in Wireless Environment

Resource Allocation for Video Streaming in Wireless Environment Resource Allocation for Video Streaming in Wireless Environment Shahrokh Valaee and Jean-Charles Gregoire Abstract This paper focuses on the development of a new resource allocation scheme for video streaming

More information

ON THE CONVERGENCE OF FARIMA SEQUENCE TO FRACTIONAL GAUSSIAN NOISE. Joo-Mok Kim* 1. Introduction

ON THE CONVERGENCE OF FARIMA SEQUENCE TO FRACTIONAL GAUSSIAN NOISE. Joo-Mok Kim* 1. Introduction JOURNAL OF THE CHUNGCHEONG MATHEMATICAL SOCIETY Volume 26, No. 2, May 2013 ON THE CONVERGENCE OF FARIMA SEQUENCE TO FRACTIONAL GAUSSIAN NOISE Joo-Mok Kim* Abstract. We consider fractional Gussian noise

More information

I 2 (t) R (t) R 1 (t) = R 0 (t) B 1 (t) R 2 (t) B b (t) = N f. C? I 1 (t) R b (t) N b. Acknowledgements

I 2 (t) R (t) R 1 (t) = R 0 (t) B 1 (t) R 2 (t) B b (t) = N f. C? I 1 (t) R b (t) N b. Acknowledgements Proc. 34th Allerton Conf. on Comm., Cont., & Comp., Monticello, IL, Oct., 1996 1 Service Guarantees for Window Flow Control 1 R. L. Cruz C. M. Okino Department of Electrical & Computer Engineering University

More information

Teletrac modeling and estimation

Teletrac modeling and estimation Teletrac modeling and estimation File 2 José Roberto Amazonas jra@lcs.poli.usp.br Telecommunications and Control Engineering Dept. - PTC Escola Politécnica University of São Paulo - USP São Paulo 11/2008

More information

r bits/frame

r bits/frame Telecommunication Systems 0 (1999) 1 14 1 MODELING PACKET DELAY IN MULTIPLEXED VIDEO TRAFFIC Charles Thompson, Kavitha Chandra Λ,Sudha Mulpur ΛΛ and Jimmie Davis ΛΛΛ Center for Advanced Computation and

More information

In Proceedings of the 13th U.K. Workshop on Performance Engineering of Computer. and Telecommunication Systems (UKPEW'97), July 1997, Ilkley, U.K.

In Proceedings of the 13th U.K. Workshop on Performance Engineering of Computer. and Telecommunication Systems (UKPEW'97), July 1997, Ilkley, U.K. In Proceedings of the 13th U.K. Workshop on Performance Engineering of Computer and Telecommunication Systems (UKPEW'97), July 1997, Ilkley, U.K. Investigation of Cell Scale and Burst Scale Eects on the

More information

VARIANCE REDUCTION IN SIMULATIONS OF LOSS MODELS

VARIANCE REDUCTION IN SIMULATIONS OF LOSS MODELS VARIANCE REDUCTION IN SIMULATIONS OF LOSS MODELS by Rayadurgam Srikant 1 and Ward Whitt 2 October 20, 1995 Revision: September 26, 1996 1 Coordinated Science Laboratory, University of Illinois, 1308 W.

More information

On the Use of Self-Similar Processes in Network Simulation

On the Use of Self-Similar Processes in Network Simulation On the Use of Self-Similar Processes in Network Simulation JOSÉ C. LÓPEZ-ARDAO, CÁNDIDO LÓPEZ-GARCÍA, ANDRÉS SUÁREZ-GONZÁLEZ, MANUEL FERNÁNDEZ-VEIGA, and RAÚL RODRÍGUEZ-RUBIO University of Vigo, Spain

More information

Modeling Video Traffic Using M/G/ Input Processes: A Compromise Between Markovian and LRD Models. Marwan M. Krunz and Armand M.

Modeling Video Traffic Using M/G/ Input Processes: A Compromise Between Markovian and LRD Models. Marwan M. Krunz and Armand M. IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 16, NO. 5, JUNE 1998 733 Modeling Video Traffic Using M/G/ Input Processes: A Compromise Between Markovian and LRD Models Marwan M. Krunz and Armand

More information

A NOVEL APPROACH TO THE ESTIMATION OF THE HURST PARAMETER IN SELF-SIMILAR TRAFFIC

A NOVEL APPROACH TO THE ESTIMATION OF THE HURST PARAMETER IN SELF-SIMILAR TRAFFIC Proceedings of IEEE Conference on Local Computer Networks, Tampa, Florida, November 2002 A NOVEL APPROACH TO THE ESTIMATION OF THE HURST PARAMETER IN SELF-SIMILAR TRAFFIC Houssain Kettani and John A. Gubner

More information

A general theory of discrete ltering. for LES in complex geometry. By Oleg V. Vasilyev AND Thomas S. Lund

A general theory of discrete ltering. for LES in complex geometry. By Oleg V. Vasilyev AND Thomas S. Lund Center for Turbulence Research Annual Research Briefs 997 67 A general theory of discrete ltering for ES in complex geometry By Oleg V. Vasilyev AND Thomas S. und. Motivation and objectives In large eddy

More information

Trac of an ATM-SMX in the NCIH. Tung Ouyang and Arne A. Nilsson. Center for Advanced Computing and Communication

Trac of an ATM-SMX in the NCIH. Tung Ouyang and Arne A. Nilsson. Center for Advanced Computing and Communication Departure Process for Periodic Real-Time Trac of an ATM-SMX in the NCIH Tung Ouyang and Arne A. Nilsson Center for Advanced Computing and Communication Department of Electrical and Computer Engineering

More information

On the Performance of an Eective Bandwidths Formula. Costas Courcoubetis, George Fouskas. and Richard Weber y. Abstract

On the Performance of an Eective Bandwidths Formula. Costas Courcoubetis, George Fouskas. and Richard Weber y. Abstract On the Performance of an Eective Bandwidths Formula Costas Courcoubetis, George Fouskas and Richard Weber y Abstract At a buered switch in an ATM (asynchronous transfer mode) network it is important to

More information

On the departure process of a leaky bucket system with long-range dependent input traffic

On the departure process of a leaky bucket system with long-range dependent input traffic Queueing Systems 28 (998) 9 24 9 On the departure process of a leaky bucket system with long-range dependent input traffic Socrates Vamvakos and Venkat Anantharam EECS Department, University of California,

More information

BUFFER PROBLEMS IN TELECOMMUNICATIONS NETWORKS. Lester R. Lipsky and John E. Hatem. Department of Computer Science and Engineering

BUFFER PROBLEMS IN TELECOMMUNICATIONS NETWORKS. Lester R. Lipsky and John E. Hatem. Department of Computer Science and Engineering BUFFER PROBLEMS IN TELECOMMUNICATIONS NETWORKS. Lester R. Lipsky and John E. Hatem Department of Computer Science and Engineering University of Connecticut Storrs, CT 06269-3155 lester@brc.uconn.edu. and.

More information

queue KTH, Royal Institute of Technology, Department of Microelectronics and Information Technology

queue KTH, Royal Institute of Technology, Department of Microelectronics and Information Technology Analysis of the Packet oss Process in an MMPP+M/M/1/K queue György Dán, Viktória Fodor KTH, Royal Institute of Technology, Department of Microelectronics and Information Technology {gyuri,viktoria}@imit.kth.se

More information

Statistical Analysis of Delay Bound Violations at an Earliest Deadline First (EDF) Scheduler Vijay Sivaraman Department of Computer Science, 3820 Boel

Statistical Analysis of Delay Bound Violations at an Earliest Deadline First (EDF) Scheduler Vijay Sivaraman Department of Computer Science, 3820 Boel Statistical Analysis of Delay Bound Violations at an Earliest Deadline First (EDF) Scheduler Vijay Sivaraman Department of Computer Science, 3820 Boelter Hall, UCLA, Los Angeles, CA 90095, U.S.A. (Email:

More information

Sensitivity of ABR Congestion Control Algorithms to Hurst Parameter Estimates

Sensitivity of ABR Congestion Control Algorithms to Hurst Parameter Estimates Sensitivity of ABR Congestion Control Algorithms to Hurst Parameter Estimates Sven A. M. Östring 1, Harsha Sirisena 1, and Irene Hudson 2 1 Department of Electrical & Electronic Engineering 2 Department

More information

e-05

e-05 Empirical Eective Bandwidths M. Falkner, M. Devetsikiotis, I. Lambadaris Department of Systems and Computer Engineering Carleton University 25 Colonel By Drive Ottawa, Ontario KS 5B6, Canada S. Tartarelli,

More information

THE key objective of this work is to bridge the gap

THE key objective of this work is to bridge the gap 1052 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 15, NO. 6, AUGUST 1997 The Effect of Multiple Time Scales and Subexponentiality in MPEG Video Streams on Queueing Behavior Predrag R. Jelenković,

More information

Part I Stochastic variables and Markov chains

Part I Stochastic variables and Markov chains Part I Stochastic variables and Markov chains Random variables describe the behaviour of a phenomenon independent of any specific sample space Distribution function (cdf, cumulative distribution function)

More information

1 Introduction Future high speed digital networks aim to serve integrated trac, such as voice, video, fax, and so forth. To control interaction among

1 Introduction Future high speed digital networks aim to serve integrated trac, such as voice, video, fax, and so forth. To control interaction among On Deterministic Trac Regulation and Service Guarantees: A Systematic Approach by Filtering Cheng-Shang Chang Dept. of Electrical Engineering National Tsing Hua University Hsinchu 30043 Taiwan, R.O.C.

More information

Performance Comparison of Two Implementations of the Leaky. LMS Adaptive Filter. Scott C. Douglas. University of Utah. Salt Lake City, Utah 84112

Performance Comparison of Two Implementations of the Leaky. LMS Adaptive Filter. Scott C. Douglas. University of Utah. Salt Lake City, Utah 84112 Performance Comparison of Two Implementations of the Leaky LMS Adaptive Filter Scott C. Douglas Department of Electrical Engineering University of Utah Salt Lake City, Utah 8411 Abstract{ The leaky LMS

More information

Some Background Information on Long-Range Dependence and Self-Similarity On the Variability of Internet Traffic Outline Introduction and Motivation Ch

Some Background Information on Long-Range Dependence and Self-Similarity On the Variability of Internet Traffic Outline Introduction and Motivation Ch On the Variability of Internet Traffic Georgios Y Lazarou Information and Telecommunication Technology Center Department of Electrical Engineering and Computer Science The University of Kansas, Lawrence

More information

STATISTICAL MODELING OF ASYNCHRONOUS IMPULSIVE NOISE IN POWERLINE COMMUNICATION NETWORKS

STATISTICAL MODELING OF ASYNCHRONOUS IMPULSIVE NOISE IN POWERLINE COMMUNICATION NETWORKS STATISTICAL MODELING OF ASYNCHRONOUS IMPULSIVE NOISE IN POWERLINE COMMUNICATION NETWORKS Marcel Nassar, Kapil Gulati, Yousof Mortazavi, and Brian L. Evans Department of Electrical and Computer Engineering

More information

On the Impact of Traffic Characteristics on Radio Resource Fluctuation in Multi-Service Cellular CDMA Networks

On the Impact of Traffic Characteristics on Radio Resource Fluctuation in Multi-Service Cellular CDMA Networks On the Impact of Traffic Characteristics on Radio Resource Fluctuation in Multi-Service Cellular CDMA Networks Keivan Navaie Sys. and Comp. Department Carleton University, Ottawa, Canada keivan@sce.carleton.ca

More information

2. SPECTRAL ANALYSIS APPLIED TO STOCHASTIC PROCESSES

2. SPECTRAL ANALYSIS APPLIED TO STOCHASTIC PROCESSES 2. SPECTRAL ANALYSIS APPLIED TO STOCHASTIC PROCESSES 2.0 THEOREM OF WIENER- KHINTCHINE An important technique in the study of deterministic signals consists in using harmonic functions to gain the spectral

More information

MODELLING OF SELF-SIMILAR TELETRAFFIC FOR SIMULATION. A thesis submitted in partial fulfilment. of the requirements for the degree of

MODELLING OF SELF-SIMILAR TELETRAFFIC FOR SIMULATION. A thesis submitted in partial fulfilment. of the requirements for the degree of MODELLING OF SELF-SIMILAR TELETRAFFIC FOR SIMULATION A thesis submitted in partial fulfilment of the requirements for the degree of Doctor of Philosophy In Computer Science in the University of Canterbury

More information

Survey of Source Modeling Techniques for ATM Networks

Survey of Source Modeling Techniques for ATM Networks Survey of Source Modeling Techniques for ATM Networks Sponsor: Sprint Yong-Qing Lu David W. Petr Victor S. Frost Technical Report TISL-10230-1 Telecommunications and Information Sciences Laboratory Department

More information

Long range dependent Markov chains with applications

Long range dependent Markov chains with applications Long range dependent Markov chains with applications Barlas Oğuz, Venkat Anantharam Department of Electrical Engineering and Computer Sciences University of California, Berkeley Email: {barlas, ananth}@eecs.berkeley.edu

More information

PRACTICAL ASPECTS OF SIMULATING SYSTEMS HAVING ARRIVAL PROCESSES WITH LONG-RANGE DEPENDENCE. Robert Geist James Westall

PRACTICAL ASPECTS OF SIMULATING SYSTEMS HAVING ARRIVAL PROCESSES WITH LONG-RANGE DEPENDENCE. Robert Geist James Westall Proceedings of the 2 Winter Simulation Conference J. A. Joines, R. R. Barton, K. Kang, and P. A. Fishwick, eds. PRACTICAL ASPECTS OF SIMULATING SYSTEMS HAVING ARRIVAL PROCESSES WITH LONG-RANGE DEPENDENCE

More information

Network Traffic Modeling using a Multifractal Wavelet Model

Network Traffic Modeling using a Multifractal Wavelet Model 5-th International Symposium on Digital Signal Processing for Communication Systems, DSPCS 99, Perth, 1999 Network Traffic Modeling using a Multifractal Wavelet Model Matthew S. Crouse, Rudolf H. Riedi,

More information

ELEG 833. Nonlinear Signal Processing

ELEG 833. Nonlinear Signal Processing Nonlinear Signal Processing ELEG 833 Gonzalo R. Arce Department of Electrical and Computer Engineering University of Delaware arce@ee.udel.edu February 15, 2005 1 INTRODUCTION 1 Introduction Signal processing

More information

CONVERGENCE TO FRACTIONAL BROWNIAN MOTION AND LOSS PROBABILITY. Jin-Chun Kim and Hee-Choon Lee

CONVERGENCE TO FRACTIONAL BROWNIAN MOTION AND LOSS PROBABILITY. Jin-Chun Kim and Hee-Choon Lee Kangweon-Kyungki Math. Jour. (2003), No., pp. 35 43 CONVERGENCE TO FRACTIONAL BROWNIAN MOTION AND LOSS PROBABILITY Jin-Chun Kim and Hee-Choon Lee Abstract. We study the weak convergence to Fractional Brownian

More information

Accurate and Fast Replication on the Generation of Fractal Network Traffic Using Alternative Probability Models

Accurate and Fast Replication on the Generation of Fractal Network Traffic Using Alternative Probability Models Accurate and Fast Replication on the Generation of Fractal Network Traffic Using Alternative Probability Models Stenio Fernandes, Carlos Kamienski & Djamel Sadok Computer Science Center, Federal University

More information

The exponential distribution and the Poisson process

The exponential distribution and the Poisson process The exponential distribution and the Poisson process 1-1 Exponential Distribution: Basic Facts PDF f(t) = { λe λt, t 0 0, t < 0 CDF Pr{T t) = 0 t λe λu du = 1 e λt (t 0) Mean E[T] = 1 λ Variance Var[T]

More information

C. Huang, M. Devetsikiotis, I. Lambadaris, and A. R. Kaye. Carleton University ABSTRACT

C. Huang, M. Devetsikiotis, I. Lambadaris, and A. R. Kaye. Carleton University ABSTRACT FAST SIMULATION OF QUEUES WITH LONG-RANGE DEPENDENT TRAFFIC C. Huang, M. Devetsikiotis, I. Lambadaris, and A. R. Kaye Department of Systems & Computer Engineering Carleton University Ottawa, Ontario K1S

More information

on a Stochastic Current Waveform Urbana, Illinois Dallas, Texas Abstract

on a Stochastic Current Waveform Urbana, Illinois Dallas, Texas Abstract Electromigration Median Time-to-Failure based on a Stochastic Current Waveform by Farid Najm y, Ibrahim Hajj y, and Ping Yang z y Coordinated Science Laboratory z VLSI Design Laboratory University of Illinois

More information

Environment (E) IBP IBP IBP 2 N 2 N. server. System (S) Adapter (A) ACV

Environment (E) IBP IBP IBP 2 N 2 N. server. System (S) Adapter (A) ACV The Adaptive Cross Validation Method - applied to polling schemes Anders Svensson and Johan M Karlsson Department of Communication Systems Lund Institute of Technology P. O. Box 118, 22100 Lund, Sweden

More information

6.435, System Identification

6.435, System Identification System Identification 6.435 SET 3 Nonparametric Identification Munther A. Dahleh 1 Nonparametric Methods for System ID Time domain methods Impulse response Step response Correlation analysis / time Frequency

More information

Optimal Rejuvenation for. Tolerating Soft Failures. Andras Pfening, Sachin Garg, Antonio Puliato, Miklos Telek, Kishor S. Trivedi.

Optimal Rejuvenation for. Tolerating Soft Failures. Andras Pfening, Sachin Garg, Antonio Puliato, Miklos Telek, Kishor S. Trivedi. Optimal Rejuvenation for Tolerating Soft Failures Andras Pfening, Sachin Garg, Antonio Puliato, Miklos Telek, Kishor S. Trivedi Abstract In the paper we address the problem of determining the optimal time

More information

Lowen and Teich, JASA 2 RECEIVED:

Lowen and Teich, JASA 2 RECEIVED: 1 The periodogram and Allan variance reveal fractal exponents greater than unity in auditory-nerve spike trains Steven B. Lowen Department of Electrical Engineering 500 West 120th Street Columbia University,

More information

v n,t n

v n,t n THE DYNAMICAL STRUCTURE FACTOR AND CRITICAL BEHAVIOR OF A TRAFFIC FLOW MODEL 61 L. ROTERS, S. L UBECK, and K. D. USADEL Theoretische Physik, Gerhard-Mercator-Universitat, 4748 Duisburg, Deutschland, E-mail:

More information

Evaluation of Effective Bandwidth Schemes for Self-Similar Traffic

Evaluation of Effective Bandwidth Schemes for Self-Similar Traffic Proceedings of the 3th ITC Specialist Seminar on IP Measurement, Modeling and Management, Monterey, CA, September 2000, pp. 2--2-0 Evaluation of Effective Bandwidth Schemes for Self-Similar Traffic Stefan

More information

On the relevance of long-tailed durations for the statistical multiplexing of large aggregations.

On the relevance of long-tailed durations for the statistical multiplexing of large aggregations. On the relevance of long-tailed durations for the statistical multiplexing of large aggregations. N. G. Duffield AT&T Laboratories Room 2C-323, 600 Mountain Avenue, Murray Hill, NJ 07974, USA duffield@research.att.com

More information

Sensitivity Analysis for Discrete-Time Randomized Service Priority Queues

Sensitivity Analysis for Discrete-Time Randomized Service Priority Queues Sensitivity Analysis for Discrete-Time Randomized Service Priority Queues George Kesidis 1, Takis Konstantopoulos 2, Michael Zazanis 3 1. Elec. & Comp. Eng. Dept, University of Waterloo, Waterloo, ON,

More information

ANALY SIS, SYNTHESIS, AND ESTIMATION OE'FRACTAL-RATE STOCHASTIC POINT PROCESSES

ANALY SIS, SYNTHESIS, AND ESTIMATION OE'FRACTAL-RATE STOCHASTIC POINT PROCESSES Fractals, Vol. 5, No.4 (1997) 565-595 World ScientificPublishing Company ANALY SIS, SYNTHESIS, AND ESTIMATION OE'FRACTAL-RATE STOCHASTIC POINT PROCESSES STEFAN THURNER, STEVEN B. LOWEN, MARKUS C. FEURSTEIN

More information

Effect of the Traffic Bursts in the Network Queue

Effect of the Traffic Bursts in the Network Queue RICE UNIVERSITY Effect of the Traffic Bursts in the Network Queue by Alireza KeshavarzHaddad A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree Master of Science Approved, Thesis

More information

ROYAL INSTITUTE OF TECHNOLOGY KUNGL TEKNISKA HÖGSKOLAN. Department of Signals, Sensors & Systems

ROYAL INSTITUTE OF TECHNOLOGY KUNGL TEKNISKA HÖGSKOLAN. Department of Signals, Sensors & Systems The Evil of Supereciency P. Stoica B. Ottersten To appear as a Fast Communication in Signal Processing IR-S3-SB-9633 ROYAL INSTITUTE OF TECHNOLOGY Department of Signals, Sensors & Systems Signal Processing

More information

Discrete Simulation of Power Law Noise

Discrete Simulation of Power Law Noise Discrete Simulation of Power Law Noise Neil Ashby 1,2 1 University of Colorado, Boulder, CO 80309-0390 USA 2 National Institute of Standards and Technology, Boulder, CO 80305 USA ashby@boulder.nist.gov

More information

Stochastic Network Calculus

Stochastic Network Calculus Stochastic Network Calculus Assessing the Performance of the Future Internet Markus Fidler joint work with Amr Rizk Institute of Communications Technology Leibniz Universität Hannover April 22, 2010 c

More information

Learning with Ensembles: How. over-tting can be useful. Anders Krogh Copenhagen, Denmark. Abstract

Learning with Ensembles: How. over-tting can be useful. Anders Krogh Copenhagen, Denmark. Abstract Published in: Advances in Neural Information Processing Systems 8, D S Touretzky, M C Mozer, and M E Hasselmo (eds.), MIT Press, Cambridge, MA, pages 190-196, 1996. Learning with Ensembles: How over-tting

More information

1 Introduction This work follows a paper by P. Shields [1] concerned with a problem of a relation between the entropy rate of a nite-valued stationary

1 Introduction This work follows a paper by P. Shields [1] concerned with a problem of a relation between the entropy rate of a nite-valued stationary Prexes and the Entropy Rate for Long-Range Sources Ioannis Kontoyiannis Information Systems Laboratory, Electrical Engineering, Stanford University. Yurii M. Suhov Statistical Laboratory, Pure Math. &

More information

Network Traffic Modeling using a Multifractal Wavelet Model

Network Traffic Modeling using a Multifractal Wavelet Model Proceedings European Congress of Mathematics, Barcelona 2 Network Traffic Modeling using a Multifractal Wavelet Model Rudolf H. Riedi, Vinay J. Ribeiro, Matthew S. Crouse, and Richard G. Baraniuk Abstract.

More information

FAKULTÄT FÜR INFORMATIK

FAKULTÄT FÜR INFORMATIK b b b b b b b b b b b b b b b b b b b b FAKULTÄT FÜR INFORMATIK der Technischen Universität München Lehrstuhl VIII Rechnerstruktur/-architektur Prof. Dr. E. Jessen Modeling of Packet Arrivals Using Markov

More information

Loss Probability Calculations and Asymptotic Analysis for Finite Buffer Multiplexers

Loss Probability Calculations and Asymptotic Analysis for Finite Buffer Multiplexers IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 9, NO. 6, DECEMBER 2001 755 Loss Probability Calculations Asymptotic Analysis for Finite Buffer Multiplexers Han S. Kim Ness B. Shroff, Senior Member, IEEE Abstract

More information

Performance Evaluation and Service Rate Provisioning for a Queue with Fractional Brownian Input

Performance Evaluation and Service Rate Provisioning for a Queue with Fractional Brownian Input Performance Evaluation and Service Rate Provisioning for a Queue with Fractional Brownian Input Jiongze Chen 1, Ronald G. Addie 2, Moshe Zukerman 1 Abstract The Fractional Brownian motion (fbm) traffic

More information

A Generator of Pseudo-Random Self-Similar Sequences Based on SRA

A Generator of Pseudo-Random Self-Similar Sequences Based on SRA A Generator of Pseudo-Random Self-Similar Sequences Based on SRA H.-D. J. Jeong,D.McNickle and K. Pawlikowski Department of Computer Science and Management University of Canterbury Christchurch, New Zealand

More information

THE need for telecommunication networks capable of

THE need for telecommunication networks capable of DUAL DEGREE SEMINAR REPORT, NOVEMBER 2000 1 Modeling of Network Traffic Student: Aditya Dua (97D07003) Guide: Prof. U.B. Desai Abstract Traffic models are at the heart of any performance evaluation of

More information

Elec4621 Advanced Digital Signal Processing Chapter 11: Time-Frequency Analysis

Elec4621 Advanced Digital Signal Processing Chapter 11: Time-Frequency Analysis Elec461 Advanced Digital Signal Processing Chapter 11: Time-Frequency Analysis Dr. D. S. Taubman May 3, 011 In this last chapter of your notes, we are interested in the problem of nding the instantaneous

More information

A Virtual Queue Approach to Loss Estimation

A Virtual Queue Approach to Loss Estimation A Virtual Queue Approach to Loss Estimation Guoqiang Hu, Yuming Jiang, Anne Nevin Centre for Quantifiable Quality of Service in Communication Systems Norwegian University of Science and Technology, Norway

More information

Gaussian distributions and processes have long been accepted as useful tools for stochastic

Gaussian distributions and processes have long been accepted as useful tools for stochastic Chapter 3 Alpha-Stable Random Variables and Processes Gaussian distributions and processes have long been accepted as useful tools for stochastic modeling. In this section, we introduce a statistical model

More information

G. Larry Bretthorst. Washington University, Department of Chemistry. and. C. Ray Smith

G. Larry Bretthorst. Washington University, Department of Chemistry. and. C. Ray Smith in Infrared Systems and Components III, pp 93.104, Robert L. Caswell ed., SPIE Vol. 1050, 1989 Bayesian Analysis of Signals from Closely-Spaced Objects G. Larry Bretthorst Washington University, Department

More information

Accelerated Simulation of Power-Law Traffic in Packet Networks

Accelerated Simulation of Power-Law Traffic in Packet Networks Accelerated Simulation of Power-Law Traffic in Packet Networks By Ho I Ma SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Supervised by Dr. John A. Schormans Department of Electronic Engineering Queen

More information

LIMITS FOR QUEUES AS THE WAITING ROOM GROWS. Bell Communications Research AT&T Bell Laboratories Red Bank, NJ Murray Hill, NJ 07974

LIMITS FOR QUEUES AS THE WAITING ROOM GROWS. Bell Communications Research AT&T Bell Laboratories Red Bank, NJ Murray Hill, NJ 07974 LIMITS FOR QUEUES AS THE WAITING ROOM GROWS by Daniel P. Heyman Ward Whitt Bell Communications Research AT&T Bell Laboratories Red Bank, NJ 07701 Murray Hill, NJ 07974 May 11, 1988 ABSTRACT We study the

More information

MODELS FOR COMPUTER NETWORK TRAFFIC

MODELS FOR COMPUTER NETWORK TRAFFIC MODELS FOR COMPUTER NETWORK TRAFFIC Murad S. Taqqu Boston University Joint work with Walter Willinger, Joshua Levy and Vladas Pipiras,... Web Site http://math.bu.edu/people/murad OUTLINE Background: 1)

More information

Wavelet Analysis of Long Range Dependent Trac. (1) CNRS URA Laboratoire de Physique - Ecole Normale Superieure de Lyon -

Wavelet Analysis of Long Range Dependent Trac. (1) CNRS URA Laboratoire de Physique - Ecole Normale Superieure de Lyon - Wavelet Analysis of Long Range Dependent Trac Patrice Abry 1, Darryl Veitch 2 (1) CNRS URA 1325 - Laboratoire de Physique - Ecole Normale Superieure de Lyon - 46, allee d'italie 69 364 LYON Cedex 07- France

More information

If we want to analyze experimental or simulated data we might encounter the following tasks:

If we want to analyze experimental or simulated data we might encounter the following tasks: Chapter 1 Introduction If we want to analyze experimental or simulated data we might encounter the following tasks: Characterization of the source of the signal and diagnosis Studying dependencies Prediction

More information

Design of IP networks with Quality of Service

Design of IP networks with Quality of Service Course of Multimedia Internet (Sub-course Reti Internet Multimediali ), AA 2010-2011 Prof. Pag. 1 Design of IP networks with Quality of Service 1 Course of Multimedia Internet (Sub-course Reti Internet

More information

THE HEAVY-TRAFFIC BOTTLENECK PHENOMENON IN OPEN QUEUEING NETWORKS. S. Suresh and W. Whitt AT&T Bell Laboratories Murray Hill, New Jersey 07974

THE HEAVY-TRAFFIC BOTTLENECK PHENOMENON IN OPEN QUEUEING NETWORKS. S. Suresh and W. Whitt AT&T Bell Laboratories Murray Hill, New Jersey 07974 THE HEAVY-TRAFFIC BOTTLENECK PHENOMENON IN OPEN QUEUEING NETWORKS by S. Suresh and W. Whitt AT&T Bell Laboratories Murray Hill, New Jersey 07974 ABSTRACT This note describes a simulation experiment involving

More information

A POPULATION-MIX DRIVEN APPROXIMATION FOR QUEUEING NETWORKS WITH FINITE CAPACITY REGIONS

A POPULATION-MIX DRIVEN APPROXIMATION FOR QUEUEING NETWORKS WITH FINITE CAPACITY REGIONS A POPULATION-MIX DRIVEN APPROXIMATION FOR QUEUEING NETWORKS WITH FINITE CAPACITY REGIONS J. Anselmi 1, G. Casale 2, P. Cremonesi 1 1 Politecnico di Milano, Via Ponzio 34/5, I-20133 Milan, Italy 2 Neptuny

More information

A Queueing System with Queue Length Dependent Service Times, with Applications to Cell Discarding in ATM Networks

A Queueing System with Queue Length Dependent Service Times, with Applications to Cell Discarding in ATM Networks A Queueing System with Queue Length Dependent Service Times, with Applications to Cell Discarding in ATM Networks by Doo Il Choi, Charles Knessl and Charles Tier University of Illinois at Chicago 85 South

More information

An Introduction to Stochastic Modeling

An Introduction to Stochastic Modeling F An Introduction to Stochastic Modeling Fourth Edition Mark A. Pinsky Department of Mathematics Northwestern University Evanston, Illinois Samuel Karlin Department of Mathematics Stanford University Stanford,

More information

The Entropy of Cell Streams as a. Trac Descriptor in ATM Networks

The Entropy of Cell Streams as a. Trac Descriptor in ATM Networks 1 The Entropy of Cell Streams as a Trac Descriptor in ATM Networks N. T. Plotkin SRI International 333 Ravenswood Avenue Menlo Park, CA 94025, USA ninatp@erg.sri.com and C. Roche Laboratoire MASI Universite

More information

Manual for SOA Exam MLC.

Manual for SOA Exam MLC. Chapter 10. Poisson processes. Section 10.5. Nonhomogenous Poisson processes Extract from: Arcones Fall 2009 Edition, available at http://www.actexmadriver.com/ 1/14 Nonhomogenous Poisson processes Definition

More information

Solution: The process is a compound Poisson Process with E[N (t)] = λt/p by Wald's equation.

Solution: The process is a compound Poisson Process with E[N (t)] = λt/p by Wald's equation. Solutions Stochastic Processes and Simulation II, May 18, 217 Problem 1: Poisson Processes Let {N(t), t } be a homogeneous Poisson Process on (, ) with rate λ. Let {S i, i = 1, 2, } be the points of the

More information

P 1j. P jm. P ij. p m

P 1j. P jm. P ij. p m Analytic Model of Performance in Telecommunication Systems, Based on On-O Trac Sources with Self-Similar Behavior Lester Lipsky Department of Computer Science and Engineering University of Connecticut,

More information

In Proceedings of the 1997 Winter Simulation Conference, S. Andradottir, K. J. Healy, D. H. Withers, and B. L. Nelson, eds.

In Proceedings of the 1997 Winter Simulation Conference, S. Andradottir, K. J. Healy, D. H. Withers, and B. L. Nelson, eds. In Proceedings of the 1997 Winter Simulation Conference, S. Andradottir, K. J. Healy, D. H. Withers, and B. L. Nelson, eds. LONG-LASTING TRANSIENT CONDITIONS IN SIMULATIONS WITH HEAVY-TAILED WORKLOADS

More information

Modelling the risk process

Modelling the risk process Modelling the risk process Krzysztof Burnecki Hugo Steinhaus Center Wroc law University of Technology www.im.pwr.wroc.pl/ hugo Modelling the risk process 1 Risk process If (Ω, F, P) is a probability space

More information

Baltzer Journals Received 7 January 1997; revised 16 April 1997 An Interpolation Approximation for the GI/G/1 Queue Based on Multipoint Pade Approxima

Baltzer Journals Received 7 January 1997; revised 16 April 1997 An Interpolation Approximation for the GI/G/1 Queue Based on Multipoint Pade Approxima Baltzer Journals Received 7 January 1997; revised 16 April 1997 An Interpolation Approximation for the GI/G/1 Queue Based on Multipoint Pade Approximation Muckai K Girish 1 and Jian-Qiang Hu 1 Telesis

More information

Queue Response to Input Correlation Functions: Continuous Spectral Analysis. University of Texas at Austin. Austin, Texas

Queue Response to Input Correlation Functions: Continuous Spectral Analysis. University of Texas at Austin. Austin, Texas Queue Response to Input Correlation Functions: Continuous Spectral Analysis San-i Li Chia-Lin Hwang Department of Electrical and Computer Engineering University of Texas at Austin Austin, Texas 78712 August

More information

Queue Analysis for Wireless Packet Data Traffic

Queue Analysis for Wireless Packet Data Traffic Queue Analysis for Wireless Packet Data Traffic Shahram Teymori and Weihua Zhuang Centre for Wireless Communications (CWC), Department of Electrical and Computer Engineering, University of Waterloo, Waterloo,

More information