R. Jain. The Art of Computer SystemsPerformance Analysis. John Wiley & Sons, Inc., 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... speedups for TCP Vegas and serve as a starting point for our more detailed evaluation of the performance implications of the individual algorithms incorporated in TCP Vegas (see Section II A) For the evaluation of the algorithms, we take the approach of a 2 k factorial design with replications [22]. This methodology allows us to determine the effect of k factors, each of them having two levels. In the case of TCP Vegas, these factors are the different algorithms, as presented in Section II A. The factor levels are on and off , indicating whether the TCP Vegas algorithm is used ( on ) or ....

....of complexity As summarized in Section V, TCP Vegas employs nine additional algorithms over TCP Reno. A complete 2 k r factorial design requires that each possible combination of the k = 9 algorithms is chosen and that the experiment described in Section IV is run r times for a specific setup [22]. This methodology would allow us to quantify the effect of each individual algorithm and the effects of all possible interactions of the algorithms, that is, for k = 9 algorithms, the experiment would result in 2 9 Gamma 1 possible effects, most of them being probably rather small 7 . To ....

[Article contains additional citation context not shown here]

R. Jain. The Art of Computer SystemsPerformance Analysis. John Wiley & Sons, Inc., 1991.

....evaluate the performance of the s 2PL and g 2PL protocols under different high speed networking latencies, simulation models of both protocols were developed using the C programming language. The simulation is a discrete event simulation using the unit time approach to advance the simulation clock [28]. We consider a dataserver database system, with a single server and multiple clients connected by a high speed network. As described earlier, the transmission delays in a high speed network can be assumed to be negligible, and the network latency consists of the signal propagation and switching ....

R. Jain, The Art of ComputerSystemsPerformanceAnalysis. 1991.

....counted in a predefined grid is inefficient because there may be many histogram bins with low probabilities of occurrence. If similar measurement results appear in groups of measured points, the analyst may try to construct workload clusters using clustering techniques as described for example in [14]. Each workload cluster can be characterized by a K dimensional box, i.e. a combination of parameter intervals. The corresponding VU list for the workload clusters depicted in Figure 8 is (also in this example, the service demand for the CPU is assumed to be constant d (i) cpu = 5, 8i) d ....

R. Jain. The Art of Computer SystemsPerformanceAnalysis. John Wiley & Sons, Inc., New York et al., 1991.

.... ; XN are independent and identically distributed (when receiver interest is stationary) A consequence of the Central Limit Theorem is that N = Z 1 Gammaff=2 oe 2 (1) where oe is the standard deviation of the weight distribution, and Z p is the p quantile of the unit normal variate [14]. Using this result, we can compute the number of samples required to establish the desired confidence level for a given ACM Multimedia November 1997 Seattle, WA 8 a = 0.01 a = 0.03 10 100 1000 10000 100000 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Number of Samples 0.825 e 82.5 ....

JAIN, R. The Art of ComputerSystemsPerformanceAnalysis. John Wiley & Sons Inc., 1991.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC