Question: Plot the results for two other cases: r = 0.2 and r = 0.75. Show the plots on top of the results given for r = 0.5.
To the expert: Dont paste irrelevant answer. NO CODES. JUST THE PLOT IS NEEDED !!!
SLIDE ON THE VOTING ALGORITHM ARE BELOW:
(20 pts) Consider the study of replica voting algorithm you had done as part of the first warm-up project. For an extended analysis, Figure 4 shows the key performance results of the voting algorithms. The performance parameters are the time to deliver a data to the end-user (TTC). the number of distinct data proposals before effecting a data delivery, and the control message overhead expended to deliver a data. One of the influential parameters is the fault severity of a mis-behaving voter: denoted as a parameter r, where r is the probability that a faulty device does in fact behave incorrectly. For e.g., r=0.1 means that a faulty device X behaves incorrectly only for 10% of the times X participates in the algorithm operations - which means that X behaves correctly for 90% of the times. The results are shown for a case of r=0.5. Empirically plot the results for two other cases: r=0.2 and r=0.75. Show the plots on top of the results given for r=0.5. N=10;datasize=30kbytesfm:#offaultydevices # of YES votes needed =6; (Tc)=50mec;(Tc)=50msec; control message: 50 bytes Tc : voter computation time Figure 4: Experimental performance results of replica voting algorithm Prototype implementation of replica system at CUNY [for simplicity, consider only the cases: fafm in the GC-72020 course project] Linux workstations (N=10) in CCNY lab cope of prototyping efforts (code written in C++ ) ** Functionality checks and protocol validation by 'stress testing' Subject the protocol to rigorous failure scenarios to ascertain its 'correct data delivery' capability and identify any design errors thereir ** Performability checks on protocol Determine the maximum achievable performance by testing the protocol under 'normal case' behaviors in the environment (i.e., message loss is very low, (Tc) is small, and fmN/2) Experiments with random injection of faulty voter behaviors, control of voter asynchrony by random spread out of the 'data ready' time of servers IP multicast transport, for the purpose of performability studies Network engineering ( contd.) Selective solicitation of votes Poll only fm voters at a time ( B specifies the voter list as a bit-map) Sample scenario for M2PC: N=5,fm=1 Network engineering ( . . contd.) Solicit votes when required \# of voters, i.e., fm+1, are likely to have data ready ("smart wait") avoids premature voting, i.e., B waits long enough to ask for votes so that most of the voters have their data reac 1 \# of data proposal (with collision suppression) 2\#s of vote messages 1 \# of data solicitation message 1# of vote solicitation message How can B determine the time of vote solicitation Tvsol ?? Assume knowledge of the probability distribution of voter computation time: PTc(x) Determine Tvsol such that the condition: l=1N1[(N1)clql(1q)N1l]l=fm holds true, where q=tTvsolPTc(x)dx
