Performance Analysis for Large IaaS Clouds - Large Scale and Big Data: Processing and Management - page 563

markov provision_decision
0_0 0_H lambda
0_H 0_0 delta_h*P_h
0_W 0_0 delta_w*P_w
0_C 0_0 delta_c
loop i,0,N-2
$(i)_H $(i+1)_H lambda
$(i)_W $(i+1)_W lambda
$(i)_C $(i+1)_C lambda
end
loop i,1,N-1
$(i)_H $(i-1)_H delta_h*P_h
$(i)_W $(i-1)_H delta_w*P_w
$(i)_C $(i-1)_H delta_c
end
loop i,0,N-1
$(i)_H $(i)_W delta_h*(1-P_h)
$(i)_W $(i)_C delta_w*(1-P_w)
end
end
end
* Blocking probability
bind P_block (prob(provision_decision,$(N-1)_H)+prob
(provision_decision,$(N-1)_W)+prob(provision_decision,$(N-1)_C))
* Dropping probability
bind P_drop (delta_c*(1-P_c)/(lambda)) * (sum(p, 0, (N-1),
prob(provision_decision, $(p)_C)))
* Overall job rejection probability
bind P_reject (P_block + P_drop)
* Mean number of jobs in RPDE
bind mean_num_RPDE sum(p, 0, (N-1), (p+1)*(prob(provision_
decision, $(p)_H)+prob(provision_decision, $(p)_W)+prob
(provision_decision, $(p)_C)))
expr P_block
expr P_drop
expr P_reject
expr mean_num_RPDE
end
The above SHARPE code shows how states and transitions can be specified in
a SHARPE input file. For example, the first line within the Markov model provi-
sion_decision shows that the model moves from state 0_0 to state 0_ H with rate
λ. Also, notice the use of loop statements to describe the structure of the CTMC.
Input parameters P h , P w , and P c are originally computed from VM provisioning sub-
models. However, we use some dummy values in the above code to demonstrate the
working of the SHARPE input file to the readers. Detailed SHARPE input file for
the interacting submodels can be found in [7]. Detailed tutorials on SHARPE can
be found in [1].