Vous êtes sur la page 1sur 5

Note: You are welcome to bring this formula sheet to the exam, but it is your

responsibility to check the accuracy of the formulas given here.

Chapter 24: Review of Probability


Pr(E1 \ E2 )
Conditional probability: Pr(E2 j E1 ) = , if Pr(E1 ) 6= 0
Pr(E1 )
t 2
f (t) = e ,t 0 with E(T ) = 1= and Var(T ) = 1=
Exponential density:
t
F (t) = Pr(T t) = 1 e
( t)n e t
Poisson density: Pr(N (t) = n) = , n = 0; 1; : : : with E[N (t)] = t
n!
1
Uniform density: f (x) = , for a x b
b a
P R1
Expected value: E(X) = k xk Pr(X = xk ), or E(X) = 0
xf (x) dx:

Variance: Var(X) = E(X 2 ) [E(X)]2


P
E(X) = y E(X j Y = y) Pr(Y = y)
Expectations by conditioning:
R1
E(X) = 0
E(X j Y = y)g(y) dy
P
Pr(A) = y Pr(A j Y = y) Pr(Y = y)
Probabilities by conditioning:
R1
Pr(A) = 0
Pr(A j Y = y)g(y) dy

Chapter 16: Markov Chains (Skip)


n-step transition matrices: P(n) = Pn . Unconditional probabilities are (n)
= (n 1)
P,
for n = 1; 2; : : : where = ( 0 ; 1 ; : : : ; M ).
P PM
Steady-state equations: j = M i=0 i pij and j=0 j = 1; or in matrix-vector notation,
0
= P and e = 1, where e = (1; 1; : : : ; 1).

Cost calculation for complex functions: Calculate using conditional probabilites from

X
M
E(Cost) = E(Cost j X = j) Pr(X = j).
j=0

P
Mean rst passage times: ij =1+ k6=j pik kj . (Note that ii = 1= i .)

1
Absorbing chains:
" #
I 0
P = , N = (I Q) 1 ,
R Q
NR = Pr(End up in an absorbing state j start elsewhere)

Cramers rule: " # " #


a b 1 1 d b
M= , M =
c d det(M ) c a

CTMC: Balance equation is from the principle of Rate Out = Rate In.

Chapter 17: Queueing Theory


Birth-Death Processes:
Rate Out = Rate In
n=0: P0 0 = P1 1
n=1: ( 1 + 1 )P1 = 0 P0 + 2 P2
n=2: ( 2 + 2 )P2 = 1 P1 + 3 P3
Any n 1 : ( n + n )Pn = n 1 Pn 1 + n+1 Pn+1 :

Solution to the above is obtained from

0 1 n 1
X
1
Pn = P 0 , n = 1; 2; : : : and Pn = 1:
1 2 n n=0

Operating Characteristics of Queues

In all models, =(s ) is the utilization factor, i.e., the expected fraction time the
P
individual server(s) are busy, where = 1 n=0 n Pn .

We use the following shorthand notation in this formula sheet.

2
Shorthand Notation Explanation

M=M=1 Poisson input, exponential service time model with 1 server

M=M=s Poisson input, exponential service time model with s servers

M=M=1=K The M=M=1 model with nite queue

M=M=s=K The M=M=s model with nite queue

M=M=1= =N The M=M=1 model with nite population

M=M=s= =N The M=M=s model with nite population

M=G=1 Poisson input, general service time model with 1 server

M=Ek =1 Poisson input, Erlang service time model with 1 server

P0 Pn

n
M=M=1 1 P0 , n 1
8
>
< ( = )n P0
n!
, n = 1; 2; : : : ; s
1
M=M=s sP1
( = )n
n!
( = )s
+ s![1 =(s )] >
: ( = )n P0 , n = s; s + 1; : : :
n=0 s!sn s
8 8
>
< 1 >
< n P0 ,
1 K+1 , 6= 6=
M=M=1=K , n = 1; : : : ; K
>
: 1 >
: 1 ,
K+1
= K+1
=
8
>
< ( = )n P0 , 1 n s
1 n!
M=M=s=K P P
s
( = )n ( = )s
+ s!
K
(s )
n s >
: ( = )n P0
n=0
n!
n=s+1 s!sn s
, s n K
n
1 N!
M=M=1= =N P
N n (N n)!
P0 , n = 1; 2; : : : ; N
N!
(N n)! ( )
n=0
8
n
>
< N!
(N n)!n!
P0 , 1 n s
1
M=M=s= =N sP1 P
N!
( )
n
+
N
N!
( )
n >
: N!
n
n=0
(N n)!n!
n=s (N n)!s!sn s
(N n)!s!sn s P0 , s n N

3
L Lq
2
M=M=1
( )
P0 ( = )s
M=M=s Lq + s!(1 )2
8
>
< (K+1) K+1
1 1 K+1
, 6=
M=M=1=K L (1 P0 )
>
: K
2
, =
8
>
< P0 ( = )s [1 K s (K s) K s (1 )]
sP1 sP1
s!(1 )2
, 6= s
M=M=s=K nPn + Lq + s 1 Pn
n=0 n=0 >
: 1 P0 ( = )s [K s+(K s)2 ]
2 s!
, =s
+
M=M=1= =N N (1 P0 ) N (1 P0 )
sP1 sP1 P
N
M=M=s= =N nPn + Lq + s 1 Pn (n s)Pn
n=0 n=0 n=s
2 2
+ 2
M=G=1 + Lq 2(1 )
2
1+k
M=Ek =1 + Lq 2k ( )

W Wq

1
M=M=1 ( )

1
M=M=s Wq + Lq =

M=M=1=K L= Lq = (1 PK )

M=M=s=K L= Lq = (1 PK )

M=M=1= =N L= Lq = (N L)

M=M=s= =N L= Lq = (N L)
1 Lq
M=G=1 Wq +
1 1+k
M=Ek =1 Wq + 2k ( )

Pr(W > t) Pr(Wq > t) Pr(Wq = 0)

(1 )t (1 )t
M=M=1 e ,t 0 e ,t 0 1
h i sP1
t P0 ( = ) s 1 e t(s 1 = )
s (1 )t
M=M=s e 1+ s!(1 ) s 1 =
[1 Pr(W = 0)]e Pn
n=0

E(T C) = E(SC) + E(W C) where E(SC) = Cs s and E(W C) = Cw L.

4
Chapter 26: The Application of Queueing Theory (Skip)
As before, E(T C) = E(SC) + E(W C), but now,
P1
with g(N ) form, E(W C) = n=0 g(n)Pn ,
R1
with h(W) form, E(W C) = 0
h(w)fW (w) dw.

Vous aimerez peut-être aussi