20.11.2012 Views

Contents Telektronikk - Telenor

Contents Telektronikk - Telenor

Contents Telektronikk - Telenor

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

A(rl ,rs ) =<br />

� �−L �<br />

ζl KB(rl) PrlB<br />

ζs<br />

(3.29)<br />

′ (rl)<br />

KB(rs) − PrsB ′ �<br />

(rs)<br />

��K �<br />

−1<br />

k=1,k�=l ζs − ζ −1�<br />

�<br />

k<br />

�K � �<br />

−1<br />

k=1,k�=l ζl − ζ−1<br />

k<br />

where ζi = ri /B(ri ) and L is the displacement<br />

in slots between the backward busy<br />

period and the busy period (for the periodic<br />

source) (see Figure 3).<br />

The generating function for the q M’s j in<br />

this case may be expressed as:<br />

K�<br />

q M j x j−1<br />

j=1<br />

= P(1 – ρ)E 1 (I + W) -1 V(x) (3.30)<br />

where E1 is the K dimensional row vector<br />

E1 = (1,0,....,0), and W = (W(i,l)) is the<br />

K × K matrix:<br />

W (i, l) = �<br />

� �M � �K−L−1 rl ζl<br />

s=K+1<br />

(3.31)<br />

and V(x) is the K dimension column vector<br />

V(x) = (V l (x)):<br />

(3.32)<br />

The overall cell loss probability for this<br />

special case may be written:<br />

(3.33)<br />

(where E T<br />

1 is the transposed of E1 ).<br />

For large M we can expand (I + W) -1 in<br />

(2.33), and by taking only the first term<br />

in the expression for W we get the following<br />

simple approximate formula for<br />

the overall cell loss probability:<br />

rs<br />

� KB(rl) − PrlB ′ (rl)<br />

KB(rs) − PrsB ′ (rs)<br />

�K k=1,k�=l<br />

ζs<br />

(x − ζk)<br />

Vl(x) = �K k=1,k�=l (ζl − ζk)<br />

�<br />

Vi(ζs)Vl(ζs)<br />

(1 − ρ)<br />

pL = −<br />

ρ E1(I + W ) −1 WE T 1<br />

pL ≈<br />

(1 − ρ) 2<br />

� �M � �K−L−1 1<br />

1<br />

rK+1 ζK+1<br />

ρ � P<br />

K (rK+1B ′ (rK+1) − B (rK+1)) �℘2<br />

(3.34)<br />

cell-loss<br />

cell-loss<br />

10 0<br />

10 -5<br />

10 -10<br />

10-15 0<br />

10 0<br />

10 -5<br />

10 -10<br />

20<br />

where ℘ is the product<br />

40<br />

60<br />

buffer size<br />

(3.35)<br />

For most of the cases of practical interest<br />

the formulas (3.34), (3.35) give satisfactory<br />

accuracy.<br />

80<br />

rho=0.8<br />

100<br />

rho=0.9<br />

5<br />

10<br />

15<br />

20<br />

speedup factor<br />

Figure 4 Cell loss probability as a function of buffer size and speed up factor at load<br />

0.9<br />

10<br />

0 10 20 30 40 50 50 60 70 80 100<br />

-15<br />

buffer size<br />

Figure 5 Cell loss probability as a function of buffer size for different speed up factors<br />

at load 0.8 and 0.9 (1 speed up = 1, 2 speed up = 5, 3 speed up = 10, 4 speed up = 15,<br />

5 speed up = 20)<br />

℘ =<br />

� K<br />

k=2 (ζK+1 − ζk)<br />

� K<br />

k=2 (1 − ζk) ζ = z<br />

B(z)<br />

5<br />

1<br />

25<br />

As for the multiserver case the product<br />

℘ may be given as an integral. We must,<br />

however, substitute for<br />

in the formulas (2.13) and (2.14) (and<br />

also changing n to K and A(z) to B(z) P ):<br />

(1 − ρB)<br />

℘ = (3.36)<br />

P (1 − ρ) (ζK+1 − 1) exp(I)<br />

(ζK+1) K<br />

1<br />

5<br />

215

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!