You are on page 1of 6

COMMUNICATION

Empiric Search of Optimal Data Displacement


in the Distributed Information System



.



.

,
,
,

.
,

,
.

/ MEIKSHAN V.

, ,

,

.
, -, ,
.

,

-
. , The article discusses the task of optimal
()
arrangement of fragments in the distributed

data base within the units of computer network
with random topology. Minimum efforts for trans mission of data through the network
, channels while processing received requests
.
is used as the criterion of optimality.

For approximate solution of the formulated
,
optimization problem, which is related
.
to the class of problems connected with nonlinear
(, mathematical programming with Boolean
)
variables, the article proposes an empiric
, procedure which works according to the scheme
of greedy algorithm. Using numerical
. , parameters, the article shows that the received

suboptimal distribution differs slightly
,
from the accurate solution found by the method
of exhaustive search.

- ,
- .
(),
- , Si : i = 1, N
- () -

30

4.indd 30

| | 4 2009

29.10.2009 11:22:46

( ..).
Si :
(),
;
,
();
.

, {F j : j = 1, M }.
( , , ,
)
.
,
X = xij : i = 1, N ; j = 1, M .
:


(
).
xij = 0
,
xkj = 1 k i .
,
,
Si Sk ,
ij .


,
. , F j
,
. , Si (i = 1, N ) j-
1, F j Si ;
xij = 0, .
(1) ij ,


ij .
, ..
,

. , j = 1, M
F j :

. ,

N
(2)
n j = xij 1.
. ,
i =1

, ,
. . , ,
() . , Si S j .
dij, dij = d ji dii = 0 (i = 1, N ).
, -
, .
. , -
[15], ( j = 1, M ) .
.
,
Si (i = 1, N ) - j- .
ij . Si
F j ( xij = 1 ), - ,

October | November | December 4 2009

4.indd 31

COMMUNICATION

31

29.10.2009 11:22:46

COMMUNICATION

,
.
,

,
[1]. (
)
,

.

,
,
. ,
. ,

.
( ),

.
, .

F j Si (
xij = 0 ), j -
,
,
ij ij .
Sm ( xmj = 1 m i ),
F j , , :

dim = min dik .


k : xkj =1

C2 ( X ) =

ij ij xkj dik .

i =1 j =1

k =1

,
:

C ( X ) = C1 ( X ) + C2 ( X ) =
N

(1 xij ) ijij

i =1 j =1
N

min dik +

k : xkj =1

ij ij xkj dik min

i =1 j =1

(3)

k =1

(1) (2).
.

,


. ,
, .

(3).
xij = 1 F j , Si , f ji .

C2 (3) :

E ji =

kj kj kj .

Sk W ji


,

, j Si :
f ji .
N M
f ji C1 ( X ) = 1 xij ijij min dik .
k : xkj =1
j - , i =1 j =1
Si . , Si

, - F j ,
, ,
: .

32

4.indd 32

| | 4 2009

29.10.2009 11:22:46

, j - f ji ,
W ji . S k W ji
,
,
, kj ,
f ji C1 (3) :

E ji =

COMMUNICATION

kj kj kj .

Sk W ji

, B ji E ji

f ji . ,
B ji > E ji
f ji


.
(3) ,
,
.

()
.

,
,
:

1 ij >0;
xij =
0 ij =0.
, ,
, X ,
0. ,
j = 1, M n j
- ,

.

(),

October | November | December 4 2009

4.indd 33

(3)
. ,
f ji B ji > E ji ,
.
,
,
ji = B ji E ji .
,
,

.
, -

33

29.10.2009 11:22:47

COMMUNICATION

,
( n j > 1 j = 1, M ),
.
- .

,
, ( N =
= 4),
.
(/)
:

,
Si S j ,


(/): dij = 8 / vij .
V :

,

,
( M = 5).
, , ( ) , , :



.
C ( X (0) ) = 9475.

(
)
. 1,
2. . 1,
.
,
11
f ji f52 .

, ( ), , C ( X (1) ) =
=6792, .. 28%.
:


. 2. ;
.
*
C ( X
)=
= 2655.

34

4.indd 34

| | 4 2009

29.10.2009 11:22:47



,
.

Q = (2 N 1) M,
N = 4 M = 5 Q = 759375.
*
) = 2375 C ( X

.

,
11,8%.
*
*
X
X
,
( n j)
( j = 1, M ) F2.
, ,
.
,
.

COMMUNICATION



.

,
.
. , ,

,
1
.
,

.
,
.

1. .. . : ,
1990.
2. .., .., .., ..

. .: , 1999.

evaluation of data al
2 3. Karlapalem K., Ahmad I., So S., Kwok Y. Empirical
location algorithms for distributed multimedia database systems // Proc. of
the 21th Annual International Conference on Computer Software and Applica
tions. Washington, 1997. P. 296301.
4. Karlsson M., Karamanolis C., Mahalingam M. A Framework for Evaluating Replica Placement Algorithms // Technical Report HPL-2002-219.
Hewlett-Packard Labs, 2002.
5. Karlsson M., Karamanolis C. Choosing replica placement heuristics for
wide-area systems // Proc. of the 24th International Conference on Distributed
Computing Systems. Tokyo, 2004. P. 350359.

October | November | December 4 2009

4.indd 35

35

29.10.2009 11:22:48