Side_1_360

(4.3)

Hence,

(4.4)

4.6.2 Average Loss Rate

Average loss rate is the number of packets that

are lost in a given time interval divided by the

duration of the time interval. The loss rate can

be defined e.g. end-to-end along a path, for an

outgoing link in a node or for a certain node.

The average unidirectional packet loss rate is

represented as shown in Table 4.2.

The average loss rate for packets travelling from

a given source ato a certain destination bis con-

sidered. The average end-to-end loss rate is de-

fined for the time window [t 1 , t 2 > and packets that

are still in transit at time t 2 are excluded. Hence,

{p|p∈Rb(t 1 , t 2 ) ∧src(p) = a∧dst(P) = b}.

The average end-to-end loss rate for packets

travelling along the path is defined by:

ηa,b(t 1 ,t 2 )= (4.5)

p∈P ̃

∑I(p)

t 2 −t 1

π(i,j)k

I(p)=

1

packet pwas lost somewhere

on its end−to−end path

0 packetcarried endpwas successfully−to−end

⎧

⎨

⎪

⎩

⎪

I(p)=

n∈pth(p)

∑ I(p,n)+

l∈lnk ( p )

∑ I*(p,l) 4.6.3 Unidirectional Packet Loss Ratio

The loss ratio is defined by the portion of pack-

ets that was lost. The loss ratio can e.g. be de-

fined for an end-to-end path, a certain node on

the path or a given link.

Then the loss ratio for packets travelling from a

given source ato a certain destination bis con-

sidered. The end-to-end loss ratio is defined for

the time window [t 1 , t 2 >. Hence, P= {p|p∈

Rb(t 1 , t 2 ) ∧src(p) = a∧dst(P) = b}.

(4.8)

|P| is the cardinality of set P.

5 Concluding Remarks

Active and passive measurements have different

pros and cons and are supplementary. Thus, an

operational measurement and monitoring plat-

form needs to include both active and passive

measurements. Passive measurements are re-

quired to e.g. monitor SLAs, perform detailed

performance measurements and collect data for

surveillance, accounting and pricing purposes.

On the other hand, active measurements are vital

to detect network failures and to test network

services. An operational measurement platform

should be tailor-made for observation of specific

parameters. Based on the parameters being ob-

served, data reduction techniques must be con-

sidered. The ultimate usage of the measurements

Proba,b(loss)=

I(p)

P ̃

∑

P

Description Definition

ηn(t 1 , t 2 ),average loss rate for node n. (4.x)

η*l(t 1 , t 2 ),average loss rate for link l= (n–1, n) η*l(t 1 ,t 2 )= (4.x)

p∈R*()n−1,n()t 1 ,t 2

∑ I*(p,l)

t 2 −t 1

ηn(t 1 ,t 2 )=

p∈Rn()t 1 ,t 2

∑ Ip( ,n)

t 2 −t 1

Table 4.2 Representation of
unidirectional packet loss rate

Description Definition

Probn(loss),loss ratio for node n. (4.6)

Prob*l(loss),loss ratio for link l= (n– 1, n). Prob*l(loss)= (4.7)

p∈R*()n−1,n()t 1 ,t 2

∑ I*(p,l)

p∈R*()n−1,n(t 1 ,t 2 )

Probn(loss)=

p∈Rn()t 1 ,t 2

∑ Ip( ,n)

p∈Rn(t 1 ,t 2 )

Table 4.3 Representation of
unidirectional packet loss ratio