Information Technology Reference
In-Depth Information
(”non aggregated”). Assuming that the ACG is built by processing messages one
by one, sorted by their total order from a service interaction log, the proposed
approach follows some rules when processing a message m :
- If m is correlated to previous messages m 1 ,m 2 , ..., m p
which are necessary
already associated with existing nodes n 1 ,n 2 , ..., n q
in the ACG then two
cases are possible:
CASE 1: Every message in
{
m 1 ,m 2 , ..., m p }
is associated to a distinct
node in
{
n 1 ,n 2 , ..., n p }
.Inthatcase, m is associated with a node n such
that
1 <i<p, CCond ( ed ( n i ,n )) = CCond ( m i ,m ) and the label of n
is the name of the service generating m . If such a node does not exist in
the ACG, then it is inserted and its correlations to the adequate other
nodes are also added.
CASE 2:
n i ∈{
n 1 ,n 2 , ..., n q }
such that two or more messages from
{
are associated with it. If those messages are correlated
to m using the same CC and n i has the service generating m as label, the
first case applies. If those messages are correlated to m using different
CCs, each of them is then moved from its existing node to a new node
inheriting all incoming edges of the original node. This is done for every
node verified until the first case can be applied.
m 1 ,m 2 , ..., m p }
- If m is not correlated to any of the previous messages then it is associated
with a predefined node in the ACG called root node .
Following those rules, it is possible to gradually build the ACG graph by pro-
cessing the service interaction log message by message. At any stage of the ACG
construction, each message to be processed is matched with the nodes of the
ACG (sets of messages) to apply the above rules.
The algorithm described below associates messages with their corresponding
nodes using a single parsing of the event log. Additional notations concerning
the relation between a message and nodes sharing some of its attribute's values
are used as follows:
m
M associated with n such that m
m
M ,if
m ,we
note n
m as an order relation between a node and a message
meaning that the node has to be correlated to the node associated
with the message. In that case, we also note CCond ( n , m )=
CCond ( m , m ) the correlation condition between the node n and
the node associated to a message m .
For each message m in the message log, the proposed algorithm allows to
identify an existing node n in the ACG or add a new node to the ACG which
can be associated to m with respect to the ACG properties previously defined in
section 3.1 of this paper.
 
Search WWH ::




Custom Search