Information Technology Reference
In-Depth Information
graph is the result of an aggregation of messages into sets of messages (nodes of
the ACG) and an edge in the ACG represents correlations between all pairs of
messages in the two sets of messages. Without oriented edges, we won't be able to
decide which messages would be aggregated. For example, let assume
m
1
≪
m
2
are correlated with a CC
c
1
and
m
3
m
4
are correlated with the same CC
c
1
. Should we aggregate
m
1
with
m
3
and
m
2
with
m
4
or should we aggregate
m
1
with
m
4
and
m
2
with
m
3
? The orientation of the edges makes the decision
easier as the sources of edges are aggregated together and the destinations of
edges are aggregated also together.
≪
Role of the Root Node in ACG.
When processing each message sequen-
tially, some messages may not be correlated to messages previously placed into
the ACG. Such messages are associated by default with the root node which
corresponds to all non correlated messages (singles) and all messages being the
starting messages of process instances (starters). Without using such a root node,
we would create a separate node for every message not correlated to previous
messages which can be very frequent and thus can lead to a larger ACG. By
aggregating those events in the root node, we do not loose any information and
we reduce drastically the number of nodes in ACG.
Additionally, the weight of an edge corresponds to the number of couples of
messages associated with, on one hand its source node and on the other hand
its destination node, being correlated with the CC of the edge. Therefore, the
more a CC is verified for couples of messages, the bigger is the weight of the
corresponding edge(s)
2
.
3.2
Step By Step Scenario of Building the ACG
In this section we describe a step by step scenario of building the ACG from
a real world service interaction log as illustrated in figure 2. The ACG is built
gradually by processing every message sequentially. At each step, an overview of
the messages being processed as well as the resulting ACG under construction.
Each node of the ACG is labeled using the service operation of its members
(messages). The root node has an additional label corresponding to the service
operation of its source message (within the root node).
Let's start with the first message in the service interaction log. In the example
below, message 0001 is placed in the root node as this is the default node for
any message that is not correlated with previously placed messages. As this is
the first message to place in the ACG, it is obviously placed in the root node.
A second message, 0002, is correlated with message 0001 with a CC ”cur-
tomerid=customerid”. However, thereisnonodeintheACGlinkedtotheroot
node with correlation condition ”curtomerid=customerid”. The message 0002
is then placed in a newly created node and that node is linked to the root
node using an edge labeled with the CC ”curtomerid=customerid”. The new
2
Many edges within the ACG could have the same CC. This is made possible if such
CC is involved in many different processes.
