![](bg4.png)
are currently not supporting transaction handling, our
model does not yet handle this parameter.
The life span defines how long an event is valid
for processing. We consider the same two values (im-
plicit, explicit) for the specification of the life span.
The consumption policy describes the order how
events are processed. (Koschel, 1999) defined four
policies. Chronicle, events should be processed by the
order of the event creation. Recent, the last received
event should be processed only. Continuous, the order
of receiving is the order of processing (FIFO). And
cumulative, events are processed as one whole group.
We consider the same values but have a special focus
on the details of ordered handling as it requires some
effort in a distributed system where events are prone
to arrive unordered. Therefore we expect some exten-
sion to the traditional ADBMS model.
The coupling mode is also one parameter for
transactional behaviors in database systems. It in-
dicates if the event happened within the transaction
(coupled) or not (decoupled). It defines also if an
event is thrown immediate or at the end of the trans-
action (deferred). In our current work we don’t cover
transaction handling explicitly and thus consider only
the deferred decoupled value.
The strategy defines how the rule execution is
triggered if multiple rules would be triggered by an
event. The ADBMS semantic definition considers the
following values: parallel, all matching rules are fired
in an unpredictable order; arbitrary, one matching
rule is picked randomly; priority, the rules have pri-
orities and the rule with the highest priority is fired;
static, a static order is given by an administrator; dy-
namic, the order is generated in runtime. In gen-
eral we aim to support all of the available parameters,
however with one important difference. As for a dis-
tributed event processing system it is usually the case
that multiple rule execution components exist and a
global ordering of the rule executions would be hard
to achieve. Thus we consider the given attributes per
processing component and not on a global scope. So
on a global scope we only support the parallel strategy
and allow a detailed specification per component.
4.2 Semantic Parameter Annotation
The Activity Service defines capsules as the glue be-
tween a raw event sender and the Activity Service
components. Aside from their task to translate data
formats, the capsule is responsible for the annotation
of the generated events with the semantic parameters.
As the capsule is specific to an event producer, it has
the knowledge to assign the parameters. Thus the
event source specific knowledge is encapsulated and
the the Activity Service is based on the generalized
semantic definitions.
5 CONCLUSIONS AND
OUTLOOK
Current cloud services for event processing lack a
well defined and vendor independent API and seman-
tics definition, which results in a high risk of a ven-
dor lock-in. With our Activity Service concept we
address these challenges by providing a generalized
event communication and processing service that can
bridge the gaps between multiple provider specific
proprietary services. It introduces a solid semantic
definition based on three categories: transport, event
and domain semantic parameters. The discussions
showed that the semantic of ADBMS can be used
as foundation for the event semantic parameters but
also showed the importance of the transport seman-
tic parameters due to the distribution in a cloud en-
vironment. The next step will thus be the definition
of these parameters. Our current implementation ef-
forts are for the moment mostly focused on overcom-
ing technological hurdles but based on our prototype
we will realize the aforementioned semantic param-
eters to provide working system that can be used to
evaluate the concepts for real world scenarios.
REFERENCES
Gatziu, S. and Koschel, A .et al.,1998 , Unbundling active
functionality. In ACM SIGMOD Rec., vol. 27, no. 1,
pp. 35–40.
Goyal, P. and Mikkilineni, R., 2009, Policy-based Event-
driven Services-oriented Architecture for Cloud Ser-
vices Operation and Management. In Proc. 2009 IEEE
Intl. Conf. on Cloud Computing.
Koschel, A. and Lockemann, P. C.,1998, Distributed events
in active database systems: Letting the genie out of
the bottle. In Data Knowl. Eng., vol. 25, no. 1-2.
Koschel, A., 1999, Ereignisgetriebene CORBA-Dienste für
heterogene, verteilte Informationssysteme. Disserta-
tion, Universität Karlsruhe, Germany.
Paton, N. W., editor, 1999, Active Rules for Databases.
Springer, New York.
Schaaf, M., Koschel, A., Gatziu, S., Astrova, I., 2010, An
Active DBMS Style Activity Service for Cloud Envi-
ronments. In Proc. of the 1st Intl. Conf. on Cloud
Computing, GRIDs, and Virtualization. Lisbon, Por-
tugal.
Schaaf M., Koschel, A., and Gatziu, S., 2011, Event pro-
cessing in the cloud environment with well defined se-
mantics. In 1st Intl. Conf. on Cloud Computing and
CLOSER2013-3rdInternationalConferenceonCloudComputingandServicesScience
348