Information Technology Reference
In-Depth Information
different capabilities in economics such as autonomic (economic) mechanisms,
self-regulation, fault tolerance, as well as autonomic market (re)engineering. To
make a market autonomic, we propose applying the extended autonomic con-
trol loop, i.e. the MAPE-K cycle, to a complex array of parameterisable (hence
adaptable) economic components, where each component can be imagined as a
traditional managed element within the Autonomic Computing paradigm. We
specifically focus on market platforms for the domain of Cloud computing, as it
is well defined domain with respect to its requirements on a market platform.
A successful implementation of the Cloud computing methodology (i.e. fulfill-
ing its promise of computing as a commodity) is only possible with appropriate
methodologies and techniques for the definition and management of Cloud mar-
ket platforms. We believe that the application of our autonomic market concept
can tackle the challenges of the paradigm. In the remainder of this section, we
first provide a motivating example for autonomic markets, before briefly describ-
ing the application of the MAPE-K loop to an economic system.
2.1 Motivating Example
Consider that a market provider decides that a market goal is the completion of
a certain number of trading transactions per unit of time. Observing the mar-
ket's adherence to this goal is trivial. However, many different events can cause
deviations from this goal, some of which may be exogenous (e.g. external out-
ages) and others observable within the platform (e.g. infrastructure bottlenecks,
a reduction in the number of active participants, etc.). To remedy deviations
with respect to this goal, several different options can be considered depending
on the cause of the deviation. Examples are: (1) scaling computing nodes up or
the infrastructure as a whole out to increase the number of concurrent trades
per unit time, or reduce the time needed to process individual trades; (2) tuning
the matching algorithm to reduce the compute time (e.g. applying a heuristic
instead of an optimal algorithm); (3) purging the order book(s) to remove redun-
dant data; and (4) tuning allocation mechanism properties (e.g. the maximum
number of entries in the order book, the clearing or pricing functions). Moreover,
combinations of these options are also valid, as well as more aggressive adap-
tions such as changing the allocation mechanism for another. Although in this
example a simple market goal has been chosen, there are many more complicated
goals as well as goal combinations that can have large impacts upon the stability
of a market (e.g. goals concerning market liquidity, revenue, e
ciency, etc.).
2.2 Applying the Autonomic Loop to a Market Platform
The MAPE-K loop contains five elements: monitoring, adaptation, planning,
execution, and a knowledge management components, which we explain below.
Monitoring
data is critical for the instrumentation of any form of adaption.
In [6], we defined a monitoring methodology for an autonomic marketplace and
demonstrated how the performance of a market platform can be measured with
respect to a specific set of market performance indicators. This task is performed
