zur Startseite


Machine Learning for Offline Availability Prediction of PaaS Components
Betreuer MSc ETH Inf.-Ing. Otto Bibartiu
Prüfer Prof. Dr. rer. nat. Dr. h. c. Kurt Rothermel

Industry 4.0 and the Internet of Things (IoT) are of great interest for industry and research. In order to
keep up with the increasing demand of sophisticated IoT solutions, Robert Bosch GmbH operates an on
premise cloud stack. The Bosch IoT Cloud (BIC) enables enterprise service applications for the emerging
field of IoT. An IoT cloud solution is composed of different cloud services which are offered as Platform-
as-a-Service (PaaS). However, cloud services must meet the Quality of Service (QoS) requirements of
the overall solution. One important aspect of QoS is availability. The unavailability of a cloud service
might lead to a failure of the overall solution, which can result in significant loss of reputation and
ultimately loss of revenue for the service provider.
The goal of this thesis is to develop concepts which use machine learning methods, e.g. deep believe
networks, to predict the availability of base cloud services offered at the PaaS layer. Cloud services
are the main building blocks to implement IoT cloud solutions, but have received little attention with
respect to availability modeling so far.
In detail, the thesis should focus on one base cloud service, namely, the message-passing middleware
RabbitMQ. RabbitMQ offers different configuration possibilities which have different effects on avail-
ability. The student should evaluate the use of a deep believe network which outputs an availability
estimate based on any given configuration of RabbitMQ. The main challenge is to find suitable test and
training sets for the network model.
The final solution should allow for a ‘‘what if‘‘ analysis, i.e., by changing the configuration parameters of
RabbitMQ. The inverse, i.e., finding a configuration for a given target availability, is beyond the scope of
this work. Existing tools and libraries can be used, for instance TensorFlow. For evaluating the solution
and facilitate evaluations, a framework shall be implemented.


Background in machine learning with focus on neuronal networks is required. Good programming skills
in general. Advanced knowledge of Cloud technologies and database systems are beneficial.

Your Benefit

During the thesis, you will be researching at Bosch in Feuerbach, so that at the end you will be an expert
in cloud technologies and have profound knowledge of the Bosch IoT Cloud.