Seminar Topics

Internet of Things -- Satish Srirama, Mohan Liyanage and Chii Chang (Responsible persons)

1. Smart Cities and the Internet of Things

In this topic, student will study what are the promising application use cases/scenarios for IoT-based smart cities.

2. A Study on Service Description Approaches in the Internet of Things

Various approaches have been introduced for description of the Internet of Things (IoT) services/resources. From the Service-Oriented Architecture (SOA) perspective, WSDL, WADL, DPWS were used. SensorML, SenML, were proposed for sensor devices. UPnP was long introduced for general office devices. Recent IETF CoAP (RFC7252) & CoRE standards also introduce their approaches for describing resources. Bluetooth has its own protocol and approach such as UUID and many different ports to describe bluetooth devices and resources. Moreover, we haven’t mention about the JSON-based service descriptions. It is now becoming impossible to rely on one single standard to enable autonomous machine-to-machine (M2M) communication because there is no global common standard for machine readable metadata. In this topic, student will study and compare all the service/resource description languages for IoT.

3. A Framework for Trustworthy Internet of Things

Security is one of the major challenges in the Internet of Things (IoT). Although various security-related works have been done for IoT, existing works were only based on the classic network security-aspect. For instance, “Cryptography alone cannot solve protecting information in IoT problem as internally compromised nodes can generate bogus information and still authenticate it using valid cryptographic” (Lize, Jingpei, & Bin, 2014). Further, the centralized solutions are not feasible for IoT since fundamentally, IoT is based on distributed environment. Assuming there can be a central management party to govern the entire environment is not realistic. Hence, IoT requires a feasible distributed trust strategy to overcome the drawback of existing security models......(see detail)

4. The ONE Simulator for DTN Protocol Evaluation

Delay Tolerant Networking (DTN), based on the store-carry-and-forward routing principle. If the next hop is not immediately available for the current node to forward a message, the node will store the message in its buffer, carry it along while moving, and forward it to other appropriate nodes until the node gets a communication opportunity which helps to forward this message farther. The ONE is a simulation environment that is capable of generating node movement using different movement models, routing messages between nodes with various DTN routing. In this topic, student will provide a comprehensive study of DTN routing algorithms with ONE simulator.

5. A Literature Survey on the Internet of Things Middleware

Student will study and compare the recent European Commission IoT projects.

6. A Study on Internet of Things Platform

Student will study one or more IoT platforms from OpenIoT, FIWARE, Amazon AWS IoT.

Mobile Cloud -- Satish Srirama (Responsible persons)

1. Augmented reality (Google Maps, GPS, info from Wikipedia or other sources)

- wifi/mobile network enabled phone gets info from gps, google streetview, match image, show information about place in phone)

- image recognition from live stream. A computer vision algorithm for looking direction detection in real-time.

2. NFC developments, possibilities, applications

"semantic city" - develop nfc app and system so tags placed in city and according to liking some of the places (by nfc touch) app suggests other places nearby to go visit or see, map

SciCloud -- Pelle Jakovits (Responsible person)

1. New Generation MapReduce - Apache Beam and Apache Tez When MapReduce first came out, it was very widely adopted. Since then, a number of higher level frameworks such as Pig or Hive have been introduced that simplify data processing but still use MapReduce internally. MapReduce is known to have many limitations, such as always having to write input, intermediate and output data to disks, and relatively slow configure and start up time. Apache Tez is designed to directly replace MapReduce for these higher level frameworks and Apache Beam is an open source alternative to MapReduce designed by google for dataflow processing. Student should give an overview of the conceptual changes that Tez and Beam introduce, describe their advantages and respective differences and demonstrate their usefulness with real examples from a chosen field of data science. Student can also choose to compare them to an earlier MapReduce alternative: Spark.
2. Real time & Large Scale data processing with Apache Storm Apache Storm is one of the first real time stream processing frameworks. Typically stream processing frameworks buffer incoming data and process them in batches, but Storm allows you to process any incoming data object in real time. Task of the student is to give an overview of the Apache Storm frameworks and its capabilities, compare it to other similar tools, implement some use cases to demonstrate its usefulness & impact.
3. GPU enabled MapReduce The goal of this topic is to study what is the state of art in using GPU's to accelerate the in-Map and in-Reduce computations in MapReduce, what is the best practices to 'move' Map & Reduce input data and GPU's, and how advanced are the Java GPU interfaces in comparison to other programming languages (C, C++, ...)
4. Automatic object recognition in SAR satellite data using distributed processing algorithms - Student should give an overview of the state of the art in this topic and try to apply the newest methods to real SAR satelite data (Satelite pictures taken of Estonia)
5. MPI on Hadoop Yarn. Hadoop Yarn is a all new Hadop version that separates the cluster resource management and HDFS from the computation model MapReduce and allows to switch it out for other computing engines. Student will study what is the state of the art in the available alternative computing engines for Yarn and will specifically focus on MPI (Message Passing Interface) implementations of Yarn. (This topic is taken)

Hadoop cloud computing projects -- Pelle Jakovits (Responsible person)

1. NEWT - A fault tolerant BSP framework on Hadoop YARN - NEWT is a HPC framework ontop of Hadoop YARN which addresses the MapReduce issues with more complex algorithms. It was developed in our group and is in a working prototype state. Student should try to use NEWT to implement one or two scientific algorithms on NEWT and measure its efficiency and parallel speedup in a cluster. One outcome could be also a tutorial on how to adapt algorithms to to NEWT.
2. HARP large scale processing framework – Student should study how HARP can be used to parallelize scientific computing algorithms or process large scale data.
3. Cloudera Impala is based on Google Dremel and aims to provide Real-Time queries ontop of Apache Hadoop. Impala raises the bar for query performance while retaining a familiar user experience. With Impala, you can query data, whether stored in HDFS or Apache HBase – including SELECT, JOIN, and aggregate functions – in real time. Furthermore, it uses the same metadata, SQL syntax (Hive SQL), ODBC driver and user interface (Hue Beeswax) as Apache Hive, providing a familiar and unified platform for batch-oriented or real-time queries. Student should study how Impala can be used to speed up MapReduce or Hive computations by replacing some of the computing tasks with Impala queries instead.

Cloud deployment -- Pelle Jakovits (Responsible persons)

1. Automatic deployment of scientific computing experiments using CloudML. CloudML enables the modelling deploying complex software systems in the cloud. CloudML is focused on enterprise service oriented applications and does not provide means to interact with already deployed system. The goal of this topic is to study what is needed to support executing scientific computing experiments on systems deployed with CloudML and how to integrate this approach with the CloudML engine.
2. An interactive not-so-random visualizer for CloudML. CloudML enables the modelling deploying complex software systems in the cloud. One of it's advantages is that the user retains the model that represents the deployed system, and thus can modify the model at any time and re-deploy it. The goal of this work is to improve the graphical user interface for CloudML which currently is in a prototype state, and to study how to keep track of changes that are made to the deployed system over time.
3. CloudML comparison to other cloud deployment managers CloudML enables the modelling deploying complex software systems in the cloud. The goal of this topic is to compare the functionality and ease of use of CloudML to other existing cloud deployment tools such as Cloudify, Puppet or Chef and to propose improvements to CloudML that would simplify it's real life use.