Skip to content

One year of academic research

The R&D Department has been very active this academical year 2010-2011 and has organized research activities in association with many academic organizations such as the Université Libre the Bruxelles (ULB-Belgium), the Université Catholique de Louvain (UCl-Belgium) and the University of Tunis El Manar (Tunisia). These collaborations have taken the form of students Master Theses and Master internships.

In total, four Master Theses have been conducted, with 2 students from the ULB and 2 students from the UCl, along with 3 internships with 2 students from the ULB and 1 student from the enginnering faculty of Tunis. Here is a summary of these work:

On the Study and Comparison of Elastic Cloud Databases.

Already presented on a blog post earlier this year, this master thesis aims at evaluating the elasticity of NoSQL databases. The architecture of these databases (Cassandra, HBase, Redis, Voldemort, …) allows large deployment on the cloud as well as online adding of additional instance nodes. In general, the elasticity, when referring to cloud-based systems, is their ability to 1°allow the performance to scale out proportionnaly to the number of running instances, without limitation on the number of runnning instances, and 2°their ability to quickly stabilize their perfromance while a new instance is added.

Nicolas Degroodt (ULB), the student who worked on this thesis came out with quantitative measures of the elasticity of NoSQL database Cassandra while in transition of a scaling out procedure, when an additional node is added to the cluster. Thibault Dory (UCl) came up with measures of the elasticity of the NoSQL Cassandra, HBase and MongoDB databases taking into account the state of the database before and after the stabilization. Dory’s great effort has also led to a conference paper (see other post also) accepted in the Second International Conference on Cloud Computing, GRIDs, and Virtualization (Rome).

Design of a Migration Architecture for a Communication Platform on the Cloud.

The goal of this Master Thesis was to propose a migration architecture for a SIP communication platform (Mobicents SIP servlets[1]), which already works well in a clustered environment, to a cloud infrastructure.

Thibault Leruitte (UCl) has came up with an appliance deployment solution based on BoxGrinder. Its solution also included the first steps towards the large scale deployment of the communication platform on the cloud as it also proposed a basic form of scaling-out of the SIP-servlet instances based on monitoring of the individual nodes with SteamCannon.

Event-driven Architecture for Intensive Care Unit Monitoring.

Nowadays the information technology has taken a big place up our everyday lives. The multitude of information platforms allow us to to efficiently live and work, be it for example when you check the status of your flight on your smartphone or when you get a traffic alert directly on your windshield. While this technology has been part of our habits for some years now, it has not yet penetrated much in medical and healthcare environments. In this context, we have collaborated with the ICU of the Mouscron (Belgium-Walloon region, represented by its ICU unit director, Dr Philippe GADISSEUX) and the ULB to develop a proof-of-concept of a sedation monitoring system based on an event-driven architecture. The sedation protocols have been followed for years now by the medical staff on the basis of a handwritten rules. The ICU having already integrated a fully computer-assisted monitoring solution for their beds, the idea is 1°to propose an EDA architecture which will track the sedation protocols on the bed screens and 2°to propose a proper language for the medical staff to easily define and refine their own protocols.

Renaud Marteleur, the ULB student working on the thesis, has come up with a solution relying on a message bus for propagating the monitoring event and a Business Rules Engines (Drools [2]) for the specification of the protocols.

On the Replacing of the Storage Engine of PigLatin (already presented on a previous post). PigLatin is a Yahoo! Project which provides a convenient language for defining parallel jobs for Hadoop, the Apache implementation of Google’s Map/Reduce. Its SQL-like semantics make easier for programmers to write Map/Reduce jobs than the native M/R language, while providing enough flexibility for custom processing.

During his internship, Arthur Lesuisse (ULB) managed to replace the storage engine of PigLatin to use the distributed NoSQL Database MongoDB. This permits to directly process data from the database (for example data collected from a backend website MongoDB database) without the need to firstly transfer it to the native PigLatin storage.

On the Cloud Computing Open Standards

These past years, numerous cloud computing offers have been flourishing for developers and end-users. These offers permit developers to deploy websites and applications with a on-demand schema on a large pool of computing resources. Amazon Web Services, one of the most famous cloud infrastructure and services providers, offers for instance cloud computing resources going from the raw computing instances to message queing and database services on the cloud. Nowadays, one the most important danger of those cloud solutions is the risk of vendor lock-in, due to the lack of (open) standards presented by these offers. Initiatives such as the Open Cloud Computing Interface (OCCi [3]) exist but are not mature yet and not widely adopted among the cloud providers. On the other side, platform abstraction services such as Rightscale or scalr exist and offer the developer the ability to define and configure services templates in an abstract way, the Rightscale [4] or scalr service then instanciates it on different cloud provider using the same configuration template. These platform abstraction service do however not make use of any open standards so far.

In an attempt to extend the open-source distribution of scalr, Amine Ghrab, our intern student from the Engineering Faculty of Tunis, has come up with an architecture and an implementation of a cloud platform abstraction service using the OCCi open standard.

Perspectives for the academical year 2011-2012.

Following up this new academical year are six new Master Theses and a Master internship. Our collaboration this year has even extended to the Université de Liège (ULg). Academic research is definetely active at Euranova!

References

[1] Mobicents – http://www.mobicents.org/

[2] Jboss Drools – www.jboss.org/drools

[3] OCCi – http://occi-wg.org/

[4] Rightscale – http://www.rightscale.com

[5] Scalr – http://www.scalr.com


 

Releated Posts

Calibrate to Interpret

Trustworthy machine learning is driving a large number of the ML community works in order to improve ML acceptance and adoption. In this paper, we show a first link between uncertainty and explainability, by studying the relation between calibration and interpretation.
Read More

Mass Estimation of Planck Galaxy Clusters using Deep Learning

Galaxy cluster masses can be inferred indirectly using measurements from X-ray band, Sunyaev-Zeldovich (SZ) effect signal or optical observations. Unfortunately, all of them are affected by some bias. Alternatively, we provide an independent estimation of the cluster masses from the Planck PSZ2 catalogue of galaxy clusters using a machine-learning method.
Read More