PrEstoCloud Banner
PrEstoCloud logo
Newsletter April 2019 - Issue 4

Message from the Coordinator

Hello everybody,

Welcome to the forth issue of the PrEstoCloud Newsletter. The project has entered its third and final year. It's exiting to see, how far research has advanced. Meanwhile we are happy about live demonstrations coming from the use case providers such as the logistic use case with 40,000 tracks, communicating with cloud services for performing big data analysis and resource management, the surveillance use case providing a large and advance system with security cameras, and last but not least the media use case providing a professional, and consumer oriented live video service with global optimization aspects and dynamic resource allocation.

In today's issue we're pleased to be able to share more details with you, such as information about the Workload Predictor, a module to monitor information and workload evolution over time to predict the workload that may be experienced in the near future. Lastly, we want to keep you up to date about events with PrEstoCloud participation, above all the appearance at the Hannover Messe, the recent workshop in Japan, and the internal Hackathon. We hope you'll find this publication useful and are looking forward to receiving your feedback!

Enjoy reading and stay curious!

With best regards,

Dirk Mayer, Software AG
Project Coordinator

Our Project Team

The consortium unites a total of eleven partner organizations from industry and research.

ACTIVEeon logo Adtitess logo cnrs logo cvs-mobile logo iccs logo
ijs logo liveu logo amram-tech logo nissatech logo Ubitech logo

PrEstoCloud Architecture and Technologies

Workload Prediction Mechanism

+ click for the full size

Introduction of Artificial Intelligence and Machine Learning in cloud infrastructure management and operations is highly excepted by all business analysts. In PrEstoCloud we open this path with our Workload Predictor.

The role of a workload predictor is to be able to predict the workload that will be experienced in the near future. In a nutshell, a workload Predictor performs predictive analytics on time series. Conceptually, it connects real-time streaming and machine learning (ML).

To design our system, we focused on finding the best technique considering the real-time Big Data processing topology of PrEstoCloud. The workload predictor collects the monitoring data from the Broker, calculates predictions related to the status of resources which may trigger an application reconfiguration through the Situation Detection Mechanism. Due to nature of the learning algorithm (on-line prediction based on a limited time window), the computation can scale properly.

The solution is based on the data-analytics approach, which start with a deep understanding of the real-time dataset through an intensive data processing pipeline. The goal of the pipeline is to determine the main characteristics of the monitored data such as CPU and Memory usage and then to prepare data for robust predictions. Therefore, our solution is based on the purely data-driven approach. It is an unsupervised approach and does not require labelled training sets.

PrEstoCloud on Tour

PrEstoCloud meeting in Nice and Hackhatons
Organised by the LiveU partner, a PrEstoCloud meeting was held on 5-7 March 2019 in Nice, France. Further steps towards project implementation were discussed. Notably, the PrEstoCloud consortium has arranged different developing hackathon events as integration sessions so far. Along these hackathons, we had two important objectives as follows:
  • Collecting feedback and evaluations on the PrEstoCloud platform, and
  • Encouraging the technical development of the PrEstoCloud platform including software developers.

We took advantage of the opportunity to join our forces especially software developers and focus on the integration activities to gather feedback and evaluations on the current version of the PrEstoCloud platform and its deployment methodologies. Therefore, three hackathons held in Nice (France) have been considered as (i) Overlay network, (ii) TOSCA specification and inventory, and (iii) Placement optimisation.

As results of the overlay network hackathon, we successfully deployed servers on two clouds in two different regions. To this end, two Amazon Web Services (AWS) gateways have been considered at Paris and US Virginia, and similarly with two Azure gateways.

The outcome of the TOSCA specification and inventory hackathon was a consensus which is needed for the representation of cloud and edge devices exploited for a particular fragment. This decision can be taken by the Control Layer, but the exact way that the TOSCA specification will be updated depends on the weighting functions which should be resolved by the continued discussion on the optimisation problem in the optimisation problem hackathon.

As results of the placement optimisation hackathon, we consider three different criteria to have an application working perfectly: (i) high availability as a security-oriented service, (ii) latency which provides an acceptable end-to-end service-level latency, and (iii) hosting costs which may be costly on a public cloud.

Joint Workshop organised by the Mellodic and PrEstoCloud projects: The International Workshop on Recent Advances for Multi-Clouds and Mobile Edge Computing (M2EC 2019)

M2EC 2019 has held in conjunction with the 33rd International Conference on Advanced Information Networking and Applications (AINA 2019), March 27-29, 2019 in Matsue, Japan.

M2EC 2019 focuses on research works addressing the challenges associated with the intersection of multi-cloud resource provisioning and mobile edge computing. Combining MEC in Multi-Cloud infrastructures can help to combat latency challenges imposed by the cloud-centric architectures. The intent of the workshop is to bring together people from research and industry, in order to provide a discussion forum for state-of-the-art topics related to cloud, multi-cloud and mobile edge computing technology, networks and applications. Accepted papers focus on context-aware authorization in multi-cloud deployments, real-time video streaming over IPv6+MPTCP technology, survey of multi-cloud computing approaches, cost benefit analyses of multi-cloud deployment of dynamic computational intelligence applications, data clustering for geographically distributed cloud deployments, situation detection on the edge as well as simulations of cross-cloud workloads.

Hannover Messe 2019
PrEstocloud belongs to one of the research projects being shown at Hannover Messe 2019 at the booth of Software AG. From 1 to 5 April, Software AG's research manager will be there to share the research findings to visitors from the manufacturing industries and the energy sector.

One of the PrEstoCloud highlights on Hannover Messe will be the interactive presentation of consortium members. Under the title "Software AG Research - Designing future of CVS Mobile Logistics with European partners in Horizon 2020," Dirk Mayer, Software AG will present the project and Marija Kokelj and Sebastijan Vagaja, both from CVS Mobile, will introduce the CVS logistic pilot, which belongs to one of the three uses cases: Vehicle telematics produce big volumes of multi-modal data; consequently, vehicle management is becoming a highly data-centric task. The use case scenario presents a transport logistics solution with analytics on telematics data that is able to extract important information through real-time computation at the edge of the network. An open discussion about transferability to other domain and industries will enable the exchange with the visitors.

Upcoming Event

  • Autonomic High Performance Computing workshop
    On July 15-19, 2019 in Dublin, Ireland, CNRS partner will organise the fifth International Workshop on Autonomic High Performance Computing (AHPC 2019). Autonomic computing requires monitoring, decision making, resource management, and actuation capabilities. The works span a wide spectrum of topics and expertise such as distributed systems, computer architecture, middleware services, databases and data-stores, high speed networks, machine learning, and control theory. The purpose of this workshop is intended to bring together specialists and researchers in those converging specialties to share views and address technical issues and developments for autonomic and high performance computing.

Dissemination Material

Join us on Facebook, Twitter and Linkedin

We invite you to use our online networking tools to make connections, exchange scientific ideas, and keep informed about the latest research findings. On Facebook, Twitter and Linkedin you will enter rooms for exchange of experiences, among professionals and stakeholders - a great opportunity to connect with a greater scientific an industrial community.

Facebook Twitter Linkedin

January 2017 - December 2019
Reference 732339 • Volume 4.2 mio. Euro • Partners 11