PANACEA Innovations

The PANACEA solution provides six main functionalities or innovations: 

  • Machine Learning Framework: Provides the ability to predict at runtime the time to crash of cloud applications and the response time of servers. When it is predicted that an application is about to fail, proactive reconfiguration mechanisms (e.g., rejuvenation) can be activated. Our solution relies on offline learning, where a number of data samples, which are collected by observing resource utilization while the system is running, are used to build the prediction model through ML techniques. This model is then used at runtime to predict the time to failure on the basis of current measurements of the system resource utilization. Potential beneficiaries of this innovation would be those ones associated with real-time and mission-critical services like telecommunication companies or smart cities.

 

  • Autonomic Cloud Service Management: Provides self-awareness (sensor mechanism) and self-configuration (effector mechanism), as well as reconfiguration operations in the Cloud Manager. With the appropriate policies implemented in the service, these mechanisms allow services to be self-managed. This innovation is very interesting for Service Providers, who can deploy unattended services (reducing OPEX) on Cloud Providers with improved availability and performance. Therefore, by offering this functionality, Cloud Providers can attract more Service Providers (especially those reluctant of the cloud due to its poor availability or performance) or build customer loyalty with the existing ones

 

  • Pervasive Monitoring: Provides a highly scalable solution for the monitoring of infrastructures and services. The monitoring system is based-on interacting agents that can read computing and network sensors to collect relevant parameters and that can make autonomous decisions on where to focus the monitoring effort. This innovation is interesting for a wide group of beneficiaries like service integrators, cloud providers, services providers and added-value resellers.

 

  • Overlay Network: Provides a self-healing, self-optimizing and highly scalable communication infrastructure that is able to monitor the quality of Internet paths between overlay nodes and to detour packets along an alternate path when the given primary path becomes unavailable or suffers from congestion. Due to the improvements offered by PANACEA over IP routing protocols, this innovation is very interesting for Developers of mission-critical applications, content providers and CDN operators.

 

  • Online QoS-driven Task Allocation: Provides the ability to dispatch incoming jobs to the best available resources in order to maintain and improve the requested QoS of the jobs. The task allocation system uses machine learning methods based on random neural networks in order to learn which job-to-resource allocations lead to good results, i.e., better QoS, and uses decisions based on the RNN to improve QoS. This innovation has especial interest for service providers, who can provide new and improved services through its use, and benefit from reduced OPEX.

 

  • Overlay Network Simulation: Provides a simulation environment for designing and optimizing overlay networks deployed over the Internet. The simulation environment is based on NEST (Network Engineering and Simulation Tool) and will enable the precise modeling of the underlying network architecture. By injecting adverse events (e.g., link/router failures, congestion, etc.) in the underlying network, we will be able to validate the self-healing and self-optimizing properties of the overlay network in a controlled environment and at a very large scale. Telecom and WAN operators are the main beneficiaries from this innovation, which would allow them to reduce OPEX and CAPEX