Architectural building blocks
- processing on the edge, including cloud computing infrastructure and edge resources
- orchestration Of distributed processing nodes, cope with sudden changes in Big Data
- self-adaptation to real-time changes, sense and cope with dynamics in velocity and variety
- proactivity regarding the anticipation of need for changes in the processing infrastructure
Technical objectives
- network virtualization
- dynamic monitoring in real-time processing architectures for Big Data
- situation-aware and context-driven adaptation recommender systems
- real-time mobile stream processing
- pro-active cloud computing
Network virtualization
-
Intersite network virtualization and security management:
- combine hardware and software network resources into a single administrative unit, including multi-cloud environments from by different cloud providers
- focus on usage control and access control technologies on different operational layers
Going beyond state-of-the-art
- deploy a virtualized OpenvSwitch infrastructure to build a set of virtual networks
- reinforced security at network level
- formulate firewalls and access detection rules
- control the SDN (software-defined network) controllers via a virtual network orchestrator
- obtain statistics about network usage and provide meaningful input to the resource manager
Dynamic monitoring in real-time Big Data processing architectures
-
Multi-layer cloud resource management and monitoring:
- monitor communication scenarios between different laaS cloud resource providers, mobile phones, embedded devices and generally resources at the edge
- coordinate and orchestrate distributed resources
Going beyond state-of-the-art
-
Detection of anomalous situations on the fly (not predefined anomalies):
- combine predective capabilities with the ability to recommend cloud resource adaptations
- develop a new semantic-based model for distributed real-time processing architectures
- monitor the real-time processing architecture
- support the meta-modelling of the adaptation process
- enable real-time changes of the processing pipeline
Situation-aware and context-driven adaptation recommender systems
-
Distribution management:
- allow for the definition of distribution constraints for proper behavior of data-intensive cloud applications (examples: response time, security constraints, …)
- constraints to be expressed during design-time, further refinement by extensions when instantiating an application, verified in real-time
- together with recommendations on new distributed opportunities
Going beyond state-of-the-art
-
Combine predictive capabilities with the ability to recommend cloud resource adaptations:
- develop a big data situation metamodel that can model situations relevant to cloud and edge resources topology, status and generic capabilites
- propose algorithms for devising proactive adaptation actions
Real-time mobile stream processing
-
Adaptive scheduling of IoT big data processing tasks between devices and the cloud:
- recommendations for the scheduling mechanisms to be given depending on the context and the situation
- scheduling system support by a mobile context analyzer, a situation detection mechanism, a resource adaptation recommender and a data-intensive application recommender
Going beyond state-of-the-art
-
Use conventional big data analytics and integrate promising concepts of edge computing:
- investigate the concept of bandwidth and medium utilization and decide on edge, cloud or hybrid computing
- investigate on acceleration concepts for a local analysis
- global analytics control local analytics and business rules
Pro-active cloud computing
-
Pro-adaptive cloud adaptation:
- realization of the integrated platform that adapts the mapping of the real-time processing network on the cloud / edge according to the dynamics of the workload
- workload dynamics are seen as variations of the data size or heterogeneity of the resources and will be recognized by a situation detection mechanism and a workload predictor
Going beyond state-of-the-art
-
Autonomous cloud management platform („proactive cloud automation”) that:
- provides a workflow catalogue system for provisioning and deployment workflows,
- using a scalable scheduler,
- the ability to connect to a variety of cloud providers