Workload Generators

The RECAP Workload Generators are application-specific and generic artificial workload generation tools that use statistical, machine learning, and simulation-based techniques to generate realistic workloads for distributed cloud, edge, and fog computing scenarios. They include:
• Simulation-based tools to generate workloads based on models for user behaviour, devices, mobility, and application interactions.
• Statistical methods for modelling and capturing workload behavior.
• Machine learning-based tools for estimating and classifying workloads.
• Diffusion-based tools for generating artificial workload traces to supplement measurements in distributed applications.

The techniques developed use a variety of application, network and domain knowledge in the data generation to capture the behaviour of distributed cloud / edge / fog applications and enable prediction capabilities that can be used in the development of proactive application and infrastructure optimisation tools. This, in turn, facilitates development and testing of application modelling and optimisation technologies without the dependencies of fine-grained simulators and testbed systems.

Benefits: The RECAP Workload Generators provide a suite of tools that enables developers and operators of distributed cloud, edge, and fog applications and infrastructures to experiment with workloads and evaluate system performance under scenario conditions that may be hard to predict or replicate algorithmically. The tools provide the capabilities needed to simulate and predictively evaluate the performance of applications in distributed networks, and are essential building blocks for parameterised scenario evaluations and ‘what-if’ analyses.

Licence: Open Source.

Contact: Umeå University, Sweden | P-OÖstberg | [email protected]

The RECAP Workload Generator/Anonymisation Tool takes as input a data set (e.g. a multidimensional time series) and uses neural networks to learn the statistical properties of the input set. Based on what it has learned, it generates new synthetic data sets with the same statistical properties as the original data set but not individual data items. In consequence, any potentially commercial sensitive or privacy sensitive information (e.g. IP addresses, user names, etc.) are removed. Such synthetic data sets can be used in all domains where production-like data is required e.g. software testing and benchmarking. Similarly, they can be used as input for developing and evaluating data analytics and machine learning mechanisms.

Benefits: This RECAP Workload Generator/Anonymisation Tool enables the creation of artificial data sets that resemble the original data sets. In the context of RECAP, it enables developers and operators of distributed cloud-edge applications and infrastructures to experiment with workloads and evaluate system performance. Being anonymised and synthetic, the generated data sets can be exchanged and analysed without the need to take privacy-preserving measures. As such, the data sets are not subject to GDPR and other privacy considerations.

Licence: Proprietary.

Contact: Ulm University, Germany | Dr. Jörg Domaschka | [email protected]