| 1. |
- Runeson, Per, et al.
(författare)
-
ESS Control System Data Lab - Executive Summary
- 2021
-
Rapport (övrigt vetenskapligt/konstnärligt)abstract
- Driven by the idea to use alarm data to explore machine learning across Industry 4.0 applications, the goal of this pilot study was to explore how to collect, store, manage and share data from the ESS Control System. Generally, we seek to make any control system data available for research and innovation but started with alarms as a feasible domain in which to explore machine learning. The goals were threefold, each explored in a work package:1. How to govern a data ecosystem, and which tools are needed to support it?2. How can alarm data be interpreted across industrial contexts, i.e., which meta dataand reference models are needed?3. How can data sharing be practically and legally handled at ESS?In summary, we identify a set of potential alleys for continued work to foster industrial innovation and collaboration in a control system data ecosystem with ESS as a catalyst.
|
|
| 3. |
- Hall, Tomas, et al.
(författare)
-
Performance Evaluation of KauNet in Physical and Virtual Emulation Environments
- 2012
-
Rapport (övrigt vetenskapligt/konstnärligt)abstract
- Evaluation of applications and protocols in the context of computer networking is often necessary to determine the efficiency and level of service they can provide. In practical testing, three different options are available for the evaluation; using a physical network as a testbed, using an emulator to simplify the infrastructure, or using a simulator to remove reliance on infrastructure entirely. As a real network is costly and difficult or even impossible to create for every scenario, emulation and simulation is often used to approximate the behavior of a network with considerably less resources required. However, while a simulator is limited only by the time required to perform the simulation, an emulator is also limited by the hardware and software used. It is therefore important to evaluate the performance of the emulator itself, to determine its ability to emulate the desired network topologies.The focus of this document is the KauNet emulator, an extension of Dummynet that adds several new features, primarily deterministic emulation of various network characteristics through the use of pre-generated patterns. A series of tests were per- formed using a testbed with KauNet in both physical and virtual environments, as well as a hybrid environment with both physical and virtual machines. While virtualization greatly increases the flexbility and utilization of resources compared to a pure physical setup, it may also reduce the overall performance and accuracy of the emulation.From the results achieved, KauNet performs well in a physical environment, with a high degree of accuracy even at high traffic loads. Virtualization on the other hand, clearly introduces several issues with both processing and packet loss that may make it undesirable for use in experiments, although it may still be sufficient for scenarios where the requirements for accuracy are lower. The hybrid environment represents a compromise, with both performance and flexibility midway between the physical and fully virtualized testbed.
|
|
