- Complexity is not well defined. The best definition I could find is a functional definition of the different characteristics of complexity, which is: https://www.britannica.com/science/complexity-scientific-theory/Surprise-generating-mechanisms
- Why is reducing complexity a good thing? It makes networks easier to deploy and manage. It also reduces the occurrence of those rare, exceptional events that cause everything to fall apart.
- Complexity is a slippery characteristic—it can easily be moved from one part of the system to another. For example, a new concept for reducing complexity can even be proved true for one characteristic, but it could increase complexity elsewhere in the system.
- Complexity mostly means a problem cannot be broken down into smaller pieces. Since the beginning of science, all our experimentation has involved breaking problems into small pieces and strongly believing in experiment isolation and replicability. If we want to tackle complexity properly, we need to address the system as a whole.
- This mostly means our models have to become more complex. A common misinterpretation is that if a model is more complex, the network itself becomes more complex. No, this only means that we observe more characteristics. The network remains the same regardless of how much we have in the simulation.
- When reducing complexity, we need to cut something out. However, we do not want to make our system less capable. So, we need to find something else that’s worth cutting. One alternative for core networks is: https://ieeexplore.ieee.org/document/9815730, practically implemented as #Fraunhofer #FOKUS Open6GCore https://www.open6gcore.org/
- Reducing sophistication is a “near concept” to reducing complexity—similar but not the same. This is easier to understand: we remove the 80% of the functionality that almost nobody uses and add an “on-demand” customization mechanism to bring it back for the few cases where it is needed. This is something we direly need for 6G.
- Miracle solutions never work—how many scientific articles suggest that if you add their functionality to the network, it will become less complex? “Add” is the word to concentrate on.
- It is becoming mentally difficult to handle the complexity. To address this, it is important to have a complex testbed environment to test whether the optimizations we develop make sense and ensure they do not negatively impact other technologies.
- When we first tackled this problem two years ago, the flexible and easy-to-use tool we needed for large-scale system simulations did not exist. So, we had to build it. More information at: OpenLANES—simulating large-scale network environments https://www.openlanes.net/.
Technology Standards[Technical Director] | 3GPP Expert - 5G/6G Standards | CTO office | Chair - Autonomous AI/ML framework - NGMN Author - Mobile Evolution - Insights | Senior Member IEEE Speaker | Patents | IETF RFCs
3 个月Reflective of managing complexity, which rises in terms of next-generation system evolution, which includes multiple sophisticated capabilities, where the operations are simplified and optimized to suit customized KPIs and KVIs, through an astute application of self-adaptive, autonomic principles, throughout the system.