I have heard and thought this phrase a lot when conducting 5G IoT experiments in recent years. An expression of discrepancy between the expected and the actual outcome of the experiment. Having an expectation upfront is of course a good habit. It prevents indiscriminate acceptance of the outcome of an experiment. If an outcome does not match the expectation, it is initially a disappointment. At the same time, it is a learning moment; where does that difference come from? From recent experience I observed some possible causes:
- Insufficient detailed information was included in the initial assessment
Causing the expected outcome to be insufficiently accurate. Sometimes tzhis happens out of sloppyness (too much happy flow thinking). But when a system is complex or insufficiently deterministic, the chance of a missed or misjudged parameter is unintentionally present.
- Errors or non-deterministic behavior in the (test)setup
No one is perfect, so mistakes do happen. Same for non-deterministic behavior, in self-developed components as well as in components from third parties. Take a 5G IoT test setup as an example. 5G technology is based on 3GPP standards, but they are quite extensive, complex, have a steep learning curve and are also continuously evolving. The 5G IoT end-to-end technology chain is also quite large and is offered in sub-components by many different parties. Think of modem, antenna, SIM, RAN, core network, cloud, etc. In addition, some implementation choices in the 3GPP standard are left to the technology and service providers themselves and they may even differ per operator/country/region. Similar sub-components therefore could behave slightly different under the same conditions. Partly due to the extensive ecosystem of technology and service providers, an interoperability challenge arises (similar to the DVD and Bluray era). A 100% test coverage is therefore a serious challenge.
- Environmental conditions
5G RF signals are weak and can easily be affected by environmental factors such as attenuation by walls and interference with other RF signals. Moreover, those environmental factors can vary greatly over time, for example due to mobility.
Something what should work fine in theory thus could behave completely different in practice. For the development of a robust 5G IoT system, it is therefore insufficient to stick to well-controlled lab tests. Of course, many problems can be discovered and solved with network simulators, but not all practical situations are easy (read: affordable) to simulate. Moreover, you can only simulate what you already know, and for that you must have hit the wall first. Field tests are therefore inevitable to reveal and mitigate all kinds of undesired behavior (unhappy flows).