Edge computing has attained significant attention in recent years as it offers several advantages for industrial applications compared to classical IoT approaches. In particular, large amounts of data can be processed locally, i.e., close to the field devices such as sensors and actuators. Typical applications include monitoring of production lines and predictive maintenance of machines. Additionally, edge computing helps to protect sensitive, enterprise-critical data, and cloud usage fees can be reduced since there is no need to send all data to the cloud, e.g., for pre-processing and analysis. Finally, latencies can be minimized enabling even real-time applications.
As a key feature of many state-of-the-art edge platforms, apps may be purchased on a marketplace and deployed remotely to one of more edge devices. Similar to smartphones, however, the licensing and billing options are rather limited today. Frequent options are lifetime purchase or subscription on a monthly or annual basis. For industrial customers, more advanced options that depend on the actual use are desirable (pay-per-use). This requires metering the resources an app consumes, e.g., total run-time, CPU time, memory consumption, network traffic, and usage of dedicated hardware like GPUs.
While modern operating systems provide various mechanisms for process monitoring, app metering remains challenging for the following reasons: Unlike cloud providers, who have full control over the hardware in the data centers, edge devices are located at the customer’s site. Consequently, it is hard to ensure that metering data is not compromised, whether intentionally or accidentally, e.g., due to misconfiguration of the operating system. This not only affects the metering itself but also transmission of the resulting data to the cloud for billing. For example, man-in-the-middle attacks may be exploited for fraudulently increasing revenues of the app vendors. Moreover, software bugs may result in wrong bills if, for example, certain events are missed such as starting or stopping an app. In case of a legal dispute, reliable information about the actual resources an app consumed is mandatory. This can be achieved by means of consistency checks that relate different kinds of information. To sum up, the integrity of metering data is essential for more sophisticated business models.
Aufgabensteller:
Prof. Dr. D. Kranzlmüller
Dauer der Arbeit:
Anzahl Bearbeiter: 1
Betreuer (in Zusammenarbeit mit einem industriellen Partner):