Main Content

Open Students Topics

Understanding the effect of collocating heterogeneous ML models on GPUs for scaling inference services

The goal of this thesis is to understand how the spatial co-deployment of different ML models affects their performance on GPUs and to derive criteria for placing and allocating model instances (as a responsibility of inference serving systems) to preserve performance at a cost-efficient level.
Flow caching for supporting efficient privacy-aware publish/subscribe over programmable switches

The goal of the thesis is to design a concept for caching flow table entries to enhance the efficiency in routing event messages while enforcing privacy by increasing the number of rules that can be assigned to a network device.
GPU Power Consumption Prediction for DL Inference Workloads

Evaluation should answer questions about the linearity or non-linearity of power consumption with co-located workloads (ideally, reasoning about the disaggregated power), and the accuracy of the prediction model (compared to the measured power draw).