Joint Compute and Connectivity Orchestration for Distributed 6G Systems

Topic description

Emerging applications increasingly distribute data processing and AI workloads across heterogeneous computing resources spanning end devices, edge infrastructure, and cloud platforms. Determining where a workload should execute can no longer be considered solely as a computing problem: moving processing from a local device to an edge resource creates communication requirements and consumes network resources.

This PhD will investigate the joint orchestration of computing and connectivity resources in 6G networks, considering communication and computation as parts of a single distributed processing infrastructure.

The research will investigate how workload characteristics and available computing resources determine whether processing should take place locally or be moved to edge infrastructure. Decisions will consider factors such as processing demand, accelerator availability, energy consumption, execution time, data volumes, and application performance requirements.

A central research challenge will be to quantify the cost of computation placement from the communication perspective. Offloading a workload may reduce local processing requirements or execution time, while simultaneously increasing data transfers, communication latency, energy consumption, and utilization of radio, transport, and edge-network resources.

The PhD will therefore develop mechanisms for dynamically coordinating compute placement and 5G-Advanced/6G network resource allocation. Rather than treating the network as a transparent transport mechanism, network conditions and resources will become explicit inputs to workload-placement decisions. Conversely, decisions to execute workloads remotely will dynamically generate connectivity requirements that must be translated into appropriate network resource allocations and QoS guarantees.

The research will explore architectures, models, optimization mechanisms, and intelligent orchestration approaches capable of jointly managing compute and connectivity resources across the device-edge continuum.

Key research questions

1. How can heterogeneous application and AI workloads be characterized in terms of their computing and communication requirements?
2. Under which conditions should processing remain local, and when should it be migrated or offloaded to edge resources?
3. What are the communication, latency, energy, and network-resource costs associated with computation offloading?
4. How should 5G-Advanced/6G resources be allocated according to workload placement and application requirements?
5. How can network state and available network resources influence real-time workload-placement decisions?
6. How can compute and network orchestrators exchange information and coordinate decisions efficiently?
7. Can learning-based or predictive mechanisms anticipate workload, compute, and network conditions and proactively optimize resource allocation?
8. What is the trade-off between the performance gains obtained through distributed processing and the communication overhead required to achieve them?
   
   Expected research outcome
   
   The PhD will develop and experimentally validate a joint compute-connectivity orchestration framework for 6G at the nextGlab testbed, capable of dynamically deciding where workloads should execute and coordinating the computing and networking resources required to support those decisions across the device-edge continuum.