FlexCRM Project

SDR presents a hard real-time computing challenge with varying (computing) system conditions. The computing resource management framework thus needs to track the states of the computing resources for being able to take advantage of the reconfiguration capabilities of mobile terminals and network elements.

An SDR processing chain, SDR application or waveform, is the part of an SDR transceiver that is implemented in software. It can be understood as a set of concurrent processes that continuously process and propagate real-time data. Such a processing chain is not specifically tailored but rather executable on any general-purpose platform with sufficient computing capacity. Therefore, an automatic mapping process needs to dynamically assign software modules to hardware resources, while meeting all computing system constraints.

Our computing resource management framework is modular. It features the computing system modeling on the one hand and the computing resource management on the other (Figure 1). The SDR computing system modeling captures the SDR platforms' computing resources and the SDR applications' computing requirements. We, therefore, suggest equivalent metrics for modeling computing resources and requirements: million operations per second (MOPS) and mega-bits per second (Mbps), particularly, model the processing and interprocessor data flow capacities and requirements.

The ALOE computing resource management is based on two simple time management principles: time slots and pipelining (Figure 2). This facilitates the synchronized execution of the waveform modules on distributed computing resources while taking advantage of the continuous data flow that characterizes wireless communications. Based on these principles, ALOE applies a general-purpose mapping algorithm, the tw-mapping, and a problem-specific cost function. The cost function implements the computing resource management objective or policy that guides the allocation of computing resources to computing requirements in a controlled manner.

The tw-mapping, the cost function, and further details about the ALOE computing resource management approach can be found in document attached below.

We currently investigate how to minimize the pipelining latency at the mapping and scheduling stages. Therefore, we examine new cost functions and evaluate their performance in different computing resource management scenarios. We, furthermore, address the scalability of the tw-mapping with the objective of applying it to large processor arrays and multiprocessor systems-on-chip (MPSoCs). We work on extensions of the tw-mapping algorithm itself, but, also, on pre and postprocessing add-ons.

Back to Main