| Speaker : | Francesca Bassi |
| SystemX | |
| Date: | 19/05/2021 |
| Time: | 2:00 pm - 3:00 pm |
| Location: | Zoom + LINCS |
Abstract
The principal tenet of C-RAN is the softwarization of the base-band signal processing, which enables the sharing of compute resources among multiple radio heads. Resource allocation in C-RAN is constrained by the hard deadline within which the computation in the shared resource pool needs to happen to preserve the functionality of the system.
When the aggregate demand of compute resources in the pool exceeds the capacity, a fraction of the radio packets is lost. A natural choice for the allocation policy is to minimize the loss rate. An alternative strategy is to maximize the PHY layer througput. Both strategies may be pursued by framing the allocation of compute resources as an optimization problem.
In order to select the best strategy it is useful to consider the impact of the choice on upper layers. We consider the performance at the MAC layer, after the HARQ process has taken place. Each packet lost due to lack of compute resources triggers an HARQ request of retransmission. If the loss happens on the last allowed transmission, all the radio resources (present and past) spent for the message go to waste. This consideration suggests that a policy of allocation of the compute resources accounting for the retransmission age of the frame may increase the performance.