Predicting excess capacity for radio access network edge systems
Disclosed are various embodiments for predicting excess capacity for radio access network (RAN) edge systems. In one embodiment, resource consumption data is generated by monitoring resource consumption of a plurality of implementations of a distributed unit (DU) and/or centralized unit (CU) used in RANs. A maximum level of resource consumption is predicted for a particular implementation of the DU/CU to be used in a particular RAN based at least in part on the resource consumption data. Excess resource capacity is predicted on a RAN-enabled edge server on which the particular implementation of the DU/CU is deployed for a cell site of the RAN based at least in part on the maximum level of resource consumption.
1 . A system, comprising:
a radio access network (RAN)-enabled edge server located at a cell site, the RAN-enabled edge server including a physical layer accelerator specialized for distributed unit (DU) physical layer communication; and
a computing device configured to at least:
generate resource consumption data by monitoring resource consumption of a plurality of implementations of a DU used in RANs;
predict a maximum level of resource consumption for a particular implementation of the DU to be used in a particular RAN based at least in part on the resource consumption data;
deploy the particular implementation of the DU on the RAN-enabled edge server; and
predict excess resource capacity on the RAN-enabled edge server based at least in part on the maximum level of resource consumption that is predicted.
2 . The system of claim 1 , wherein the excess resource capacity includes excess capacity on the physical layer accelerator that is predicted based at least in part on a maximum level of Layer 1 resource consumption that is predicted.
3 . The system of claim 1 , wherein the computing device is further configured to at least train a machine learning model based at least in part on the resource consumption data and data indicating corresponding spectrum assignments, wherein the maximum level of resource consumption is determined through the machine learning model.
4 . The system of claim 3 , wherein the computing device is further configured to at least update the machine learning model based at least in part on actual resource consumption of the particular implementation of the DU on the RAN-enabled edge server.
5 . The system of claim 1 , wherein the computing device is further configured to at least predict a quantity of reserved machine instances that can be allocated in the excess capacity based at least in part on the resource consumption data.
6 . The system of claim 1 , wherein the computing device is further configured to at least predict a quantity of ephemeral machine instances that can be allocated in the excess capacity based at least in part on the resource consumption data.
7 . A computer-implemented method, comprising:
generating resource consumption data by monitoring resource consumption of a plurality of implementations of at least one of: a distributed unit (DU) or a centralized unit (CU) used in radio access networks (RANs);
predicting a level of resource consumption for a particular implementation of the DU or the CU to be used in a particular RAN based at least in part on the resource consumption data; and
predicting excess resource capacity on a RAN-enabled edge server on which the particular implementation of the DU or the CU is deployed for a cell site of the RAN based at least in part on the level of resource consumption.
8 . The computer-implemented method of claim 7 , further comprising offering the excess resource capacity of the RAN-enabled edge server to third-party customers of a cloud provider network.
9 . The computer-implemented method of claim 7 , wherein the plurality of implementations of the DU or the CU corresponds to a plurality of different vendors of the DU or the CU.
10 . The computer-implemented method of claim 7 , further comprising:
performing a static analysis on the particular implementation of the DU or the CU; and
wherein predicting the level of resource consumption for the particular implementation of the DU or the CU is further based at least in part on the static analysis.
11 . The computer-implemented method of claim 7 , further comprising:
determining spectrum assignments of the cell site of the RAN; and
wherein predicting the level of resource consumption for the particular implementation of the DU or the CU is further based at least in part on the spectrum assignments.
12 . The computer-implemented method of claim 7 , wherein predicting the level of resource consumption for the particular implementation of the DU further comprises:
predicting a maximum level of Layer 1 resource consumption for the particular implementation of the DU, wherein the maximum level of Layer 1 resource consumption is static; and
predicting a maximum level of Layer 2 resource consumption for the particular implementation of the DU, wherein the maximum level of Layer 1 resource consumption is dynamic.
13 . The computer-implemented method of claim 12 , further comprising predicting excess capacity of a physical layer accelerator specialized for DU physical layer communication used in the RAN-enabled edge server based at least in part on the maximum level of Layer 1 resource consumption.
14 . The computer-implemented method of claim 12 , further comprising predicting a quantity of processor cores in the RAN-enabled edge server corresponding to excess capacity based at least in part on the maximum level of Layer 2 resource consumption.
15 . A computer-implemented method, comprising:
generating resource consumption data by monitoring resource consumption of a plurality of implementations of a distributed unit (DU) used in radio access networks (RANs);
predicting excess resource capacity on a RAN-enabled edge server on which a particular implementation of the DU is deployed for a cell site of a RAN based at least in part on the resource consumption data; and
determining a change in the excess resource capacity of the RAN-enabled edge server based at least in part on a change in Layer 2 resource consumption by the DU.
16 . The computer-implemented method of claim 15 , further comprising:
predicting a demand for the excess resource capacity on the RAN-enabled edge server; and
determining to enable or disable one or more processor cores of the RAN-enabled edge server based at least in part on the demand that is predicted.
17 . The computer-implemented method of claim 15 , further comprising offering the excess resource capacity on the RAN-enabled edge server for customers of a cloud provider network as a cellular availability zone of the cloud provider network.
18 . The computer-implemented method of claim 15 , further comprising migrating a workload of a customer of a cloud provider network away from the RAN-enabled edge server based at least in part on the change in the excess resource capacity in order to accommodate an increase in Layer 2 resource consumption by the DU.
19 . The computer-implemented method of claim 18 , wherein the workload is migrated to a different RAN-enabled edge server selected based at least in part on a proximity to the RAN-enabled edge server.
20 . The computer-implemented method of claim 15 , further comprising migrating a workload of a customer of a cloud provider network to the RAN-enabled edge server based at least in part on the change in the excess resource capacity from a decrease in Layer 2 resource consumption by the DU.