Apparatus and method for dynamic resource allocation in cloud radio access networks
View Patent ↗Disclosed is an apparatus for dynamic resource allocation in cloud radio access networks, and the dynamic resource allocation apparatus includes: a deep reinforcement learning unit learning load fluctuation of a remote radio head by using deep reinforcement learning and predicting the load fluctuation of the remote radio head; a calculation unit calculating a computational resource of a virtual machine corresponding to the remote radio head by using the predicted load fluctuation; and an allocation unit allocating the calculated computational resource to the virtual machine.
1. An apparatus for allocating a dynamic resource, the apparatus comprising:
a processor configured to:
learn load fluctuation of a remote radio head by using deep reinforcement learning,
predict the load fluctuation of the remote radio head;
calculate a computational resource of a virtual machine corresponding to the remote radio head by using the predicted load fluctuation; and
allocate the calculated computational resource to the virtual machine,
wherein the processor is further configured to:
acquire a reward value from the virtual machine; and
predict the load fluctuation of the remote radio head by using the reward value,
wherein the processor is further configured to repeatedly perform learning until a difference value between a required value of the computational resource and an allocation value of the computational resource reaches a predetermined value or less, and
wherein the processor is further configured to, when there are a plurality of virtual machines, calculate the difference value by adding absolute values of required values of the computational resources for respective virtual machines and the allocation value of the computational resource.
2. The apparatus of claim 1 , wherein the processor is further configured to:
learn the load fluctuation of the remote radio head by using the deep reinforcement learning and
predict the load fluctuation of the remote radio head.
3. The apparatus of claim 1 , wherein the processor is further configured to learn the load fluctuation of the remote radio head by dividing a time domain and predicts the load fluctuation to correspond to the time domain.
4. An operation method of a dynamic resource allocation apparatus, the method comprising:
(a) learning load fluctuation of a remote radio head by using deep reinforcement learning and predicting the load fluctuation of the remote radio head;
(b) calculating a computational resource of a virtual machine corresponding to the remote radio head by using the predicted load fluctuation;
(c) allocating the calculated computational resource to the virtual machine;
(d) acquiring a reward value from the virtual machine and predicts the load fluctuation of the remote radio head by using the reward value,
wherein the predicting of the load fluctuation of the remote radio head by using the reward value is repeatedly performed until a difference value between a required value of the computational resource and an allocation value of the computational resource reaches a predetermined value or less, and
wherein when there are a plurality of virtual machines, the difference value is calculated by adding absolute values of required values of the computational resources for respective virtual machines and the allocation value of the computational resource.
5. The method of claim 4 , wherein in the predicting of the load fluctuation, the load fluctuation of the remote radio head is learned by dividing a time domain and the load fluctuation is predicted to correspond to the time domain.