Software driven long-term-evolution core network for mobile access
Systems and methods for managing network resources, including managing a generated virtualized data plane network using a central controller. Virtual machine (VM) resources are assigned to two or more different network functions at a local data center. Traffic is dynamically optimized based on at least one of aggregate traffic demands and quality of service (QoS) goals, and resource allocations and inter-data center (DC) bandwidth resources are determined for VMs for a plurality of services. VMs for each middlebox function and a routing plane for each service are configured based on the determined resource allocation, and flows are routed based on the resource allocation and one or more configured network paths using an overlay-routing framework.
1. A computer implemented method for managing network resources, comprising:
managing a generated virtualized data plane network using a central controller, the managing further comprising:
assigning virtual machine (VM) resources to two or more different network functions at a local data center;
dynamically optimizing traffic based on at least one of aggregate traffic demands and quality of service (QoS) goals; and
determining resource allocations and inter-data center (DC) bandwidth resources for VMs for a plurality of services;
configuring VMs for each middlebox function and a routing plane for each service based on the determined resource allocation, said middlebox function including transforming, inspecting, filtering and manipulating traffic for purposes other than packet forwarding; and
routing flows based on the resource allocation and one or more configured network paths using an overlay-routing framework.
2. The computer-implemented method of claim 1 , wherein the flows are routed across network function chains and a plurality of geographically distributed DCs.
3. The computer-implemented method of claim 1 , further comprising:
iteratively reassigning VMS and inter-DC resources across two of more DCs to minimize end-to-end delays until a threshold is reached.
4. The computer-implemented method of claim 1 , wherein the network paths are configured by:
classifying uplink and downlink traffic for one of more services using one or more routers;
performing label-switching using intermediate nodes; and
selecting a VM for a next function in a chain of service.
5. The computer-implemented method of claim 3 , wherein the two or more DCs are geographically remote DCs.
6. The computer-implemented method of claim 1 , wherein the controller is periodically updated with at least one of a total number of VMs available, an estimated current demand of each service obtained from a Serving Gateway (SGW)/Packet Data Network Gateway (PDN-GW), a location of VMs for each service on one or more remote DCs, and average end-to-end delays when each service is routed through a local DC and each of a plurality of remote DCs in a DC pool.
7. The computer-implemented method of claim 1 , further comprising leveraging consistent hashing to uniformly distribute one or more UEs across active Serving Gateway (SGW)/Packet Data Network Gateway (PDN-GW) VMs, wherein the SGW and PDN-GW are included in an Evolved Packet Core (EPC).
8. The computer-implemented method of claim 7 , wherein each SGW VM is represented by a number of tokens, the tokens being hashed to a hash ring with each VM being assigned to multiple points on the ring.
9. The computer-implemented method of claim 7 , wherein the one or more UEs are reassigned to new VMs by copying a state from an existing VM to a new VM to scale as a total number of UEs increases.