Dynamic inter-cloud placement of virtual network functions for a slice
Examples can include an optimizer that dynamically determines where to place virtual network functions for a slice in a distributed Telco cloud network. The optimizer can determine a slice path that complies with a service level agreement and balances network load. The virtual network functions of the slice can be provisioned at clouds identified by the optimal slice path. In one example, performance metrics are normalized, and tenant-selected weights can be applied. This can allow the optimizer to prioritize particular SLA attributes in choosing an optimal slice path.
1. A method for dynamic inter-cloud placement of virtual network functions (“VNFs”) in a slice path, comprising:
determining candidate slice paths relative to an edge cloud, wherein the candidate slice paths are different permutations of VNF-to-cloud assignments, the permutations being limited by a number of VNFs or a maximum number of intercloud links;
for the candidate slice paths, determining a load based on each cloud in the candidate slice path;
identifying a best composite slice path based on a composite score that includes a performance metric and the load, wherein the performance metric corresponds to a service level agreement (“SLA”) attribute; and
provisioning the VNFs at corresponding clouds specified by the best composite slice path.
2. The method of claim 1 , further comprising:
creating a matrix that maps the performance metric to each different cloud in the candidate slice paths, wherein the performance metric of a candidate slice path is derived from values in the matrix; and
when at least one candidate slice path complies with the SLA attribute, eliminating candidate slice paths that are non-compliant with the SLA requirement.
3. The method of claim 2 , wherein the SLA attribute is a maximum round-trip time.
4. The method of claim 1 , further comprising:
detecting network congestion or a performance metric below an SLA requirement;
determining new candidate slice paths; and
provisioning at least one of the VNFs at a new cloud specified by a new best composite slice path, the new best composite slice path having a new best composite score based on a combination of a weighted performance metric and weighted load for the new candidate slice paths.
5. The method of claim 1 , further comprising:
for each candidate slice path, calculating a normalized cost of the performance metric; and
when no candidate slice path includes a normalized cost that meets an SLA threshold, ranking the candidate slice paths based on how close the normalized cost is to the SLA threshold.
6. The method of claim 1 , wherein determining the candidate cloud paths includes:
determining a first set of slice paths that includes all permutations of VNF-to-cloud assignments that are within the number of VNFs and maximum number of intercloud links;
determining a second set of slice paths that complies with the SLA attribute; and
taking an intersection of the first and second sets.
7. The method of claim 1 , wherein the best composite slice path has the lowest composite score of multiple composite slice paths, wherein the composite score includes adding a weighted load to a weighted performance metric.
8. A non-transitory, computer-readable medium comprising instructions that, when executed by a processor, perform stages for dynamic inter-cloud virtual network function (“VNF”) placement in a slice, the stages comprising:
determining candidate slice paths relative to an edge cloud, wherein the candidate slice paths are different permutations of VNF-to-cloud assignments, the permutations being limited by a number of VNFs or a maximum number of intercloud links;
for the candidate slice paths, determining a load based on each cloud in the candidate slice path;
identifying a best composite slice path having a best composite score based on a performance metric and the load, wherein the performance metric corresponds to a service level agreement (“SLA”) attribute; and
provisioning the VNFs at corresponding clouds specified by the best composite slice path.
9. The non-transitory, computer-readable medium of claim 8 , the stages further comprising:
creating a matrix that maps the performance metric to each different cloud in the candidate slice paths, wherein the performance metric of a candidate slice path is derived from values in the matrix; and
when at least one candidate slice path complies with the SLA attribute, eliminating candidate slice paths that are non-compliant with the SLA requirement.
10. The non-transitory, computer-readable medium of claim 9 , wherein the SLA attribute is a maximum round-trip time.
11. The non-transitory, computer-readable medium of claim 8 , the stages further comprising:
detecting network congestion or a performance metric below an SLA requirement;
determining new candidate slice paths; and
provisioning at least one of the VNFs at a new cloud specified by a new best composite slice path, the new best composite slice path having a new best composite score based on a combination of a weighted performance metric and weighted load for the new candidate slice paths.
12. The non-transitory, computer-readable medium of claim 8 , the stages further comprising:
for each candidate slice path, calculating a normalized cost of the performance metric; and
when no candidate slice path includes a normalized cost that meets an SLA threshold, ranking the candidate slice paths based on how close the normalized cost is to the SLA threshold.
13. The non-transitory, computer-readable medium of claim 8 , wherein determining the candidate cloud paths includes:
determining a first set of slice paths that includes all permutations of VNF-to-cloud assignments that are within the number of VNFs and maximum number of intercloud links;
determining a second set of slice paths that complies with the SLA attribute; and
taking an intersection of the first and second sets.
14. The non-transitory, computer-readable medium of claim 8 , wherein the best composite slice path has the lowest composite score of multiple composite slice paths, wherein the composite score includes adding a weighted load to a weighted performance metric.
15. A system for dynamic inter-cloud virtual network function (“VNF”) placement in a slice, comprising:
a non-transitory, computer-readable medium containing instructions; and
a processor that executes the instructions to perform stages comprising:
determining candidate slice paths relative to an edge cloud, wherein the candidate slice paths are different permutations of VNF-to-cloud assignments, the permutations being limited by a number of VNFs or a maximum number of intercloud links;
for the candidate slice paths, determining a load based on each cloud in the candidate slice path;
identifying a best composite slice path having a best composite score based on a performance metric and the load, wherein the performance metric corresponds to a service level agreement (“SLA”) attribute; and
provisioning the VNFs at corresponding clouds specified by the best composite slice path.
16. The system of claim 15 , the stages further comprising:
creating a matrix that maps the performance metric to each different cloud in the candidate slice paths, wherein the performance metric of a candidate slice path is derived from values in the matrix; and
when at least one candidate slice path complies with the SLA attribute, eliminating candidate slice paths that are non-compliant with the SLA requirement.
17. The system of claim 16 , wherein the SLA attribute is a maximum round-trip time.
18. The system of claim 15 , the stages further comprising:
detecting network congestion or a performance metric below an SLA requirement;
determining new candidate slice paths; and
provisioning at least one of the VNFs at a new cloud specified by a new best composite slice path, the new best composite slice path having a new best composite score based on a combination of a weighted performance metric and weighted load for the new candidate slice paths.
19. The system of claim 15 , the stages further comprising:
for each candidate slice path, calculating a normalized cost of the performance metric; and
when no candidate slice path includes a normalized cost that meets an SLA threshold, ranking the candidate slice paths based on how close the normalized cost is to the SLA threshold.
20. The system of claim 15 , wherein determining the candidate cloud paths includes:
determining a first set of slice paths that includes all permutations of VNF-to-cloud assignments that are within the number of VNFs and maximum number of intercloud links;
determining a second set of slice paths that complies with the SLA attribute; and
taking an intersection of the first and second sets.