Provisioning of containers for virtualized applications
In a general aspect, a computer-implemented method can include receiving a request to provision a plurality of containers of an application across a plurality of data center hosts and iteratively placing the plurality of containers on the plurality of data center hosts. The containers can be selected for placement based on one of a locality constraint and an association with previously-placed containers. Placement of a selected container can be based on, at least, compute requirements of the selected container, network requirements of the selected container, configuration of the plurality of data center hosts, and performance metrics for the plurality of data center hosts.
1. A computer-implemented method including executing, by at least one processor of a computing device, instructions to implement the method, the instructions being recorded on a non-transitory computer-readable storage medium, the method comprising:
receiving a request to provision a plurality of containers of an application across a plurality of data center hosts;
identifying, from the plurality of containers, a first container to provision based on a computing resource to be used to implement the first container and network requirements between the first container and the computing resource, the network requirements between the first container and the computing resource defined in a manifest file for the first container;
selecting a first host from the plurality of data center hosts to implement the first container based on compute requirements of the first container, the network requirements between the first container and the computing resource, configuration of the plurality of data center hosts, and dynamically measured performance metrics of the plurality of data center hosts;
identifying, from the plurality of containers, a second container to provision based on the second container meeting network requirements between the second container and the first container from among remaining unplaced containers;
automatically generating affinity requirements between the first container and the second container based on one or more affinity rules automatically derived from the manifest file, wherein the one or more affinity rules include affinity rules related to shared container resources; and
selecting a second host from the plurality of data center hosts to implement the second container based on compute requirements of the second container, the network requirements between the second container and the first container, the configuration of the plurality of data center hosts, the automatically generated affinity requirements, and the dynamically measured performance metrics of the plurality of data center hosts.
2. The computer-implemented method of claim 1 , wherein:
the compute requirements of the first container include a processor requirement of the first container and a storage requirement of the first container; and
the network requirements between the first container and the computing resource include at least one of a bandwidth requirement, a maximum number of hops requirement, and a latency requirement.
3. The computer-implemented method of claim 1 , wherein:
the compute requirements of the second container include a processor requirement and a storage requirement; and
the network requirements between the second container and the first container include at least one of a bandwidth requirement, a maximum number of hops requirement, and a latency requirement.
4. The computer-implemented method of claim 1 , wherein selecting the first host from the plurality of data center hosts to implement the first container is further based on, at least one of, security requirements of the first container, scaling requirements of the second container, and availability requirements of the first container.
5. The computer-implemented method of claim 1 , wherein selecting the second host from the plurality of data center hosts to implement the second container is further based on, at least one, of security requirements of the second container, scaling requirements of the second container, and availability requirements of the second container.
6. The computer-implemented method of claim 1 , wherein the one or more affinity rules further include one or more of affinity rules related to defined data center fault regions, a shared data file, a shared filesystem, a shared IP address, placement on a same host, and a proximity to a shared resource.
7. The computer-implemented method of claim 1 , further comprising:
identifying, from the plurality of containers, a third container to provision based on network requirements between the third container and the second container; and
selecting a third host from the plurality of data center hosts to implement the third container based on compute requirements of the third container, network requirements between the third container and the second container, the configuration of the plurality of data center hosts, and the performance metrics of the plurality of data center hosts.
8. The computer-implemented method of claim 7 , wherein selecting the third host from the plurality of data center hosts to implement the third container is further based on, at least one of, security requirements of the third container, scaling requirements of the third container, affinity requirements between the first container, the second container and the third container, and availability requirements of the third container.
9. A computer-implemented method including executing, by at least one processor of a computing device, instructions to implement the method, the instructions being recorded on a non-transitory computer-readable storage medium, the method comprising:
receiving a request to provision a plurality of containers of an application across a plurality of data center hosts;
identifying, from the plurality of containers, a first container to provision based on a computing resource to be used to implement the first container and network requirements between the first container and the computing resource, the network requirements between the first container and the computing resource defined in a manifest file for the first container;
selecting a first host from the plurality of data center hosts to implement the first container based on compute requirements of the first container, network requirements of the first container, configuration of the plurality of data center hosts, and performance metrics of the plurality of data center hosts;
identifying, from the plurality of containers, a second container to provision based on the second container meeting network requirements between the second container and the first container from among remaining unplaced containers;
automatically generating affinity requirements between the first container and the second container based on one or more affinity rules automatically derived from the manifest file, wherein the one or more affinity rules include affinity rules related to shared container resources; and
selecting a second host from the plurality of data center hosts to implement the second container based on compute requirements of the second container, network requirements between the first container and the second container, security requirements of the second container, scaling requirements of the second container, the automatically generated affinity requirements between the first container and the second container, availability requirements of the second container, the configuration of the plurality of data center hosts, and the performance metrics of the plurality of data center hosts.
10. The computer-implemented method of claim 9 , wherein the computing resource to be used to implement the first container is one of data storage and Internet access.
11. The computer-implemented method of claim 9 , wherein the first container is a first container type, and the second container is a second container type, the second container type being different than the first container type.
12. The computer-implemented method of claim 9 , wherein the first container and the second container are a same container type.
13. The computer-implemented method of claim 9 , wherein the performance metrics of the plurality of data center hosts are determined dynamically.
14. The computer-implemented method of claim 9 , wherein:
the network requirements between the second container and the first container is included in a manifest file for the second container.
15. The computer-implemented method of claim 9 , wherein:
selecting the first host from the plurality of data center hosts to implement the first container is further based on, at least one of, a security requirement of the first container and an availability requirement of the first container.
16. The computer-implemented method of claim 9 , wherein:
selecting the first host from the plurality of data center hosts to implement the first container is further based on relative weights for the compute requirements of the first container, the network requirements between the first container and the computing resource, and an availability requirement for the first container.
17. The computer-implemented method of claim 9 , wherein the one or more affinity rules further include one or more of affinity rules related to defined data center fault regions, a shared data file, a shared filesystem, a shared IP address, placement on a same host, and a proximity to a shared resource.
18. The computer-implemented method of claim 9 , further comprising:
identifying, from the plurality of containers, a third container to provision based on network requirements between the third container and the second container; and
selecting a third host from the plurality of data center hosts to implement the third container based on compute requirements of the third container, network requirements of the third container, security requirements of the third container, scaling requirements of the third container, affinity requirements between the first container, the second container, and the third container, availability requirements of the third container, the configuration of the plurality of data center hosts, and the performance metrics of the plurality of data center hosts.
19. The computer-implemented method of claim 18 , wherein the third container is of a type that is different than a type of the first container and different than a type of the second container.
20. A computer-implemented method including executing, by at least one processor of a computing device, instructions to implement the method, the instructions being recorded on a non-transitory computer-readable storage medium, the method comprising:
receiving a request to provision an application having an N number of containers across a plurality of data center hosts; and
determining respective placements for the N number of containers of the application in the plurality of data center hosts, until the respective placements for the N number of containers are determined, iteratively by:
selecting, for placement, a container from unplaced containers of the N number of containers, the selecting being based on one of a locality constraint and an association with previously-placed containers of the N number of containers;
automatically generating affinity requirements between the selected container and the previously-placed containers based on one or more affinity rules automatically derived from a manifest file, wherein the one or more affinity rules include affinity rules related to shared container resources;
identifying a set of data center hosts from the plurality of data center hosts based on compute requirements of the selected container, network requirements of the selected container, and the automatically generated affinity requirements;
assigning a respective score to each of the data center hosts of the set of data center hosts based on one or more scoring weights for the selected container, configuration of the plurality of data center hosts, and performance metrics for the data center hosts of the identified set of data center hosts;
selecting, based on the respective scores, a host from the set of data center hosts for placement of the selected container; and
updating a container configuration table with provisioning information for the selected container, the provisioning information including the selected host and one or more logical network paths for the selected container; and
provisioning the N number of containers on the plurality of data center hosts based on the container configuration table.
21. The computer-implemented method of claim 20 , further comprising:
receiving an indication of at least one of a change in the plurality of data center hosts and a change in the application;
identifying one or more containers to reprovision based on the indication; and
reprovisioning each of the identified one or more containers by, for each of the identified one or more containers being reprovisioned the method includes:
identifying a modified set of data center hosts from the plurality of data center hosts based on the indication, compute requirements of the container being reprovisioned, and network requirements of the container being reprovisioned;
assigning a respective score to each of the data center hosts of the modified set of data center hosts based on one or more scoring weights for the container being reprovisioned, configuration of the plurality of data center hosts, and performance metrics for the data center hosts of the modified set of data center hosts;
selecting, based on the respective scores, a host from the modified set of data center hosts for placement of the container being reprovisioned; and
updating the container configuration table with provisioning information for the container being reprovisioned, the provisioning information for the container being reprovisioned including the selected host from the modified set of data center hosts and one or more logical network paths for the container being reprovisioned.
22. The computer-implemented method of claim 20 , wherein placing the selected container includes placing multiple instances of the selected container in the plurality of data center hosts based on at least one of a scaling requirement for the selected container and an availability requirement of the selected container.
23. The computer-implemented method of claim 20 , wherein the one or more affinity rules further include one or more of affinity rules related to defined data center fault regions, a shared data file, a shared filesystem, a shared IP address, placement on a same host, and a proximity to a shared resource.