Management and utilization of network interface card resources for workload execution
A method comprises receiving a request for at least one network interface card resource, and identifying one or more network interface cards comprising the at least one network interface card resource. In the method, at least one virtualized computing resource image is received, and a workload is executed using an instance of the at least one virtualized computing resource image running on the one or more network interface cards. Workload execution includes implementation of a binding mechanism using one or more control groups and a given protocol. In addition, a cluster including one or more nodes comprising the identified one or more network interface cards is generated.
1. A method, comprising:
receiving, in a cloud-based system comprising a plurality of host devices and a plurality of network interface cards, a request for at least one network interface card resource, the host devices having respective first operating systems and the network interface cards having respective second operating systems that are separate from the first operating systems of the host devices, the cloud-based system providing a multi-tenant environment in which network card resources are shared among multiple tenants, the request being received from a user device associated with a particular one of the tenants;
identifying, in the cloud-based system, one or more of the network interface cards comprising the at least one network interface card resource;
receiving, in the cloud-based system, at least one virtualized computing resource image;
determining, in the cloud-based system, at least one associated kernel image for the one or more network interface cards, the kernel image for the one or more network interface cards being different than kernel images of the host devices and identifying one or more namespaces associated with the particular tenant, the one or more namespaces being utilized in the cloud-based system to provide isolation in workload execution, on the one or more network interface cards, between the particular tenant and other ones of the multiple tenants; and
executing, in the cloud-based system, a workload using an instance of the at least one virtualized computing resource image running on the one or more network interface cards utilizing the one or more namespaces identified by the kernel image;
wherein the method is performed by at least one processing device of the cloud-based system, the at least one processing device comprising a processor coupled to a memory.
2. The method of claim 1 wherein the at least one virtualized computing resource image comprises one of a container image and a virtual machine image.
3. The method of claim 1 further comprising generating a cluster including one or more nodes comprising the identified one or more network interface cards comprising the at least one network interface card resource.
4. The method of claim 3 further comprising monitoring one or more metrics of the one or more nodes.
5. The method of claim 1 wherein the at least one processing device is a component of a cloud service provider hosting the one or more network interface cards.
6. The method of claim 5 wherein the at least one virtualized computing resource image is uploaded to the cloud service provider from a user device.
7. The method of claim 5 wherein the at least one virtualized computing resource image is executable on the one or more network interface cards and on a host device of the cloud service provider.
8. The method of claim 1 wherein the one or more network interface cards comprise smart network interface cards, and the at least one network interface card resource comprises at least one of a central processing unit, a graphics processing unit, a partition of a central processing unit, graphics processing unit computational time slices and graphics processing unit memory.
9. The method of claim 1 wherein executing the workload comprises implementing a binding mechanism using one or more control groups and a given protocol.
10. The method of claim 9 wherein the one or more control groups isolate the workload.
11. The method of claim 9 wherein the given protocol provides a programmable network data path in a kernel.
12. The method of claim 1 further comprising eliminating the instance of the at least one virtualized computing resource image following completion of the executing of the workload.
13. The method of claim 1 further comprising releasing the at least one network interface card resource following completion of the executing of the workload.
14. An apparatus, comprising:
at least one processor and at least one memory storing computer program instructions wherein, when the at least one processor executes the computer program instructions, the apparatus is configured:
to receive, in a cloud-based system comprising a plurality of host devices and a plurality of network interface cards, a request for at least one network interface card resource, the host devices having respective first operating systems and the network interface cards having respective second operating systems that are separate from the first operating systems of the host devices, the cloud-based system providing a multi-tenant environment in which network card resources are shared among multiple tenants, the request being received from a user device associated with a particular one of the tenants;
to identify, in the cloud-based system, one or more of the network interface cards comprising the at least one network interface card resource;
to receive, in the cloud-based system, at least one virtualized computing resource image;
to determine, in the cloud-based system, at least one associated kernel image for the one or more network interface cards, the kernel image for the one or more network interface cards being different than kernel images of the host devices and identifying one or more namespaces associated with the particular tenant, the one or more namespaces being utilized in the cloud-based system to provide isolation in workload execution, on the one or more network interface cards, between the particular tenant and other ones of the multiple tenants; and
to execute, in the cloud-based system, a workload using an instance of the at least one virtualized computing resource image running on the one or more network interface cards utilizing the one or more namespaces identified by the kernel image.
15. The apparatus of claim 14 wherein the apparatus is further configured to generate a cluster including one or more nodes comprising the identified one or more network interface cards comprising the at least one network interface card resource.
16. The apparatus of claim 14 wherein, in executing the workload, the apparatus is configured to implement a binding mechanism using one or more control groups and a given protocol.
17. The apparatus of claim 14 wherein the apparatus is further configured to eliminate the instance of the at least one virtualized computing resource image following completion of the executing of the workload.
18. A computer program product stored on a non-transitory computer-readable medium and comprising machine executable instructions, the machine executable instructions, when executed, causing a processing device:
to receive, in a cloud-based system comprising a plurality of host devices and a plurality of network interface cards, a request for at least one network interface card resource, the host devices having respective first operating systems and the network interface cards having respective second operating systems that are separate from the first operating systems of the host devices, the cloud-based system providing a multi-tenant environment in which network card resources are shared among multiple tenants, the request being received from a user device associated with a particular one of the tenants;
to identify, in the cloud-based system, one or more of the network interface cards comprising the at least one network interface card resource;
to receive, in the cloud-based system, at least one virtualized computing resource image;
to determine, in the cloud-based system, at least one associated kernel image for the one or more network interface cards, the kernel image for the one or more network interface cards being different than kernel images of the host devices and identifying one or more namespaces associated with the particular tenant, the one or more namespaces being utilized in the cloud-based system to provide isolation in workload execution, on the one or more network interface cards, between the particular tenant and other ones of the multiple tenants; and
to execute, in the cloud-based system, a workload using an instance of the at least one virtualized computing resource image running on the one or more network interface cards utilizing the one or more namespaces identified by the kernel image.
19. The computer program product of claim 18 wherein in executing the workload, the machine executable instructions cause the processing device to implement a binding mechanism using one or more control groups and a given protocol.
20. The computer program product of claim 18 wherein the machine executable instructions further cause the processing device to eliminate the instance of the at least one virtualized computing resource image following completion of the executing of the workload.