Aperiodic snapshot creation recommendation engine
An aperiodic snapshot recommendation engine running in a storage system aperiodically generates hints that a new snapshot should be created. The hints are sent to host servers to prompt snapshot generation commands to be sent to the storage system. The hints may be generated based on current storage system workload conditions using a model of a snapshot scheduler running on a host server for which the storage system maintains data. The model may be created using a machine learning technique. For example, machine learning may be used to model the host's snapshot scheduler in terms of storage system workload conditions existing when the snapshot scheduler commands generation of new snapshots during a training phase.
1. An apparatus comprising:
a storage system comprising at least one computing node that manages access to a plurality of non-volatile drives, the storage system comprising a snapshot recommendation engine that aperiodically generates a message indicating that a new snapshot should be created, the message being sent to a host server that comprises a periodic snapshot scheduler, wherein the snapshot recommendation engine generates a model of the periodic snapshot scheduler using features associated with storage system workload, thereby prompting the host server to aperiodically send a command to the storage system to cause a new snapshot to be created.
2. The apparatus of claim 1 wherein the snapshot recommendation engine generates the message indicating that the new snapshot should be created based on storage system workload since a most recent snapshot.
3. The apparatus of claim 1 wherein the host server comprises an aperiodic snapshot scheduler and the snapshot recommendation engine generates a model of the aperiodic snapshot scheduler using features associated with storage system workload.
4. The apparatus of claim 3 wherein the snapshot recommendation engine generates the message indicating that the new snapshot should be created based on storage system workload since a most recent snapshot.
5. The apparatus of claim 1 wherein the host server comprises an aperiodic snapshot scheduler and the snapshot recommendation engine generates a model of the aperiodic snapshot scheduler that predicts at least one of time of day and day of week at which new snapshots should be created.
6. The apparatus of claim 5 wherein the snapshot recommendation engine generates the message indicating that the new snapshot should be created based on the predicted time of day or day of week.
7. A method comprising:
using a storage system comprising at least one computing node that manages access to a plurality of non-volatile drives and a host server comprising a periodic snapshot scheduler:
using a snapshot recommendation engine generating a model of the periodic snapshot scheduler using features associated with storage system workload;
using the model to determine that a new snapshot should be created aperiodically;
aperiodically generating a message indicating that the new snapshot should be created;
sending the message to the host server, thereby prompting the host server to send a command to the storage system to cause the new snapshot to be created; and
responsive to the command, aperiodically creating the new snapshot.
8. The method of claim 7 comprising the snapshot recommendation engine generating the message indicating that the new snapshot should be created based on storage system workload since a most recent snapshot.
9. The method of claim 7 wherein the host server comprises an aperiodic snapshot scheduler and comprising the snapshot recommendation engine generating a model of the aperiodic snapshot scheduler using features associated with storage system workload.
10. The method of claim 9 comprising the snapshot recommendation engine generating the message indicating that the new snapshot should be created based on storage system workload since a most recent snapshot.
11. The method of claim 7 wherein the host server comprises an aperiodic snapshot scheduler and comprising the snapshot recommendation engine generating a model of the aperiodic snapshot scheduler that predicts at least one of time of day and day of week at which new snapshots should be created.
12. The method of claim 11 comprising the snapshot recommendation engine generating the message indicating that the new snapshot should be created based on the predicted time of day or day of week.
13. A computer program product on a non-transitory computer-readable medium, comprising:
instructions that generate a model of a periodic snapshot scheduler of a host server using features associated with storage system workload, wherein the model determines when a new snapshot should be created aperiodically;
instructions that aperiodically generate a message indicating that the new snapshot should be created, the message being sent to the host server, thereby prompting the host server to return a command to cause the new snapshot to be created; and
instructions responsive to the command to aperiodically create the new snapshot.
14. The computer program product on the non-transitory computer-readable medium of claim 13 wherein the instructions that generate the message do so based on storage system workload since a most recent snapshot.
15. The computer program product on the non-transitory computer-readable medium of claim 13 wherein the host server comprises an aperiodic snapshot scheduler and comprising instructions that generate the model based on storage system workload.
16. The computer program product on the non-transitory computer-readable medium of claim 15 wherein the instructions that generate the message do so based on storage system workload since a most recent snapshot.
17. The computer program product on the non-transitory computer-readable medium of claim 13 wherein the host server comprises an aperiodic snapshot scheduler and comprising instructions that generate the model by modeling at least one of time of day and day of week at which new snapshots should be created.