Data prefetching operations with trustworthy energy awareness
Energy aware data prefetching is disclosed. Data stored in a source storage may be retrieved or prefetched to a target storage based on a prefetching time predicted or estimated by an awareness engine. The awareness engine may include models configured to predict the prefetching time so that the prefetching time accounts for a cost and/or source of energy. The prefetching operation is performed at the prefetching time. The models of the awareness engine may also be configured to recommend a source for the prefetching operation from among multiple potential sources for the prefetching operation in an energy aware manner.
1 . A method comprising:
requesting a prefetching time for performing a prefetching operation to prefetch data from a source storage to a target storage, wherein the prefetching time is requested from an awareness engine that includes a prefetching model configured to estimate the prefetching time;
retrieving the data from the source storage to the target storage at the prefetching time, wherein the prefetching time determined by the awareness engine accounts for an energy considerations associated with the source storage when prefetching the data from the source storage to the target storage, wherein the energy considerations include a type of energy and an energy cost; and
storing the prefetched data at the target storage at the prefetching time, wherein the target storage is local to a client.
2 . The method of claim 1 , wherein the prefetching time comprises a prefetching window and an estimated time to perform the prefetching operation.
3 . The method of claim 1 , further comprising generating the prefetching time by inputting features including weather features, and features of the source storage and features of the target storage into a trained prefetching module configured to predict the prefetching time.
4 . The method of claim 1 , wherein the prefetching time comprises a prefetching time window, further comprising waiting for the prefetching time window to arrive prior to performing the prefetching operation, wherein the prefetching time window identifies a window of time for initiating the prefetching operation.
5 . The method of claim 1 , wherein the prefetching time is generated by at least one machine learning model trained on data associated with historical prefetching operations, historical energy costs, and factors influencing the energy costs.
6 . The method of claim 1 , further comprising accounting for a source of energy or an estimated cost of the energy when performing the prefetching operation.
7 . The method of claim 1 , wherein the prefetching time is associated with lower energy costs.
8 . The method of claim 1 , further comprising generating a recommended source of the data using a model configured to recommend a source for the prefetching operation, wherein the source storage was recommended by the model and accounts for at least energy cost.
9 . The method of claim 1 , further comprising receiving a request at the target storage from the client for a data block, determining whether the data block has been retrieved, and downloading the data block if not present at the target storage.
10 . The method of claim 1 , further comprising reading a queue of data that is anticipated to be requested by the client, determining the prefetching time for the data in the queue, and prefetching the data based on a recommendation from the awareness engine.
11 . A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations comprising:
requesting a prefetching time for performing a prefetching operation to prefetch data from a source storage to a target storage, wherein the prefetching time is requested from an awareness engine that includes a prefetching model configured to estimate the prefetching time;
retrieving the data from the source storage to the target storage at the prefetching time, wherein the prefetching time determined by the awareness engine accounts for an energy considerations associated with the source storage when prefetching the data from the source storage to the target storage, wherein the energy considerations include a type of energy and an energy cost; and
storing the prefetched data at the target storage at the prefetching time, wherein the target storage is local to a client.
12 . The non-transitory storage medium of claim 11 , wherein the prefetching time comprises a prefetching window and an estimated time to perform the prefetching operation.
13 . The non-transitory storage medium of claim 11 , further comprising generating the prefetching time by inputting features including weather features, and features of the source storage and features of the target storage into a trained prefetching module configured to predict the prefetching time.
14 . The non-transitory storage medium of claim 11 , wherein the prefetching time comprises a prefetching time window, further comprising waiting for the prefetching time window to arrive prior to performing the prefetching operation, wherein the prefetching time window identifies a window of time for initiating the prefetching operation.
15 . The non-transitory storage medium of claim 11 , wherein the prefetching time is generated by at least one machine learning model trained on data associated with historical prefetching operations, historical energy costs, and factors influencing the energy costs.
16 . The non-transitory storage medium of claim 11 , further comprising accounting for a source of energy or an estimated cost of the energy when performing the prefetching operation.
17 . The non-transitory storage medium of claim 11 , wherein the prefetching time is associated with lower energy costs.
18 . The non-transitory storage medium of claim 11 , further comprising generating a recommended source of the data using a model configured to recommend a source for the prefetching operation, wherein the source storage was recommended by the model and accounts for at least energy cost.
19 . The non-transitory storage medium of claim 11 , further comprising receiving a request at the target storage from the client for a data block, determining whether the data block has been retrieved, and downloading the data block if not present at the target storage.
20 . The non-transitory storage medium of claim 11 , further comprising reading a queue of data that is anticipated to be requested by the client, determining the prefetching time for the data in the queue, and prefetching the data based on a recommendation from the awareness engine.