Virtual storage mechanism to optimize fetch performance of named objects in enterprise systems
Embodiments relate to enhancing a virtual storage mechanism to optimize fetch performance of named objects in enterprise systems. A technique includes receiving from a requester a request to retrieve an object and determining that the object is available for retrieval in a virtual data space, where a machine learning model is configured to cause the object to be placed in the virtual data space, the machine learning model being trained on fetch data for objects. The technique includes retrieving the object from the virtual data space instead of a storage system and providing the object to the requester for execution.
1 . A computer-implemented method comprising:
predicting by a machine learning model a location for placing an object, wherein the location predicted is a virtual data space of a local memory, the virtual data space comprising tiers, wherein the tiers comprise a fixed virtual lookaside facility (VLF) having a highest weightage for storing first objects assigned by the machine learning model, a pageable VLF having a lower weightage than the highest weightage for storing second objects, and a temporary VLF for storing third objects incoming to the machine learning model prior to a retraining of the machine learning model, wherein the retraining of the machine learning model is based at least in part on the third objects received prior to the retraining in order to swap one or more objects within the tiers of the virtual data space;
receiving from a requester a request to retrieve the object;
determining that the object is available for retrieval in the virtual data space of the local memory, wherein the machine learning model is configured to cause the object to be placed in the virtual data space, the machine learning model being trained on fetch data for objects;
retrieving the object from the virtual data space of the local memory instead of a storage system that is remote from the local memory, wherein the machine learning model is retrained based on a trigger; and
providing the object to the requester for execution from the local memory instead of the storage system, in response to the machine learning model predicting the location for the object as the virtual data space of the local memory.
2 . The computer-implemented method of claim 1 ,
wherein the first objects in the fixed VLF are searchable and retrievable faster than the second objects in the pageable VLF; and
wherein the second objects in the pageable VLF are searchable and retrievable faster than the third objects in the temporary VLF.
3 . The computer-implemented method of claim 1 , wherein the object is retrieved from one of the tiers in the virtual data space; and
wherein the machine learning model is trained to identify one of a first, second, and third tiers in which to place new objects.
4 . The computer-implemented method of claim 1 , wherein the trigger comprises a setting in a graphical user interface having selectable buttons for controlling fetch performance tuning, wherein training data comprises historical data including a type of the objects, names of the objects, a fetch time of the objects, an execution time of the objects, and a number of fetches of the objects, wherein training the machine learning model comprises associating labels with the historical data in order to modify weights in the machine learning model such that a placement of one or more of the objects is modified within the tiers of the virtual data space;
wherein the graphical user interface comprises options for changing weight factors applied to the objects in the tiers of the virtual data space; and
wherein the fetch performance tuning provides the options for increasing or decreasing weights to modify staging and swapping of the objects for the placement in the tiers of the virtual data space.
5 . The computer-implemented method of claim 1 , wherein the trigger further comprises at least one of a predefined time period or a predefined change in the fetch data.
6 . The computer-implemented method of claim 1 , wherein a library lookaside is automatically updated based on any new object being added to the virtual data space, changed from one tier to another tier in the virtual data space, and removed from the virtual data space.
7 . The computer-implemented method of claim 1 , wherein the storage system comprises auxiliary storage that requires input and output requests to access data; and
wherein the machine learning model is trained to determine to store a given object in the fixed VLF for a first period of a day and swap the given object to the pageable VLF for a second period of time during the day.
8 . A system comprising:
a memory having computer readable instructions; and
one or more processors for executing the computer readable instructions, the computer readable instructions controlling the one or more processors to perform operations comprising:
predicting by a machine learning model a location for placing an object, wherein the location predicted is a virtual data space of a local memory, the virtual data space comprising tiers, wherein the tiers comprise a fixed virtual lookaside facility (VLF) having a highest weightage for storing first objects assigned by the machine learning model, a pageable VLF having a lower weightage than the highest weightage for storing second objects, and a temporary VLF for storing third objects incoming to the machine learning model prior to a retraining of the machine learning model, wherein the retraining of the machine learning model is based at least in part on the third objects received prior to the retraining in order to swap one or more objects within the tiers of the virtual data space;
receiving from a requester a request to retrieve the object;
determining that the object is available for retrieval in the virtual data space of the local memory, wherein the machine learning model is configured to cause the object to be placed in the virtual data space, the machine learning model being trained on fetch data for objects;
retrieving the object from the virtual data space of the local memory instead of a storage system that is remote from the local memory, wherein the machine learning model is retrained based on a trigger; and
providing the object to the requester for execution from the local memory instead of the storage system, in response to the machine learning model predicting the location for the object as the virtual data space of the local memory.
9 . The system of claim 8 ,
wherein the first objects in the fixed VLF are searchable and retrievable faster than the second objects in the pageable VLF; and
wherein the second objects in the pageable VLF are searchable and retrievable faster than the third objects in the temporary VLF.
10 . The system of claim 8 , wherein the object is retrieved from one of the tiers in the virtual data space; and
wherein the machine learning model is trained to identify one of a first, second, and third tiers in which to place new objects.
11 . The system of claim 8 , wherein the machine learning model is trained using the fetch data that has been updated based on recent fetches for the objects.
12 . The system of claim 11 , wherein the trigger further comprises at least one of a predefined time period or a predefined change in the fetch data.
13 . The system of claim 8 , wherein a library lookaside is automatically updated based on any new object being added to the virtual data space, changed from one tier to another tier in the virtual data space, and removed from the virtual data space.
14 . The system of claim 8 , wherein the storage system comprises auxiliary storage that requires input and output requests to access data.
15 . A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by one or more processors to cause the one or more processors to perform operations comprising:
predicting by a machine learning model a location for placing an object, wherein the location predicted is a virtual data space of a local memory, the virtual data space comprising tiers, wherein the tiers comprise a fixed virtual lookaside facility (VLF) having a highest weightage for storing first objects assigned by the machine learning model, a pageable VLF having a lower weightage than the highest weightage for storing second objects, and a temporary VLF for storing third objects incoming to the machine learning model prior to a retraining of the machine learning model, wherein the retraining of the machine learning model is based at least in part on the third objects received prior to the retraining in order to swap one or more objects within the tiers of the virtual data space;
receiving from a requester a request to retrieve the object;
determining that the object is available for retrieval in the virtual data space of the local memory, wherein the machine learning model is configured to cause the object to be placed in the virtual data space, the machine learning model being trained on fetch data for objects;
retrieving the object from the virtual data space of the local memory instead of a storage system that is remote from the local memory, wherein the machine learning model is retrained based on a trigger; and
providing the object to the requester for execution from the local memory instead of the storage system, in response to the machine learning model predicting the location for the object as the virtual data space of the local memory.
16 . The computer program product of claim 15 ,
wherein the first objects in the fixed VLF are searchable and retrievable faster than the second objects in the pageable VLF; and
wherein the second objects in the pageable VLF are searchable and retrievable faster than the third objects in the temporary VLF.
17 . The computer program product of claim 15 , wherein the object is retrieved from one of the tiers in the virtual data space; and
wherein the machine learning model is trained to identify one of a first, second, and third tiers in which to place new objects.
18 . The computer program product of claim 15 , wherein the machine learning model is trained using the fetch data that has been updated based on recent fetches for the objects.
19 . The computer program product of claim 18 , wherein the trigger comprises at least one of a predefined time period or a predefined change in the fetch data.
20 . The computer program product of claim 15 , wherein a library lookaside is automatically updated based on any new object being added to the virtual data space, changed from one tier to another tier in the virtual data space, and removed from the virtual data space.