||This article provides insufficient context for those unfamiliar with the subject. (October 2009)|
AutoFetch is a mechanism for automatically tuning object-relational mapping queries.
Improving the modularity of application code
AutoFetch uses the program state (typically the Call Stack) to classify queries. This means that AutoFetch can tune the same query depending on HOW it was called.
For example, a Data Access API such as
findCustomer(int customerId) can be tuned differently depending on the callers of the method.
Reducing the development and maintenance burden from Developers
Because AutoFetch automatically gathers the profiling information developers do not need to manually try and gather this information and apply performance optimizations to the code.
AutoFetch can monitor the object graph usage so that if the usage changes the tuning of the query can change to suit. This reduces the code maintenance issues for developers and maintains optimal performance as usage changes over time.
How it works
AutoFetch collects object graph usage when queries are executed. It collects the 'profile' information which is later used to automatically tune the query on subsequent executions. The ORM query is tuned by determining and automatically adding the correct prefetch directives for each query.
This improves the performance of the application by reducing "lazy loading". For ORMs that support partial objects AutoFetch can also tune the query by just including the properties that the application uses rather than all properties.
AutoFetch profiling can also be used to select optimal query strategies such as "array fetching" where object graph use is known to be uneven (typically weighted to the first object graphs in a list).
This could lead to much more efficient and higher performing object graph traversal than is currently possible with current techniques employed by ORM tools.