You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
Since Gaffer 1 there have been various changes to how the cache works. Most importantly on this topic, the removal of the internal cache of the "Cache" (#2457). However, this was done without any performance tests, and I assert the cache has become unacceptably performant.
Expected behaviour
The cache should perform a lot closer to how it did in Gaffer 1.23.0. I would suggest performance tests to ensure this behaviour is achieved and maintained.
Additional context
As mentioned above, the internal cache was removed. Not only should this have been tested, but the "Cache" code should have been refactored in accordance with the new behaviour. Calls to the cache which were once cheap and done many times in an unoptimised way are now no longer acceptable and create massive performance issues. I think the "Cache", and many things that use it like NamedOperations and the FederatedStore, need to be refactored with performance in mind.
One example of this performance issue is if you start a FederatedStore with a graph in it that also has NamedOperations in the cache. It now takes 3 times as long to start and spams tens of log lines mentioning the NamedOperationResolver.
The text was updated successfully, but these errors were encountered:
Describe the bug
Since Gaffer 1 there have been various changes to how the cache works. Most importantly on this topic, the removal of the internal cache of the "Cache" (#2457). However, this was done without any performance tests, and I assert the cache has become unacceptably performant.
Expected behaviour
The cache should perform a lot closer to how it did in Gaffer 1.23.0. I would suggest performance tests to ensure this behaviour is achieved and maintained.
Additional context
As mentioned above, the internal cache was removed. Not only should this have been tested, but the "Cache" code should have been refactored in accordance with the new behaviour. Calls to the cache which were once cheap and done many times in an unoptimised way are now no longer acceptable and create massive performance issues. I think the "Cache", and many things that use it like NamedOperations and the FederatedStore, need to be refactored with performance in mind.
One example of this performance issue is if you start a FederatedStore with a graph in it that also has NamedOperations in the cache. It now takes 3 times as long to start and spams tens of log lines mentioning the NamedOperationResolver.
The text was updated successfully, but these errors were encountered: