During block reports we resolve the file path in reverse using the inode IDs. Depending on the path length his uses multiple FTIS operations.
The performance can be improved in two ways
1) Create a Unique Index on the inode ID. This will avoid frag scans on the database side.
2) Use the NN cache to resolve the file path.
The current solution option 1 is selected as it delivers satisfactory performance and the load on the DB is low
Creating a unique index does not work because then ndb upgrade every read-committed lock to read.