Consider changing repo layout to work with "large" (perhaps 16MB) files kind of like hadoop.
Consider changing repo layout to work with "large" perhaps 16MB files kind of like hadoop. Using small files like now (avg size ~32k) makes repo deletes take forever and makes copying the repo take forever. For our repo, we could pack as many files as will fit wholly fit into ~16MB. Larger files would be saved as individual files. The whole point is to consolidate the tiny files to use fewer inodes and thus speed up delagent and repo copies.