NSAmelchev opened a new pull request, #13710: URL: https://github.com/apache/ignite/pull/13710
Before: a local partition of a compressed group was copied densely with `transferTo`, so the holes of the sparse snapshot file became zeros on disk. Then the copy was opened as a `FilePageStore`, every page was read again and a hole was punched after the compressed bytes of each page with a separate `fallocate` call. Now: the partition is copied page by page. The compressed size is taken from the page header and only these bytes are written at the page offset, so the file system leaves the rest of the page unallocated and the copy is sparse without the second pass. Uncompressed pages are written whole. The snapshot is opened read-only, as for uncompressed groups. Partitions received from remote nodes are written by the file receiver and are punched as before. Disk Page Compressions 4 (TEST_SCALE_FACTOR=1.0, 4-CPU agents): 37:42 on master vs 28:44. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
