| To: | linux-xfs@xxxxxxxxxxx |
|---|---|
| Subject: | guaranteed increasing extent allocation |
| From: | "Rick Smith" <rgsmith72@xxxxxxxxxxx> |
| Date: | Fri, 20 Jun 2003 10:50:26 -0700 |
| Sender: | linux-xfs-bounce@xxxxxxxxxxx |
|
Hello, Is there a way to guarantee that if you reserve space for a series of files at once using the XFS_RESVSP64 ioctl (in say, a for-loop before I/O takes place) that the extents for each respective allocation will be greater than the previous file's extents for the current reserve set? In other words, we are looking to maintain high performance for files read/written in sequence by laying them out in the filesystem in order to minimize disk head seeks. With a fresh XFS filesystem, we seem to get consecutive extent allocation, but as the filesystem ages and files are created and deleted, it seems less and less likely to be able to reserve consecutive files. Can this be done currently or would a new ioctl be necessary to reserve files relative to each other be the ticket? I have included a couple examples of what we are seeing below. Any help appreciated. Rick Smith The following shows 4 files reserved with XFS_RESVSP64 in a for-loop with each consecutive file reserved in-front (with a lower starting extent) than the previous one (xfs_bmap-v): test.20.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1417438960..1417449759 168 (8154160..8164959) 10800 test.21.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1417428160..1417438959 168 (8143360..8154159) 10800 test.22.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1417417360..1417428159 168 (8132560..8143359) 10800 test.23.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1417406560..1417417359 168 (8121760..8132559) 10800 The following shows an alternating reserve location for files reserved in order: test.6.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1405059864..1405070663 167 (4163664..4174463) 10800 test.7.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1404984224..1404995023 167 (4088024..4098823) 10800 test.8.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1405070664..1405081463 167 (4174464..4185263) 10800 test.9.data: EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL 0: [0..10799]: 1404973424..1404984223 167 (4077224..4088023) 10800 _________________________________________________________________ Tired of spam? Get advanced junk mail protection with MSN 8. http://join.msn.com/?page=features/junkmail |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: XFS, md (raid5) and "switching cache buffer size", Steve Lord |
|---|---|
| Next by Date: | TAKE - fix initializer for sync interval, Steve Lord |
| Previous by Thread: | XFS, md (raid5) and "switching cache buffer size", Gordon Henderson |
| Next by Thread: | TAKE - fix initializer for sync interval, Steve Lord |
| Indexes: | [Date] [Thread] [Top] [All Lists] |