* xfs speculative preallocation -- fragmentation issue with sparse file handling?
@ 2013-02-18 21:08 Brian Foster
2013-02-18 21:24 ` Mark Tinguely
0 siblings, 1 reply; 4+ messages in thread
From: Brian Foster @ 2013-02-18 21:08 UTC (permalink / raw)
To: xfs; +Cc: Dave Chinner
Hi guys,
I was running a sanity check of my quota throttling stuff rebased
against the updated speculative prealloc algorithm:
a1e16c26 xfs: limit speculative prealloc size on sparse files
... and ran into an interesting behavior on my baseline test (quota
disabled).
The test I'm running is a concurrent write of 32 files (10GB each) via
iozone (I'm not testing performance, just using it as a concurrent writer):
iozone -w -c -e -i 0 -+n -r 4k -s 10g -t 32 -F /mnt/data/file{0..31}
... what I noticed is that from monitoring du during the test,
speculative preallocation seemed to be ineffective. From further
tracing, I observed that imap[0].br_blockcount in
xfs_iomap_eof_prealloc_initial_size() was fairly consistently maxed out
at around 32768 blocks (128MB).
Without the aforementioned commit, preallocation occurs as expected and
the files result in 7-9 extents after the test. With the commit, I'm in
the 70s to 80s range of number of extents with a max extent size of
128MB. A couple examples of xfs_bmap output are appended to this
message. It seems like initial fragmentation might be throwing the
algorithm out of whack..?
Brian
--- Baseline
/mnt/data/file0:
EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET
TOTAL
0: [0..65535]: 327912..393447 0 (327912..393447)
65536
1: [65536..262143]: 4980968..5177575 0 (4980968..5177575)
196608
2: [262144..524287]: 11272424..11534567 0
(11272424..11534567) 262144
3: [524288..1048575]: 25690344..26214631 0
(25690344..26214631) 524288
4: [1048576..2097151]: 58720488..59769063 0
(58720488..59769063) 1048576
5: [2097152..4194303]: 128975080..131072231 0
(128975080..131072231) 2097152
6: [4194304..8388607]: 227541352..231735655 0
(227541352..231735655) 4194304
7: [8388608..16777215]: 504365536..512754143 0
(504365536..512754143) 8388608
8: [16777216..20870671]: 588251584..592345039 0
(588251584..592345039) 4093456
9: [20870672..20971519]: 2684353896..2684454743 1
(536870256..536971103) 100848
--- Sparse file prealloc algorithm
/mnt/data/file0:
EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET
TOTAL
0: [0..65535]: 491752..557287 0 (491752..557287)
65536
1: [65536..196607]: 5734632..5865703 0 (5734632..5865703)
131072
2: [196608..327679]: 10846440..10977511 0 (10846440..10977511)
131072
3: [327680..458751]: 15433960..15565031 0 (15433960..15565031)
131072
4: [458752..720895]: 21463272..21725415 0 (21463272..21725415)
262144
5: [720896..983039]: 25919720..26181863 0 (25919720..26181863)
262144
6: [983040..1245183]: 35225888..35488031 0 (35225888..35488031)
262144
7: [1245184..1507327]: 44401016..44663159 0 (44401016..44663159)
262144
8: [1507328..1769471]: 53314008..53576151 0 (53314008..53576151)
262144
9: [1769472..2031615]: 62489064..62751207 0 (62489064..62751207)
262144
10: [2031616..2293759]: 72450536..72712679 0 (72450536..72712679)
262144
11: [2293760..2555903]: 76120552..76382695 0 (76120552..76382695)
262144
12: [2555904..2818047]: 84771304..85033447 0 (84771304..85033447)
262144
13: [2818048..3080191]: 93422056..93684199 0 (93422056..93684199)
262144
14: [3080192..3342335]: 102597096..102859239 0
(102597096..102859239) 262144
15: [3342336..3604479]: 110723560..110985703 0
(110723560..110985703) 262144
16: [3604480..3866623]: 119374312..119636455 0
(119374312..119636455) 262144
17: [3866624..4128767]: 128811496..129073639 0
(128811496..129073639) 262144
18: [4128768..4390911]: 133530088..133792231 0
(133530088..133792231) 262144
19: [4390912..4653055]: 142180840..142442983 0
(142180840..142442983) 262144
20: [4653056..4915199]: 151355880..151618023 0
(151355880..151618023) 262144
21: [4915200..5177343]: 159482344..159744487 0
(159482344..159744487) 262144
22: [5177344..5439487]: 167608808..167870951 0
(167608808..167870951) 262144
23: [5439488..5701631]: 176783848..177045991 0
(176783848..177045991) 262144
24: [5701632..5963775]: 186483176..186745319 0
(186483176..186745319) 262144
25: [5963776..6225919]: 195133928..195396071 0
(195133928..195396071) 262144
26: [6225920..6488063]: 199066088..199328231 0
(199066088..199328231) 262144
27: [6488064..6750207]: 207978984..208241127 0
(207978984..208241127) 262144
28: [6750208..7012351]: 216367592..216629735 0
(216367592..216629735) 262144
29: [7012352..7274495]: 224494056..224756199 0
(224494056..224756199) 262144
30: [7274496..7536639]: 232882664..233144807 0
(232882664..233144807) 262144
31: [7536640..7798783]: 241533416..241795559 0
(241533416..241795559) 262144
32: [7798784..8060927]: 251494888..251757031 0
(251494888..251757031) 262144
33: [8060928..8323071]: 255689192..255951335 0
(255689192..255951335) 262144
34: [8323072..8585215]: 264077800..264339943 0
(264077800..264339943) 262144
35: [8585216..8847359]: 272990696..273252839 0
(272990696..273252839) 262144
36: [8847360..9109503]: 281379304..281641447 0
(281379304..281641447) 262144
37: [9109504..9371647]: 290554344..290816487 0
(290554344..290816487) 262144
38: [9371648..9633791]: 300253672..300515815 0
(300253672..300515815) 262144
39: [9633792..9895935]: 304447976..304710119 0
(304447976..304710119) 262144
40: [9895936..10158079]: 313098728..313360871 0
(313098728..313360871) 262144
41: [10158080..10420223]: 322011624..322273767 0
(322011624..322273767) 262144
42: [10420224..10682367]: 330138088..330400231 0
(330138088..330400231) 262144
43: [10682368..10944511]: 339575272..339837415 0
(339575272..339837415) 262144
44: [10944512..11206655]: 349536744..349798887 0
(349536744..349798887) 262144
45: [11206656..11468799]: 353468904..353731047 0
(353468904..353731047) 262144
46: [11468800..11730943]: 362381800..362643943 0
(362381800..362643943) 262144
47: [11730944..11993087]: 370770408..371032551 0
(370770408..371032551) 262144
48: [11993088..12255231]: 379159016..379421159 0
(379159016..379421159) 262144
49: [12255232..12517375]: 388071912..388334055 0
(388071912..388334055) 262144
50: [12517376..12779519]: 397771240..398033383 0
(397771240..398033383) 262144
51: [12779520..13041663]: 401965544..402227687 0
(401965544..402227687) 262144
52: [13041664..13303807]: 410878440..411140583 0
(410878440..411140583) 262144
53: [13303808..13565951]: 419529192..419791335 0
(419529192..419791335) 262144
54: [13565952..13828095]: 427655656..427917799 0
(427655656..427917799) 262144
55: [13828096..14090239]: 436568552..436830695 0
(436568552..436830695) 262144
56: [14090240..14352383]: 446792168..447054311 0
(446792168..447054311) 262144
57: [14352384..14614527]: 450986472..451248615 0
(450986472..451248615) 262144
58: [14614528..14876671]: 459637224..459899367 0
(459637224..459899367) 262144
59: [14876672..15138815]: 468287976..468550119 0
(468287976..468550119) 262144
60: [15138816..15400959]: 476414440..476676583 0
(476414440..476676583) 262144
61: [15400960..15663103]: 485327336..485589479 0
(485327336..485589479) 262144
62: [15663104..15925247]: 495288808..495550951 0
(495288808..495550951) 262144
63: [15925248..16187391]: 499483112..499745255 0
(499483112..499745255) 262144
64: [16187392..16449535]: 507871720..508133863 0
(507871720..508133863) 262144
65: [16449536..16711679]: 517046760..517308903 0
(517046760..517308903) 262144
66: [16711680..16973823]: 525173224..525435367 0
(525173224..525435367) 262144
67: [16973824..17235967]: 533823976..534086119 0
(533823976..534086119) 262144
68: [17235968..17498111]: 544309736..544571879 0
(544309736..544571879) 262144
69: [17498112..17760255]: 548504040..548766183 0
(548504040..548766183) 262144
70: [17760256..18022399]: 556892648..557154791 0
(556892648..557154791) 262144
71: [18022400..18284543]: 565805544..566067687 0
(565805544..566067687) 262144
72: [18284544..18546687]: 573932008..574194151 0
(573932008..574194151) 262144
73: [18546688..18808831]: 582058472..582320615 0
(582058472..582320615) 262144
74: [18808832..19070975]: 591495656..591757799 0
(591495656..591757799) 262144
75: [19070976..19333119]: 596214248..596476391 0
(596214248..596476391) 262144
76: [19333120..19595263]: 604865000..605127143 0
(604865000..605127143) 262144
77: [19595264..19857407]: 613253608..613515751 0
(613253608..613515751) 262144
78: [19857408..20119551]: 621380072..621642215 0
(621380072..621642215) 262144
79: [20119552..20381695]: 630555112..630817255 0
(630555112..630817255) 262144
80: [20381696..20643839]: 635273704..635535847 0
(635273704..635535847) 262144
81: [20643840..20867199]: 643662312..643885671 0
(643662312..643885671) 223360
82: [20867200..20971519]: 653360704..653465023 0
(653360704..653465023) 104320
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: xfs speculative preallocation -- fragmentation issue with sparse file handling?
2013-02-18 21:08 xfs speculative preallocation -- fragmentation issue with sparse file handling? Brian Foster
@ 2013-02-18 21:24 ` Mark Tinguely
2013-02-18 23:29 ` Brian Foster
0 siblings, 1 reply; 4+ messages in thread
From: Mark Tinguely @ 2013-02-18 21:24 UTC (permalink / raw)
To: Brian Foster; +Cc: Dave Chinner, xfs
On 02/18/13 15:08, Brian Foster wrote:
> Hi guys,
>
> I was running a sanity check of my quota throttling stuff rebased
> against the updated speculative prealloc algorithm:
>
> a1e16c26 xfs: limit speculative prealloc size on sparse files
>
> ... and ran into an interesting behavior on my baseline test (quota
> disabled).
>
> The test I'm running is a concurrent write of 32 files (10GB each) via
> iozone (I'm not testing performance, just using it as a concurrent writer):
>
> iozone -w -c -e -i 0 -+n -r 4k -s 10g -t 32 -F /mnt/data/file{0..31}
>
> ... what I noticed is that from monitoring du during the test,
> speculative preallocation seemed to be ineffective. From further
> tracing, I observed that imap[0].br_blockcount in
> xfs_iomap_eof_prealloc_initial_size() was fairly consistently maxed out
> at around 32768 blocks (128MB).
>
> Without the aforementioned commit, preallocation occurs as expected and
> the files result in 7-9 extents after the test. With the commit, I'm in
> the 70s to 80s range of number of extents with a max extent size of
> 128MB. A couple examples of xfs_bmap output are appended to this
> message. It seems like initial fragmentation might be throwing the
> algorithm out of whack..?
>
> Brian
... the patched version increases in doubles
+ if (imap[0].br_startblock == HOLESTARTBLOCK)
+ return 0;
vvvvvv
+ if (imap[0].br_blockcount <= (MAXEXTLEN >> 1))
+ return imap[0].br_blockcount;
^^^^^^
+ return XFS_B_TO_FSB(mp, XFS_ISIZE(ip));
+}
have you experimented without the middle if statement.
If I remember correctly when I reviewed the code, that should be moving
code closer to the original code; namely use the file size as the
preallocation value.
--Mark.
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: xfs speculative preallocation -- fragmentation issue with sparse file handling?
2013-02-18 21:24 ` Mark Tinguely
@ 2013-02-18 23:29 ` Brian Foster
2013-02-18 23:39 ` Dave Chinner
0 siblings, 1 reply; 4+ messages in thread
From: Brian Foster @ 2013-02-18 23:29 UTC (permalink / raw)
To: Mark Tinguely; +Cc: Dave Chinner, xfs
On 02/18/2013 04:24 PM, Mark Tinguely wrote:
> On 02/18/13 15:08, Brian Foster wrote:
>> Hi guys,
>>
>> I was running a sanity check of my quota throttling stuff rebased
>> against the updated speculative prealloc algorithm:
>>
>> a1e16c26 xfs: limit speculative prealloc size on sparse files
>>
>> ... and ran into an interesting behavior on my baseline test (quota
>> disabled).
>>
>> The test I'm running is a concurrent write of 32 files (10GB each) via
>> iozone (I'm not testing performance, just using it as a concurrent
>> writer):
>>
>> iozone -w -c -e -i 0 -+n -r 4k -s 10g -t 32 -F /mnt/data/file{0..31}
>>
>> ... what I noticed is that from monitoring du during the test,
>> speculative preallocation seemed to be ineffective. From further
>> tracing, I observed that imap[0].br_blockcount in
>> xfs_iomap_eof_prealloc_initial_size() was fairly consistently maxed out
>> at around 32768 blocks (128MB).
>>
>> Without the aforementioned commit, preallocation occurs as expected and
>> the files result in 7-9 extents after the test. With the commit, I'm in
>> the 70s to 80s range of number of extents with a max extent size of
>> 128MB. A couple examples of xfs_bmap output are appended to this
>> message. It seems like initial fragmentation might be throwing the
>> algorithm out of whack..?
>>
>> Brian
>
> ... the patched version increases in doubles
>
> + if (imap[0].br_startblock == HOLESTARTBLOCK)
> + return 0;
>
> vvvvvv
> + if (imap[0].br_blockcount <= (MAXEXTLEN >> 1))
> + return imap[0].br_blockcount;
> ^^^^^^
>
> + return XFS_B_TO_FSB(mp, XFS_ISIZE(ip));
> +}
>
> have you experimented without the middle if statement.
> If I remember correctly when I reviewed the code, that should be moving
> code closer to the original code; namely use the file size as the
> preallocation value.
>
Just a quick update...
I've tested the change above and a suggestion Dave made on IRC to return
(imap[0].br_blockcount << 1) and both resolve the immediate issue. I
need to verify the original test case still works and I'll post a patch.
Thanks...
Brian
> --Mark.
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: xfs speculative preallocation -- fragmentation issue with sparse file handling?
2013-02-18 23:29 ` Brian Foster
@ 2013-02-18 23:39 ` Dave Chinner
0 siblings, 0 replies; 4+ messages in thread
From: Dave Chinner @ 2013-02-18 23:39 UTC (permalink / raw)
To: Brian Foster; +Cc: xfs, Mark Tinguely, Dave Chinner
On Mon, Feb 18, 2013 at 06:29:02PM -0500, Brian Foster wrote:
> On 02/18/2013 04:24 PM, Mark Tinguely wrote:
> > On 02/18/13 15:08, Brian Foster wrote:
> > ... the patched version increases in doubles
> >
> > + if (imap[0].br_startblock == HOLESTARTBLOCK)
> > + return 0;
> >
> > vvvvvv
> > + if (imap[0].br_blockcount <= (MAXEXTLEN >> 1))
> > + return imap[0].br_blockcount;
> > ^^^^^^
> >
> > + return XFS_B_TO_FSB(mp, XFS_ISIZE(ip));
> > +}
> >
> > have you experimented without the middle if statement.
> > If I remember correctly when I reviewed the code, that should be moving
> > code closer to the original code; namely use the file size as the
> > preallocation value.
> >
>
> Just a quick update...
>
> I've tested the change above and a suggestion Dave made on IRC to return
> (imap[0].br_blockcount << 1) and both resolve the immediate issue. I
> need to verify the original test case still works and I'll post a patch.
> Thanks...
Though I think you'll find Mark's suggestion reverts preallocation
on sparse files back to the old, undesirable behaviour as it will
prealloc based on the file size on the second 4k of the write at
EOF....
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2013-02-18 23:39 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-02-18 21:08 xfs speculative preallocation -- fragmentation issue with sparse file handling? Brian Foster
2013-02-18 21:24 ` Mark Tinguely
2013-02-18 23:29 ` Brian Foster
2013-02-18 23:39 ` Dave Chinner
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox