Architecture - Free Space Management - Revision history

Docadmin at 21:11, 22 January 2010

2010-01-22T21:11:09Z

← Older revision		Revision as of 14:11, 22 January 2010
Line 1:		Line 1:
	'''''Note:''''' The content on this page reflects the state of design of a Lustre feature at a particular point in time and may contain outdated information.		'''''Note:''''' ''The content on this page reflects the state of design of a Lustre feature at a particular point in time and may contain outdated information.''

	== Free Space Management ==		== Free Space Management ==

Docadmin: Protected "Architecture - Free Space Management" ([edit=sysop] (indefinite) [move=sysop] (indefinite))

2010-01-21T19:45:27Z

Protected "Architecture - Free Space Management" ([edit=sysop] (indefinite) [move=sysop] (indefinite))

← Older revision	Revision as of 12:45, 21 January 2010
(No difference)

Docadmin: /* Free Space Management */

2010-01-18T23:58:05Z

Free Space Management

← Older revision		Revision as of 16:58, 18 January 2010
Line 1:		Line 1:
			'''''Note:''''' The content on this page reflects the state of design of a Lustre feature at a particular point in time and may contain outdated information.

	== Free Space Management ==		== Free Space Management ==

Docadmin: /* References */

2010-01-14T22:26:02Z

References

← Older revision		Revision as of 15:26, 14 January 2010
Line 49:		Line 49:
	via a weighting -- if OST2 has twice as much free space as OST1, it will		via a weighting -- if OST2 has twice as much free space as OST1, it will
	be twice as ''likely'' to be used, but is still not ''guaranteed'' to be used.		be twice as ''likely'' to be used, but is still not ''guaranteed'' to be used.

	~~== References ==~~

	~~[[Category:Architecture]]~~

Docadmin: moved Architecture Free Space Management to Architecture - Free Space Management

2010-01-14T21:47:05Z

moved Architecture Free Space Management to Architecture - Free Space Management

← Older revision	Revision as of 14:47, 14 January 2010
(No difference)

Docadmin: 1 revision

2010-01-14T21:01:08Z

1 revision

← Older revision	Revision as of 14:01, 14 January 2010
(No difference)

Bsneitzel: /* References */

2007-12-11T16:23:14Z

References

New page

== Free Space Management ==

In Lustre 1.6, the MDS assigns file stripe objects to OSTs using a mix of load balancing and free space considerations in order to optimize filesystem performance and OSS space utilizations. Emptier OSTs are prefered for stripes, and stripes are distributed evenly over OSSs to increase network bandwidth utilization. The weighting factor between these two optimizations is user-adjustable.

There are two stripe allocation methods. The method is determined by the amount of free-space imbalance on the OSTs. The '''weighted allocator''' is used when any two OSTs are imbalanced by more than 20%. Until then, a faster '''round-robin allocater''' is used. The round-robin order maximizes network balancing.

== Round-Robin Allocator ==

When OSTs have approximately the same free space (within 20%), an efficient round-robin allocator is used. The round-robin allocator alternates stripes between OSTs on different
OSSs. Here are some example round-robin stripe orders (the same letter represents the different OSTs on a single OSS):

{| border=1 cellspacing=1
|-
|3: AAA ||a single 3-OST OSS
|-
|3+3: ABABAB ||2 3-OST OSS's
|-
| 3+4: BBABABA ||a 3-OST OSS (A) and and 4-OST OSS (B)
|-
| 3+5: BBABBABA
|-
| 3+5+1: BBABABABC
|-
| 3+5+2: BABABCBABC
|-
| 4+6+2: BABABCBABABC
|}

=== Weighted Allocator ===

When the OSTs free space difference is significant, then a weighting
algorithm is used to influence OST ordering based on size and location.
Note that these are weightings for a ''random'' algorithm and so will not necessarily strictly choose the "emptiest" OST every time. On ''average'', it will fill the emptier OSTs
faster.

==== Adjusting the weighting between free space and location ====
This priority can be adjusted via the proc file ''/proc/fs/lustre/lov/lustre-mdtlov/qos_prio_free''. The default in the future will be 90%.
You can set this permanently on existing betas with this command on the MGS:
{| border=1 cellspacing=1
|-
|lctl conf_param <fsname>-MDT0000.lov.qos_prio_free=90
|}
Increasing the value will put more weighting on free space. When set to 100% free space priority, then
location is no longer used in the stripe ordering calculations, and the weighting is based entirely on free space.

Note that setting the priority to 100% just means that OSS distribution
doesn't count in the weighting, but the stripe assignment is still done
via a weighting -- if OST2 has twice as much free space as OST1, it will
be twice as ''likely'' to be used, but is still not ''guaranteed'' to be used.

== References ==

[[Category:Architecture]]