I/O priorities - Revision history

90.165.252.186: Corrected a mispelling error, but still there is an important question unanswered, what is HZ?

2016-04-24T14:02:03Z

Corrected a mispelling error, but still there is an important question unanswered, what is HZ?

Kir: fix kernel version

2013-12-13T11:13:59Z

fix kernel version

Kir: improved

2013-12-13T11:13:07Z

improved

Kir: add link to i/o limits; added into howto cat

2013-12-13T11:12:01Z

add link to i/o limits; added into howto cat

Kir: Kir moved page I/O priorities for containers to I/O priorities: simplify

2013-12-13T11:10:48Z

Kir moved page I/O priorities for containers to I/O priorities: simplify

66.45.235.121 at 20:35, 7 December 2012

2012-12-07T20:35:59Z

98.222.188.208 at 20:23, 12 May 2010

2010-05-12T20:23:40Z

Kir: VE->CT

2009-02-12T08:29:14Z

VE->CT

Kir: I/O priorities for VE moved to I/O priorities for containers: VE->CT

2009-02-12T08:27:16Z

I/O priorities for VE moved to I/O priorities for containers: VE->CT

Kir: added see also

2009-02-12T08:26:36Z

added see also

@@ Line 13: / Line 13: @@
 The higher the value you use, the more I/O time your container will receive.
-The mapping from priority to time is the following: if <tt>0</tt> prio corresponds to time slice <tt>t</tt>, than <tt>8</tt> prio corresponds to time slice <tt>2 * t</tt>. Default time slice value is <tt>HZ/2</tt>. The main criteria for fairness at the moment is that if you set I/O prio of CT 1 to <tt>p1</tt> and I/O prio of CT 2 to <tt>p2</tt>, and <tt>p1 > p2</tt> then CT 1 should do more I/O than CT 2. In simple form the priorities are relative with the higher priority getting more I/O time.
+The mapping from priority to time is the following: if <tt>0</tt> prio corresponds to time slice <tt>t</tt>, then <tt>7</tt> prio corresponds to time slice <tt>2 * t</tt>. Default time slice value is <tt>HZ/2</tt>. The main criteria for fairness at the moment is that if you set I/O prio of CT 1 to <tt>p1</tt> and I/O prio of CT 2 to <tt>p2</tt>, and <tt>p1 > p2</tt> then CT 1 should do more I/O than CT 2. In simple form the priorities are relative with the higher priority getting more I/O time.
 == See also ==

← Older revision		Revision as of 11:13, 13 December 2013
Line 1:		Line 1:
−	The I/O priorities feature is implemented in OpenVZ since kernel <tt>2.6.18-~~028stable021~~</tt>, <tt>vzctl 3.0.16</tt>. This feature allows to assign I/O priority to any [[container]]. Priority range is <tt>0-7</tt>. The more priority a container has, the more time for using block devices this container will obtain. This feature is based on CFQ I/O scheduler, so this scheduler should be used for block device in question. Default I/O priority is <tt>4</tt>.	+	The I/O priorities feature is implemented in OpenVZ since kernel <tt>2.6.18-028stab021</tt>, <tt>vzctl 3.0.16</tt>. This feature allows to assign I/O priority to any [[container]]. Priority range is <tt>0-7</tt>. The more priority a container has, the more time for using block devices this container will obtain. This feature is based on CFQ I/O scheduler, so this scheduler should be used for block device in question. Default I/O priority is <tt>4</tt>.

	== Examples ==		== Examples ==

@@ Line 1: / Line 1: @@
-The I/O priorities feature is implemented in OpenVZ since kernel <tt>2.6.18-028stable021</tt>, <tt>vzctl 3.0.16</tt>. This feature allows to assign I/O priority to any [[container]]. Priority range is <tt>0-7</tt>. The more priority a container has, the more time for using block devices this container will obtain. This feature is based on CFQ I/O scheduler, so this scheduler should be used for block device in question. Default I/O priority is <tt>4</tt>. Examples of using:
+The I/O priorities feature is implemented in OpenVZ since kernel <tt>2.6.18-028stable021</tt>, <tt>vzctl 3.0.16</tt>. This feature allows to assign I/O priority to any [[container]]. Priority range is <tt>0-7</tt>. The more priority a container has, the more time for using block devices this container will obtain. This feature is based on CFQ I/O scheduler, so this scheduler should be used for block device in question. Default I/O priority is <tt>4</tt>.
   vzctl set 101 --ioprio 0 --save
   vzctl set 101 --ioprio 7 --save
 The higher the value you use, the more I/O time your container will receive.
 The mapping from priority to time is the following: if <tt>0</tt> prio corresponds to time slice <tt>t</tt>, than <tt>8</tt> prio corresponds to time slice <tt>2 * t</tt>. Default time slice value is <tt>HZ/2</tt>. The main criteria for fairness at the moment is that if you set I/O prio of CT 1 to <tt>p1</tt> and I/O prio of CT 2 to <tt>p2</tt>, and <tt>p1 > p2</tt> then CT 1 should do more I/O than CT 2. In simple form the priorities are relative with the higher priority getting more I/O time.
 == See also ==

@@ Line 12: / Line 12: @@
 == See also ==
+* [[I/O limits]]
 * [[IO accounting]]
 [[Category: Resource management]]

@@ Line 4: / Line 4: @@
   vzctl set 101 --ioprio 7 --save
 The mapping from priority to time is the following: if <tt>0</tt> prio corresponds to time slice <tt>t</tt>, than <tt>8</tt> prio corresponds to time slice <tt>2 * t</tt>. Default time slice value is <tt>HZ/2</tt>. The main criteria for fairness at the moment is that if you set I/O prio of CT 1 to <tt>p1</tt> and I/O prio of CT 2 to <tt>p2</tt>, and <tt>p1 > p2</tt> then CT 1 should do more I/O than CT 2. In simple form the priorities are relative with the higher priority getting more I/O time.
 == See also ==

← Older revision	Revision as of 11:10, 13 December 2013
(No difference)

← Older revision	Revision as of 08:27, 12 February 2009
(No difference)

@@ Line 7: / Line 7: @@
 </pre>
 The mapping from priority to time is the following: if <tt>0</tt> prio corresponds time slice <tt>t</tt>, than <tt>8</tt> prio corresponds to time slice <tt>2 * t</tt>. Default time slice value is <tt>HZ/2</tt>. The main criteria for fairness at the moment is that if you set I/O prio of VE 1 to <tt>p1</tt> and I/O prio of VE 2 to <tt>p2</tt>, and <tt>p1 > p2</tt> then VE 1 should do more I/O than VE 2.
 [[Category: Resource management]]