5 cgroup subsys "blkio" implements the block io controller. There seems to be
6 a need of various kinds of IO control policies (like proportional BW, max BW)
7 both at leaf nodes as well as at intermediate nodes in a storage hierarchy.
8 Plan is to use the same cgroup based management interface for blkio controller
9 and based on user options switch IO policies in the background.
11 In the first phase, this patchset implements proportional weight time based
12 division of disk policy. It is implemented in CFQ. Hence this policy takes
13 effect only on leaf nodes when CFQ is being used.
17 You can do a very simple testing of running two dd threads in two different
18 cgroups. Here is what you can do.
20 - Enable group scheduling in CFQ
21 CONFIG_CFQ_GROUP_IOSCHED=y
23 - Compile and boot into kernel and mount IO controller (blkio).
25 mount -t cgroup -o blkio none /cgroup
28 mkdir -p /cgroup/test1/ /cgroup/test2
30 - Set weights of group test1 and test2
31 echo 1000 > /cgroup/test1/blkio.weight
32 echo 500 > /cgroup/test2/blkio.weight
34 - Create two same size files (say 512MB each) on same disk (file1, file2) and
35 launch two dd threads in different cgroup to read those files.
38 echo 3 > /proc/sys/vm/drop_caches
40 dd if=/mnt/sdb/zerofile1 of=/dev/null &
41 echo $! > /cgroup/test1/tasks
42 cat /cgroup/test1/tasks
44 dd if=/mnt/sdb/zerofile2 of=/dev/null &
45 echo $! > /cgroup/test2/tasks
46 cat /cgroup/test2/tasks
48 - At macro level, first dd should finish first. To get more precise data, keep
49 on looking at (with the help of script), at blkio.disk_time and
50 blkio.disk_sectors files of both test1 and test2 groups. This will tell how
51 much disk time (in milli seconds), each group got and how many secotors each
52 group dispatched to the disk. We provide fairness in terms of disk time, so
53 ideally io.disk_time of cgroups should be in proportion to the weight.
55 Various user visible config options
56 ===================================
57 CONFIG_CFQ_GROUP_IOSCHED
58 - Enables group scheduling in CFQ. Currently only 1 level of group
61 CONFIG_DEBUG_CFQ_IOSCHED
62 - Enables some debugging messages in blktrace. Also creates extra
63 cgroup file blkio.dequeue.
65 Config options selected automatically
66 =====================================
67 These config options are not user visible and are selected/deselected
68 automatically based on IO scheduler configuration.
71 - Block IO controller. Selected by CONFIG_CFQ_GROUP_IOSCHED.
73 CONFIG_DEBUG_BLK_CGROUP
74 - Debug help. Selected by CONFIG_DEBUG_CFQ_IOSCHED.
76 Details of cgroup files
77 =======================
79 - Specifies per cgroup weight. This is default weight of the group
80 on all the devices until and unless overridden by per device rule.
81 (See blkio.weight_device).
82 Currently allowed range of weights is from 100 to 1000.
85 - One can specify per cgroup per device rules using this interface.
86 These rules override the default value of group weight as specified
89 Following is the format.
91 #echo dev_maj:dev_minor weight > /path/to/cgroup/blkio.weight_device
92 Configure weight=300 on /dev/sdb (8:16) in this cgroup
93 # echo 8:16 300 > blkio.weight_device
94 # cat blkio.weight_device
98 Configure weight=500 on /dev/sda (8:0) in this cgroup
99 # echo 8:0 500 > blkio.weight_device
100 # cat blkio.weight_device
105 Remove specific weight for /dev/sda in this cgroup
106 # echo 8:0 0 > blkio.weight_device
107 # cat blkio.weight_device
112 - disk time allocated to cgroup per device in milliseconds. First
113 two fields specify the major and minor number of the device and
114 third field specifies the disk time allocated to group in
118 - number of sectors transferred to/from disk by the group. First
119 two fields specify the major and minor number of the device and
120 third field specifies the number of sectors transferred by the
121 group to/from the device.
123 - blkio.io_service_bytes
124 - Number of bytes transferred to/from the disk by the group. These
125 are further divided by the type of operation - read or write, sync
126 or async. First two fields specify the major and minor number of the
127 device, third field specifies the operation type and the fourth field
128 specifies the number of bytes.
131 - Number of IOs completed to/from the disk by the group. These
132 are further divided by the type of operation - read or write, sync
133 or async. First two fields specify the major and minor number of the
134 device, third field specifies the operation type and the fourth field
135 specifies the number of IOs.
137 - blkio.io_service_time
138 - Total amount of time between request dispatch and request completion
139 for the IOs done by this cgroup. This is in nanoseconds to make it
140 meaningful for flash devices too. For devices with queue depth of 1,
141 this time represents the actual service time. When queue_depth > 1,
142 that is no longer true as requests may be served out of order. This
143 may cause the service time for a given IO to include the service time
144 of multiple IOs when served out of order which may result in total
145 io_service_time > actual time elapsed. This time is further divided by
146 the type of operation - read or write, sync or async. First two fields
147 specify the major and minor number of the device, third field
148 specifies the operation type and the fourth field specifies the
149 io_service_time in ns.
152 - Total amount of time the IOs for this cgroup spent waiting in the
153 scheduler queues for service. This can be greater than the total time
154 elapsed since it is cumulative io_wait_time for all IOs. It is not a
155 measure of total time the cgroup spent waiting but rather a measure of
156 the wait_time for its individual IOs. For devices with queue_depth > 1
157 this metric does not include the time spent waiting for service once
158 the IO is dispatched to the device but till it actually gets serviced
159 (there might be a time lag here due to re-ordering of requests by the
160 device). This is in nanoseconds to make it meaningful for flash
161 devices too. This time is further divided by the type of operation -
162 read or write, sync or async. First two fields specify the major and
163 minor number of the device, third field specifies the operation type
164 and the fourth field specifies the io_wait_time in ns.
167 - Total number of bios/requests merged into requests belonging to this
168 cgroup. This is further divided by the type of operation - read or
169 write, sync or async.
172 - Total number of requests queued up at any given instant for this
173 cgroup. This is further divided by the type of operation - read or
174 write, sync or async.
176 - blkio.avg_queue_size
177 - Debugging aid only enabled if CONFIG_DEBUG_CFQ_IOSCHED=y.
178 The average queue size for this cgroup over the entire time of this
179 cgroup's existence. Queue size samples are taken each time one of the
180 queues of this cgroup gets a timeslice.
182 - blkio.group_wait_time
183 - Debugging aid only enabled if CONFIG_DEBUG_CFQ_IOSCHED=y.
184 This is the amount of time the cgroup had to wait since it became busy
185 (i.e., went from 0 to 1 request queued) to get a timeslice for one of
186 its queues. This is different from the io_wait_time which is the
187 cumulative total of the amount of time spent by each IO in that cgroup
188 waiting in the scheduler queue. This is in nanoseconds. If this is
189 read when the cgroup is in a waiting (for timeslice) state, the stat
190 will only report the group_wait_time accumulated till the last time it
191 got a timeslice and will not include the current delta.
194 - Debugging aid only enabled if CONFIG_DEBUG_CFQ_IOSCHED=y.
195 This is the amount of time a cgroup spends without any pending
196 requests when not being served, i.e., it does not include any time
197 spent idling for one of the queues of the cgroup. This is in
198 nanoseconds. If this is read when the cgroup is in an empty state,
199 the stat will only report the empty_time accumulated till the last
200 time it had a pending request and will not include the current delta.
203 - Debugging aid only enabled if CONFIG_DEBUG_CFQ_IOSCHED=y.
204 This is the amount of time spent by the IO scheduler idling for a
205 given cgroup in anticipation of a better request than the exising ones
206 from other queues/cgroups. This is in nanoseconds. If this is read
207 when the cgroup is in an idling state, the stat will only report the
208 idle_time accumulated till the last idle period and will not include
212 - Debugging aid only enabled if CONFIG_DEBUG_CFQ_IOSCHED=y. This
213 gives the statistics about how many a times a group was dequeued
214 from service tree of the device. First two fields specify the major
215 and minor number of the device and third field specifies the number
216 of times a group was dequeued from a particular device.
219 - Writing an int to this file will result in resetting all the stats
224 /sys/block/<disk>/queue/iosched/group_isolation
226 If group_isolation=1, it provides stronger isolation between groups at the
227 expense of throughput. By default group_isolation is 0. In general that
228 means that if group_isolation=0, expect fairness for sequential workload
229 only. Set group_isolation=1 to see fairness for random IO workload also.
231 Generally CFQ will put random seeky workload in sync-noidle category. CFQ
232 will disable idling on these queues and it does a collective idling on group
233 of such queues. Generally these are slow moving queues and if there is a
234 sync-noidle service tree in each group, that group gets exclusive access to
235 disk for certain period. That means it will bring the throughput down if
236 group does not have enough IO to drive deeper queue depths and utilize disk
237 capacity to the fullest in the slice allocated to it. But the flip side is
238 that even a random reader should get better latencies and overall throughput
239 if there are lots of sequential readers/sync-idle workload running in the
242 If group_isolation=0, then CFQ automatically moves all the random seeky queues
243 in the root group. That means there will be no service differentiation for
244 that kind of workload. This leads to better throughput as we do collective
245 idling on root sync-noidle tree.
247 By default one should run with group_isolation=0. If that is not sufficient
248 and one wants stronger isolation between groups, then set group_isolation=1
249 but this will come at cost of reduced throughput.
253 - Currently only sync IO queues are support. All the buffered writes are
254 still system wide and not per group. Hence we will not see service
255 differentiation between buffered writes between groups.