2011-03-07 22:53:43

by Justin TerAvest

Subject: [PATCH] blk-cgroup: Lower minimum weight from 100 to 10.

We've found that we still get good, useful isolation at weights this
low. I'd like to adjust the minimum so that any other changes can take
these values into account.
---
Documentation/cgroups/blkio-controller.txt | 2 +-
block/blk-cgroup.h | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/Documentation/cgroups/blkio-controller.txt b/Documentation/cgroups/blkio-controller.txt
index 4ed7b5c..24314fd 100644
--- a/Documentation/cgroups/blkio-controller.txt
+++ b/Documentation/cgroups/blkio-controller.txt
@@ -140,7 +140,7 @@ Proportional weight policy files
- Specifies per cgroup weight. This is default weight of the group
on all the devices until and unless overridden by per device rule.
(See blkio.weight_device).
- Currently allowed range of weights is from 100 to 1000.
+ Currently allowed range of weights is from 10 to 1000.

- blkio.weight_device
- One can specify per cgroup per device rules using this interface.
diff --git a/block/blk-cgroup.h b/block/blk-cgroup.h
index ea4861b..57e7234 100644
--- a/block/blk-cgroup.h
+++ b/block/blk-cgroup.h
@@ -240,7 +240,7 @@ static inline char *blkg_path(struct blkio_group *blkg) { return NULL; }

#endif

-#define BLKIO_WEIGHT_MIN 100
+#define BLKIO_WEIGHT_MIN 10
#define BLKIO_WEIGHT_MAX 1000
#define BLKIO_WEIGHT_DEFAULT 500

--
1.7.3.1

2011-03-08 14:31:14

by Vivek Goyal

Subject: Re: [PATCH] blk-cgroup: Lower minimum weight from 100 to 10.

On Mon, Mar 07, 2011 at 02:53:14PM -0800, Justin TerAvest wrote:
> We've found that we still get good, useful isolation at weights this
> low. I'd like to adjust the minimum so that any other changes can take
> these values into account.
> ---
> Documentation/cgroups/blkio-controller.txt | 2 +-
> block/blk-cgroup.h | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>

Looks good to me.

Acked-by: Vivek Goyal <[email protected]>

Thanks
Vivek

> diff --git a/Documentation/cgroups/blkio-controller.txt b/Documentation/cgroups/blkio-controller.txt
> index 4ed7b5c..24314fd 100644
> --- a/Documentation/cgroups/blkio-controller.txt
> +++ b/Documentation/cgroups/blkio-controller.txt
> @@ -140,7 +140,7 @@ Proportional weight policy files
> - Specifies per cgroup weight. This is default weight of the group
> on all the devices until and unless overridden by per device rule.
> (See blkio.weight_device).
> - Currently allowed range of weights is from 100 to 1000.
> + Currently allowed range of weights is from 10 to 1000.
>
> - blkio.weight_device
> - One can specify per cgroup per device rules using this interface.
> diff --git a/block/blk-cgroup.h b/block/blk-cgroup.h
> index ea4861b..57e7234 100644
> --- a/block/blk-cgroup.h
> +++ b/block/blk-cgroup.h
> @@ -240,7 +240,7 @@ static inline char *blkg_path(struct blkio_group *blkg) { return NULL; }
>
> #endif
>
> -#define BLKIO_WEIGHT_MIN 100
> +#define BLKIO_WEIGHT_MIN 10
> #define BLKIO_WEIGHT_MAX 1000
> #define BLKIO_WEIGHT_DEFAULT 500
>
> --
> 1.7.3.1

2011-03-08 14:37:03

by Jens Axboe

Subject: Re: [PATCH] blk-cgroup: Lower minimum weight from 100 to 10.

On 2011-03-07 23:53, Justin TerAvest wrote:
> We've found that we still get good, useful isolation at weights this
> low. I'd like to adjust the minimum so that any other changes can take
> these values into account.

Justin, this looks fine to me if Vivek agrees. You did not sign-off on
the patch though, so I can't apply it yet. Please re-send with the ack
from Vivek added and your signed-off-by as well.

--
Jens Axboe

2011-03-08 15:04:11

by Vivek Goyal

Subject: Re: [PATCH] blk-cgroup: Lower minimum weight from 100 to 10.

On Tue, Mar 08, 2011 at 03:36:55PM +0100, Jens Axboe wrote:
> On 2011-03-07 23:53, Justin TerAvest wrote:
> > We've found that we still get good, useful isolation at weights this
> > low. I'd like to adjust the minimum so that any other changes can take
> > these values into account.
>
> Justin, this looks fine to me if Vivek agrees. You did not sign-off on
> the patch though, so I can't apply it yet. Please re-send with the ack
> from Vivek added and your signed-off-by as well.

The primary reason I am fine with wider weight range because there are
many situations where we just don't get effective 10 times service
differentiation for various reasons like workload don't seem to be
active at the service tree at the same time or if we decide to swith
off idling on faster storage etc.

So because of this leaky nature of service differentiation, it might help to
have a wider range of allowed weights.

Thanks
Vivek