Definition of IO depth distribution in the terse output

Definition of IO depth distribution in the terse output

[asomers]

The man page section for the terse output format describes the IO
depth distribution fields as "<=1, 2, 4, 8, 16, 32, >=64".  Should the
last entry be ">32" instead?  Similarly, in the section for IO latency
distribution, should the final value be ">2000" instead of ">=2000"?

-Alan

Re: Definition of IO depth distribution in the terse output

[Jens Axboe]

On Thu, Jul 25 2013, asomers@gmail.com wrote:
> The man page section for the terse output format describes the IO
> depth distribution fields as "<=1, 2, 4, 8, 16, 32, >=64".  Should the
> last entry be ">32" instead?  Similarly, in the section for IO latency
> distribution, should the final value be ">2000" instead of ">=2000"?

The bucket named 32 is 32 and up (but not including) 64. And so forth.

-- 
Jens Axboe

Re: Definition of IO depth distribution in the terse output

[asomers]

I still don't understand how there isn't a gap.  Are you saying that
the buckets go like this:
<=1, >=2, >=4, >=8, >=16, >=32, >=64 ?

Is the lowest bucket misnamed, or is there a gap between 1 and 2?

On Thu, Jul 25, 2013 at 9:10 PM, Jens Axboe <axboe@kernel.dk> wrote:
> On Thu, Jul 25 2013, asomers@gmail.com wrote:
>> The man page section for the terse output format describes the IO
>> depth distribution fields as "<=1, 2, 4, 8, 16, 32, >=64".  Should the
>> last entry be ">32" instead?  Similarly, in the section for IO latency
>> distribution, should the final value be ">2000" instead of ">=2000"?
>
> The bucket named 32 is 32 and up (but not including) 64. And so forth.
>
> --
> Jens Axboe
>

Re: Definition of IO depth distribution in the terse output

[Jens Axboe]

On 07/26/2013 08:51 AM, asomers@gmail.com wrote:
> I still don't understand how there isn't a gap.  Are you saying that
> the buckets go like this:
> <=1, >=2, >=4, >=8, >=16, >=32, >=64 ?
> 
> Is the lowest bucket misnamed, or is there a gap between 1 and 2?

Bucket0		1
Bucket1		2..3
Bucket2		4..7
Bucket3		8..15
Bucket4		16..31
Bucket5		32..63
Bucket6		64..

-- 
Jens Axboe

Re: Definition of IO depth distribution in the terse output

[asomers]

On Fri, Jul 26, 2013 at 9:07 AM, Jens Axboe <axboe@kernel.dk> wrote:
> On 07/26/2013 08:51 AM, asomers@gmail.com wrote:
>> I still don't understand how there isn't a gap.  Are you saying that
>> the buckets go like this:
>> <=1, >=2, >=4, >=8, >=16, >=32, >=64 ?
>>
>> Is the lowest bucket misnamed, or is there a gap between 1 and 2?
>
> Bucket0         1

Should this be 0..1?  Or is there some guarantee that there will
always be at least one outstanding IO?

> Bucket1         2..3
> Bucket2         4..7
> Bucket3         8..15
> Bucket4         16..31
> Bucket5         32..63
> Bucket6         64..
>
> --
> Jens Axboe
>