Re: [Qemu-devel] [PATCH V3 for-1.6 0/5] Leaky bucket throttling and features

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1574 bytes --]

On 07/23/2013 10:29 AM, Benoît Canet wrote:
> The first patch fixes the throttling which was broken by a previous commit.
> 
> The next patch replace the existing throttling algorithm by the well described
> leaky bucket algorithm.
> 
> Third patch implement bursting by adding *_threshold parameters to
> qmp_block_set_io_throttle.
> 
> The last one allow to define the max size of an io when throttling by iops via
> iops_sector_count to avoid vm users cheating on the iops limit.
> 
> The last patch adds a metric reflecting how much the I/O are throttled.
> 
> since v1:
>     Add throttling percentage metric [Benoît]
> 
> since v2:
>     Enable timer only during I/O activity [Stefan]
>     Mark function as coroutine_fn [Stefan]
>     Guard these function checking they are in a coroutine [Stefan]
>     Use defines to access buckets [Stefan]
>     Fix typo [Stefan]
>     reset throttling metric on iddle [Benoît]
>     rename invalid to check_io_limit [Stefan]
> 
> Benoît Canet (5):
>   block: Repair the throttling code.

This is probably okay for 1.6 (in fact, now's the RIGHT time to be
fixing bugs).

>   block: Modify the throttling code to implement the leaky bucket
>     algorithm.

but this sounds too risky now that we have passed soft freeze; as
mentioned on the thread on blockdev-add, this part is definitely 1.7
material, and the series should be split and subject lines updated to
reflect that.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 621 bytes --]