Il 25/11/2013 16:24, Peter Lieven ha scritto:
> On 25.11.2013 16:14, Paolo Bonzini wrote:
>> Il 25/11/2013 16:07, Peter Lieven ha scritto:
>>>>> since the convert process is basically a sync operation it might
>>>>> be benificial in some case to change the hardcoded I/O buffer
>>>>> size to an alternate (greater) value.
>>>> Do you really need the extra knob? You can just add to BlockLimits the
>>>> optimal transfer length, and use it unconditionally.
>>> If you say patch 5 and 3 are ok. What could be done is to remove
>>> this knob and increase the iobuf_size to cluster_size if cluster_size
>>> is greater.
>> Yes, that makes sense because it avoids unnecessary COW.
>>
>>> I do not want to increase the default iobuf size to anything
>>> greater than 2MB. I do not know why this was choosen, but maybe
>>> there was a reason for it.
>> I think it is fine to increase it to the cluster_size or even to the
>> optimal transfer length (new BlockLimits field).
> okay scsi speaking:
> bs->bl.optimal_transfer_length = max(iscsilun->bl.opt_xfer_len,
> iscsilun->bl.opt_unmap_gran) ?
I was thinking more of
bs->bl.optimal_transfer_length = lun2qemu(iscsilun->bl.opt_xfer_len);
...
iobuf_size =
max(bs->bl.optimal_transfer_length, cluster_sectors)*512;
iobuf_size = min(16MB, max(2MB, iobuf_size));
> bdi->cluster_size as in Patch 3?
Yes, that one's fine.
>> * clamp maximum size to optimal transfer length.
> or increase it if its larger. for the dell equallogic storages we use
> the opt_unmap_gran
> is 15MB!!!
You're right, maximum size is really bounded anyway by iobuf_size.
>> * then, round final sector down to unmap granularity
> okay, i will try your modification to only round down the last sector.
> would you use bdi->cluster_size or the unmap alignment field from the
> bs->bl?
bdi->cluster_size. The qemu-img reason for the rounding is to avoid
unnecessary COW, which is what bdi->cluster_size is about. It just so
happens that it helps your benchmark too. :)
Paolo
> Peter
