[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v2 11/12] mirror: support more than one in-fligh
From: |
Paolo Bonzini |
Subject: |
Re: [Qemu-devel] [PATCH v2 11/12] mirror: support more than one in-flight AIO operation |
Date: |
Mon, 21 Jan 2013 13:55:31 +0100 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130110 Thunderbird/17.0.2 |
Il 21/01/2013 13:35, Kevin Wolf ha scritto:
> I'm wondering if a whole bitmap is really appropriate when you have at
> most 16 parallel requests in flight. Other places in qemu (like
> copy-on-read or qcow2 cluster allocation) use lists of in-flight
> requests instead.
>
> I'm not requesting a change here, just wondering what the reasons are
> and whether this, or the other places, or none of both should be changed
> long term.
The reason is simply that the code is reasoning in bitmaps a lot
(cow_bitmap and of course the dirty bitmap), so it was a natural pick
and the memory usage is not important. It is simpler and more efficient
than a linear scan.
I have the required information already in the MirrorOp struct indeed,
but I didn't need those in a list.
Paolo
- Re: [Qemu-devel] [PATCH v2 08/12] mirror: allow customizing the granularity, (continued)
- [Qemu-devel] [PATCH v2 10/12] mirror: add buf-size argument to drive-mirror, Paolo Bonzini, 2013/01/16
- [Qemu-devel] [PATCH v2 11/12] mirror: support more than one in-flight AIO operation, Paolo Bonzini, 2013/01/16
- [Qemu-devel] [PATCH v2 12/12] mirror: support arbitrarily-sized iterations, Paolo Bonzini, 2013/01/16
- [Qemu-devel] [PATCH v2 06/12] block: return count of dirty sectors, not chunks, Paolo Bonzini, 2013/01/16
- Re: [Qemu-devel] [PATCH v2 00/12] Drive mirroring performance improvements, Eric Blake, 2013/01/16
- Re: [Qemu-devel] [PATCH v2 00/12] Drive mirroring performance improvements, Kevin Wolf, 2013/01/21