Re: [PATCH 08/18] KVM test: Add basic file transfer test

From: Lucas Meneghel Rodrigues <lmr@redhat.com>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: kvm@vger.kernel.org, autotest@test.kernel.org
Subject: Re: [PATCH 08/18] KVM test: Add basic file transfer test
Date: Wed, 29 Sep 2010 14:43:13 -0300	[thread overview]
Message-ID: <1285782193.2794.19.camel@freedom> (raw)
In-Reply-To: <20100928132425.GF14385@redhat.com>

On Tue, 2010-09-28 at 15:24 +0200, Michael S. Tsirkin wrote:
> On Mon, Sep 27, 2010 at 06:43:54PM -0400, Lucas Meneghel Rodrigues wrote:
> > From: Amos Kong <akong@redhat.com>
> > 
> > This test is the basic test of transfering file between host and guest.
> > Try to transfer a large file from host to guest, and transfer it back
> > to host, then compare the files by calculate their md5 hash.
> > 
> > The default file size is 4000M, scp timeout is 1000s. It means if the
> > average speed is less than 4M/s, this test will be fail.
> > We can extend
> > this test by using another disk later, then we can transfer larger
> > files without the limit of first disk size.
> > 
> > Changes from v1:
> > - Use md5 to verify the integrity of files
> > - Try to use autotest API, such as, utils.system()
> > 
> > Signed-off-by: Amos Kong <akong@redhat.com>
> 
> Why scp_timeout? Not transfer_timeout?
> Is this really only scp file transfer to/from linux guest?
> Need to either name it so or generalize. Other things that
> need testing are NFS for linux guest, scp from windows, samba
> for linux and windows guests.

Ok, the new version of the test is ready. Here is the output of the
tests under Fedora 13 64 bit and Windows 7 64 bit, conveniently trimmed
down:

13:22:42 INFO | Creating 100MB file on host
13:22:42 DEBUG| Running 'dd if=/dev/urandom of=/root/autotest/client/tmp/tmpmOpO1z_kvm.smp2.Win7.64.file_transfer.remote/a.out bs=1M count=100'
13:22:56 INFO | Transfering file host -> guest, timeout: 1000s
13:23:04 INFO | File transfer host -> guest succeed, estimated throughput: 11.77MB/s
13:23:04 INFO | Transfering file guest -> host, timeout: 1000s
13:23:13 INFO | File transfer guest -> host succeed, estimated throughput: 11.59MB/s
13:23:13 DEBUG| Size of a.out: 104857600B
13:23:13 DEBUG| Size of c.out: 104857600B
13:23:13 INFO | Cleaning temp file on guest
13:23:13 DEBUG| Sending command: del C:\b.out
13:23:14 INFO | Cleaning temp files on host

13:12:28 INFO | Creating 100MB file on host
13:12:28 DEBUG| Running 'dd if=/dev/urandom of=/root/autotest/client/tmp/tmpp0B85Z_kvm.virtio_blk.smp2.Fedora.13.64.virtio_net.file_transfer.remote/a.out bs=1M count=100'
13:12:42 INFO | Transfering file host -> guest, timeout: 1000s
13:12:46 INFO | File transfer host -> guest succeed, estimated throughput: 25.41MB/s
13:12:46 INFO | Transfering file guest -> host, timeout: 1000s
13:12:50 INFO | File transfer guest -> host succeed, estimated throughput: 27.23MB/s
13:12:50 DEBUG| Size of a.out: 104857600B
13:12:50 DEBUG| Size of c.out: 104857600B
13:12:50 INFO | Cleaning temp file on guest
13:12:50 DEBUG| Sending command: rm -f /tmp/b.out
13:12:50 INFO | Cleaning temp files on host

I chose the file size 100MB just to test my changes, of course the
default size is still 4000MB. I'll send the updated version of the test,
won't send the full patchset again until the other issues are properly
fixed.

If you guys are wondering why the difference in the throughput figures,
well, I am also wondering. Maybe this is because windows is not using
virtio, maybe rss is slower, I don't know yet.

Cheers,

Lucas