RE: [Xen-devel] Rate adaption during live migration

[Daniel Stodden <stodden@xxxxxxxxxx>]

Hi Ian.

On Sun, 2007-10-14 at 21:54 +0100, Ian Pratt wrote:
> > I always thought adaptive transmission rates in live migration were
> > switched on by default. Now I notice this does not seem to even be the
> > case.
> > 
> > Is there any specific reason? Just because I'm interested. Won't work?
> 
> It certainly used to work, and I can't think of any reason why it would have 
> bit rotted. 

Well then why, forgive my perseverance, isn't ADAPTIVE_SAVE defined?

> > More importantly: Regarding the behavior when turned off, this turns
> > off > RATE_IS_MAX() as well. Have been playing with totally pathetic cases
> > and was surprised to learn that the present configuration would happily
> > exercise them through maximum number of iterations. Is this
> > intentional?
> 
> Yep, it goes flat out if the rate limiting is disabled.

Sorry, I do not understand this, perhaps due to may ignorance.

There is
        (sent_this_iter > sent_last_iter) && RATE_IS_MAX()

as one primary condition to leave the precopy loop. To me this reads as
'We're onverging the WWS'. But undef'd ADAPTIVE_SAVE states

#define RATE_IS_MAX() 0

which unless I've been terribly missing something should rather be 1,
since (a) it es indeed writing full speed in that mode and (b) this
nullifies the whole check regarding the WWS (unless the latter is < 50
pages).

Any help appreciated.

Thanks for listening,
daniel

-- 
Daniel Stodden
LRR     -      Lehrstuhl fÃr Rechnertechnik und Rechnerorganisation
Institut fÃr Informatik der TU MÃnchen             D-85748 Garching
http://www.lrr.in.tum.de/~stodden         mailto:stodden@xxxxxxxxxx
PGP Fingerprint: F5A4 1575 4C56 E26A 0B33  3D80 457E 82AE B0D8 735B



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel