RE: [PATCH 2 of 8] libxl: introduce libxl_set_relative_memory_target

[Dan Magenheimer <dan.magenheimer@oracle.com>]

> From: Jeremy Fitzhardinge [mailto:jeremy@goop.org]
> 
>  On 09/01/2010 01:03 PM, Dan Magenheimer wrote:
> > Indeed, that's what selfballooning does.  The xenstore watch
> > is irrelevant for selfballooning (though the watch also can be used
> > asynchronously for backwards compatibility).
> 
> There's no mechanism to make the balloon driver ignore the target
> watch,
> so any updates to xenstore will update the driver's target.

The selfballooning patch currently applies to the balloon driver
so it could easily disable the target watch, though it does not.
 
> > So, frankly, I think the "xm memset" functionality is largely
> > useless, but agree that it should be maintained in xl for backwards
> > compatibility.  But trying to comingle the concepts of maxmem
> > and target is a bad idea.
> 
> In the general case I think you're probably right (I can't see it being
> useful in a VPS hosting service, for example), but there are definitely
> special cases where it is useful.  Squashing down existing domains to
> make room for a new one, for example, or more app-specific uses.

Agreed in general, though I suspect sysadmins would be rather
peeved if/when a simple xm command in dom0 causes kernel OOMs
and application kills... or, worse, guest kernel crashes (which are
circumvented by minimum-target code in the linux-xen balloon
driver but NOT in the pvops balloon driver).

So "useless" is an overstatement, but "must be used extremely
carefully with knowledge of current activity in the guest
and/or willingness for the targeted guest or its applications
to die for a greater cause" is not an overstatement.
 
> Giving domains some real incentive to be economical with memory would
> probably change the landscape a lot.  But I don't think there's a real
> solution without knowing the specifics of that incentive.

Agreed.  Lacking a clear incentive though, reducing the disincentive
is a reasonable approach... which is what tmem+selfballooning does.

My long term view of the incentive is something like a VPS hoster
that offers service for $10/month, but offers it for $5/month
if using a tmem(+selfballooning)-enabled kernel.  This is essentially
like the electric utility companies that give customers a discount
if the customer allows them a remote-kill switch to turn off your air
conditioning if system-wide conditions warrant.

Dan