Yeah, I'm planning on using pv_ops for both the dom0 and domU kernels.
Any movement towards making TMEM available would be great; it's one of
things we're looking forward to testing and using (along with Remus)
I'm systems engineer at a company running about 50 blogs/social
, all on
top of a heavily virtualized infrastructure (Xen [v3.2] is our sole
platform) . We've got about 80 dom0's, holding about 125 domU's and on
a few of
those boxes, especially those holding one or more of our web server
run into "out of memory" crashes from time to time. So, we're hoping
to see if TMEM helps reduce our memory woes. We tend to mix and match
domU's within dom0's, so some boxes will have 1 memory-hungry web
alongside a number of smaller, low-usage, low-memory, developer boxes.
what we've read, TMEM might be perfect for our setup.
On 05/14/2010 08:26 AM, Dan Magenheimer wrote:
is primarily of use in PV guests, less so in dom0.
Are you putting pv_ops kernels in your PV guests (domUs)?
so, Jeremy has been a big proponent of tmem and we can work
together to get it into his pv_ops tree. But I think most of the
focus for the pv_ops work has been to provide dom0 support. Upstream
kernels already provide domU support, so I’ve been instead working on
linux-kernel-mailing-list toward getting tmem patches accepted upstream.
not, just using a tmem-modified dom0 will probably not help
you. I have been working on tmem-modified rpm’s for various
RedHat-ish kernels and I know Jan Beulich has been working on putting
support into various SuSE kernels.
also be interested in more information about your
planned usage model for tmem.
I'm building Jeremy's pv_ops dom0 kernel (18.104.22.168) for use with Xen
it looks like the Transcendant Memory options arent in the kernel
TMEM not available in pv_ops? I've been looking forward to using it
but also really wanna work with pv_ops moving forward.