3:19 p.m.
--Apple-Mail-FC5CF2F4-77D9-4AB3-AECA-43D5365F3844
Content-Type: text/plain;
charset=us-ascii
Content-Transfer-Encoding: quoted-printable
Are there any documents going more deeply in these sort of
considerations?=
This is exactly what I have been searching for lately, too. Please let me k=
now if you find anything (or blog posts, forums, books, etc). Thank you!
Sent from my mobile phone
On 6 Sep 2017, at 12:47, Gianluca Cecchi
<gianluca.cecchi(a)gmail.com> wrote=
:
=20
Hello,
I was talking with a guy expert in VMware and discussing performance of VM=
s in
respect of virtual cpus assigned to them in relation with mapping with t=
he real hw of the hypervisor underneath.
=20
One of the topics was numa usage and its overheads in case of a "too" big V=
M, in terms of both number of vcpus and memory amount.
Eg:=20
suppose host has 2 intel based sockets, with 6 cores and HT enabled and ha=
s 96Gb
of ram (distributed 48+48 between the 2 processors)
suppose I configure a VM with 16 vcpus (2:4:2): would be the mapping
respe=
cted at physical level or only a sort of "hint" for the
hypervisor?
Can I say that it would perform better if I configure it 12 vcpus and
mapp=
ing 1:6:2, because it can stay all inside one cpu?
=20
And what if I define a VM with 52Gb of ram? Can I say that it would perfor=
m in
general better if I try to get it all in one cpu related memory slots (=
eg not more than 48Gb in my example)?
>=20
Are there any documents going more deeply in these sort of
considerations?=
=20
Also, if one goes and sizes so that the biggest VM is able to all-stay ins=
ide one
cpu-memory, does it make sense to say that it will perform better in=
this scenario a cluster composed by 4 nodes, each one with 1 socket and 48G=
b of memory instead of a cluster of 2 nodes, each one with 2 sockets and 96G=
b of ram?
=20
Hope I have clarified my questions/doubts.
=20
=20
Thanks in advance for any insight,
Gianluca =20
_______________________________________________
Users mailing list
Users(a)ovirt.org
http://lists.ovirt.org/mailman/listinfo/users
--Apple-Mail-FC5CF2F4-77D9-4AB3-AECA-43D5365F3844
Content-Type: text/html;
charset=utf-8
Content-Transfer-Encoding: quoted-printable
<html><head><meta http-equiv=3D"content-type"
content=3D"text/html; charset=3D=
utf-8"></head><body
dir=3D"auto"><div><span></span></div><div><meta
http-equ=
iv=3D"content-type" content=3D"text/html;
charset=3Dutf-8"><div><blockquote t=
ype=3D"cite"><div dir=3D"ltr"><font
color=3D"#000000"><span style=3D"backgro=
und-color: rgba(255, 255, 255, 0);">Are there any documents going more deepl=
y in these sort of
considerations?</span></font></div></blockquote><div id=3D=
"AppleMailSignature"><br></div>This is exactly what I have been
searching fo=
r lately, too. Please let me know if you find anything (or blog posts,=
forums, books, etc). Thank you!<br><br>Sent from my mobile
phone</div=
<div><br>On 6 Sep 2017, at 12:47, Gianluca Cecchi
<<a href=3D"mailto:gia=
nluca.cecchi@gmail.com">gianluca.cecchi(a)gmail.com</a>&gt;
wrote:<br><br></di=
v><blockquote type=3D"cite"><div><div
dir=3D"ltr">Hello,<div>I was talking w=
ith a guy expert in VMware and discussing performance of VMs in respect of v=
irtual cpus assigned to them in relation with mapping with the real hw of th=
e hypervisor underneath.</div><div><br></div><div>One of the
topics was numa=
usage and its overheads in case of a "too" big VM, in terms of both number o=
f vcpus and memory
amount.</div><div>Eg: </div><div>suppose host has 2 i=
ntel based sockets, with 6 cores and HT enabled and has 96Gb of ram (distrib=
uted 48+48 between the 2 processors)</div><div>suppose I configure a VM with=
16 vcpus (2:4:2): would be the mapping respected at physical level or only a=
sort of "hint" for the hypervisor?</div><div>Can I say that it
would perfor=
m better if I configure it 12 vcpus and mapping 1:6:2, because it can stay a=
ll inside one cpu?</div><div><br></div><div>And what if I
define a VM with 5=
2Gb of ram? Can I say that it would perform in general better if I try to ge=
t it all in one cpu related memory slots (eg not more than 48Gb in my exampl=
e)?</div><div><br></div><div>Are there any documents going
more deeply in th=
ese sort of considerations?</div><div><br></div><div>Also,
if one goes and s=
izes so that the biggest VM is able to all-stay inside one cpu-memory, does i=
t make sense to say that it will perform better in this scenario a cluster c=
omposed by 4 nodes, each one with 1 socket and 48Gb of memory instead of a c=
luster of 2 nodes, each one with 2 sockets and 96Gb of
ram?</div><div><br></=
div><div>Hope I have clarified my
questions/doubts.</div><div><br></div><div=
<br></div><div>Thanks in advance for any
insight,</div><div>Gianluca =
</div></div>
</div></blockquote><blockquote
type=3D"cite"><div><span>____________________=
___________________________</span><br><span>Users mailing
list</span><br><sp=
an><a
href=3D"mailto:Users@ovirt.org">Users@ovirt.org</a></span><br><span><a=
href=3D"http://lists.ovirt.org/mailman/listinfo/users">http:...
rg/mailman/listinfo/users</a></span><br></div></blockquote></div></body></ht=
ml>=
--Apple-Mail-FC5CF2F4-77D9-4AB3-AECA-43D5365F3844--