Re: Resource Allocation Question
you have 3 servers of 8 cores @ 2.4GHz - but your turbo boost frequency is
3.2 GHZ (
and for unknown reason, this Turbo Boost frequency is what ACS sees - but
that only means, say 30% effective over-provisioning of CPU - not issues in
other things at all !
Your total CPU/MHz math looks like this: 3 servers x 8 cores (HT) x 3.2GHZ
= 24 x 3.2GHz = 76.8 GHz - so the math is clear :)
You should be able to use up to "
cluster.cpu.allocated.capacity.disablethreshold " as percentage of your
total CPU/GHz - which by default (unless I'm wrong) 0.8, meaning 80% of
your total CPU/MHz = 0.8 x 76.8GHz = 61.44 GHz... Whatever is the color on
Dashboard - you should be able to use up to total of 61.44 GHz really -
after that you will receive error about "Insufficient Capacity..."
There is lower threshold used just to notify you
"cluster.cpu.allocated.capacity.notificationthreshold" when you are close
to limit... (you get email only once !)
These 2 thresholds can be set as part of Global Settings (requires MGMT
server restart afterwards) - or you can set it on the Cluster level (no
need for MGMT restart)
You can also set CPU over-provisioning as neeeded, for testing purposes,
say to 3.0 so you ACS will see 300% of your real GHz = 230.4 GHz
Hope that makes sense.
NOTE: one thing that I see perhaps confuses you - in Compute Offerings, you
define number of CPU cores (say, 4 cores), but also a speed of each CPU
core (say, 2.0 GHz), instead of real core frequency = this means your VM
will effectively be allocated 4 cores x 2.0GHz = 8GHz. So this means that
your whole cluster (but also a host which will be chosen eventually) needs
to have at least 8GHz free/available in order to successfully deploy VM.
So forget about real number of cores vs. virtual cores (as part of Compute
Offering) - better observe only GHz (Dashboard does display only GHz not
number of Cores, afaik?)
On Tue, 18 Dec 2018 at 18:56, Asai <asai@xxxxxxxxxxxxxxxxxxxxx> wrote:
> Thank you for your response.
> So, is there any way we can take advantage of the processor speed that
> doesn’t seem to be allocated? We’re not clear on how this works. It seems
> like there’s a tremendous amount of underutilization here? Although we
> have allocated 20 out of 24 cores, it seems like we’re not able to access
> the resources the processors like they should be able to be.
> What are we missing here?
> > On Dec 13, 2018, at 2:27 PM, Andrija Panic <andrija.panic@xxxxxxxxx>
> > In my case, hyperthreading number of cores (whatever OS sees as number of
> > cores) times the TurboBoost GHZ is what I get, so for 2x8 cored 2.6Ghz
> > cores with hyperthreading) I get 32 x 3.4 Ghz, which is not what you can
> > achieve at any time (that frequency can be obtained on only very few
> > at any moment, not nearly all cores).
> > Go to Host in GUI and check its statistics, you will see - take into
> > account any overprovisioning applied....
> > Cheers
> > On Thu, Dec 13, 2018, 20:19 Rafael Weingärtner <
> > wrote:
> >> You have a four core CPU, with 8 threads each? Then, each thread is 2.4
> >> GHz. So, 2.4 * 4 * 8.
> >> On Thu, Dec 13, 2018 at 6:14 PM Asai <asai@xxxxxxxxxxxxxxxxxxxxx>
> >>> Greetings,
> >>> I have a simple question regarding Cloudstack resource allocation.
> >>> In the dashboard, I’m seeing that our Memory is 25 out of 30 GB and the
> >>> circle is colored red. # of CPUs is 20 out of 24 and is colored red.
> >> But
> >>> CPU is 33 GHz out of 76 GHz. We have a single Xeon E5-2620 v3 @
> >>> This is a 4 core / 8 thread CPU. How is that we have 76 GHz available?
> >> It
> >>> seems like we’d have around 57 or so GHz available.
> >>> Thanks for your assistance.
> >>> Asai
> >> --
> >> Rafael Weingärtner