Open Stack

It depends on the use-case, there is a point when most of the time the VMs
on the compute node are idle.

A use-case yahoo! is doing is letting developers have many VMs, in this
case those VMs are mostly idle.

Given a hypervisor with 12 cores, we can place say 12x2 core VMs on there,
this is a total of 24 VM cores, in reality most of the time *all* the
developers using those VMs will not be utilizing all 24 'virtual' cores
(highly unlikely that all those users conspired to do this at the same
time, even if they do this is where the linux scheduler will get
involved). So there are reasons to do this (save money, improve
efficiency), of course figuring out the right balance is up to u. Likely
don't do 8 vms cores on a 2 core box, I would recommend buying better
hardware before u do this ;)

On 12/23/13, 12:55 PM, "Cristian Falcas" <cristi.falcas at gmail.com> wrote:

>There is no point in using 8 virtual cores in compute node with 2
>cores. The same is valid for using swap as memory to reach the desired
>12gb.
>
>Of course, if you don't plan on using that machine for any real work,
>you can do it.
>
>
>
>On Mon, Dec 23, 2013 at 6:39 PM, Joshua Harlow <harlowja at yahoo-inc.com>
>wrote:
>> Nope, u can over provision on most all of the resources (CPU, ram,
>>disk) u
>> described there. Ram is the tricky one as the Linux oom killer may
>>start to
>> get involved when u push the ram limits to high. But there is nothing
>> stopping u from running 8 or more vms on a box, depending on the over
>> provision ratio u are ok with...
>>
>> Sent from my really tiny device...
>>
>> On Dec 23, 2013, at 3:55 AM, "Vikas Parashar" <para.vikas at gmail.com>
>>wrote:
>>
>> Thanks Cristian,
>>
>> Will elasticity  be limited to 4 Cores/4GB  (The max capacity of a
>>physical
>> host) ?
>>
>>
>> On Mon, Dec 23, 2013 at 5:00 PM, Cristian Falcas
>><cristi.falcas at gmail.com>
>> wrote:
>>>
>>> Hi,
>>>
>>> From what I know you can resize a machine, but this involves
>>> rebuilding the instance: openstack will create a snapshot of the
>>> machine an recreate the instance with the new snapshot and a new
>>> flavor. This is not very fast from my experience, so you will have a
>>> considerable downtime doing this, depending on the size of the current
>>> instance and how fast is your storage.
>>>
>>> Best regards,
>>> Cristian Falcas
>>>
>>>
>>>
>>> On Mon, Dec 23, 2013 at 12:03 PM, Vikas Parashar <para.vikas at gmail.com>
>>> wrote:
>>> > Hi,
>>> >
>>> > IaaS is all about elastic computing. I can stretch resources as per
>>>my
>>> > need
>>> > - increasing/decreasing the number of cores, RAM allocated etc..
>>> >
>>> > My question is - how does openStack achieve this elasticity for both
>>> > computation and RAM.
>>> >
>>> > If I create an image with 2 cores and 4 GB RAM (and one day I need to
>>> > increase this to, lets say - 6 Cores and 12 GB RAM), but all the
>>> > physical
>>> > hosts that I currently have (for Compute and RAM) at my disposal
>>>have a
>>> > max
>>> > of 4 Cores and 4 GB RAM each..
>>> >
>>> > Using openStack -
>>> >
>>> > a) is this possible (as long as the total cores and total RAM
>>>required
>>> > is
>>> > less than the group-total) ? If yes, how is this achieved.
>>> >
>>> > b) or the elasticity will be limited to 4 Cores/4GB  (The max
>>>capacity
>>> > of a
>>> > physical host) ? If no, then is it possible to achieve it ?
>>> >
>>> > _______________________________________________
>>> > Mailing list:
>>> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>>> > Post to     : openstack at lists.openstack.org
>>> > Unsubscribe :
>>> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>>> >
>>
>>
>> _______________________________________________
>> Mailing list: 
>>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>> Post to     : openstack at lists.openstack.org
>> Unsubscribe : 
>>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack