[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] Condor to reboot a machine

On Jan 9, 2014, at 1:20 PM, Rich Pieri <ratinox@xxxxxxx> wrote:

> Tiago Macarios wrote:
>> I'll have a look at it. AFAIR we tried using EC2 with HVM instances a
>> while back and the penalty was way more than 2-3%. I'll double check on
>> that.
> I can't speak to EC2. I have currently a rack full of OpenStack and our
> actual production runs on bare metal vs. user-mode virtualized are
> showing a 3.5% overhead for virtualization. These are older PowerEdge
> 1950 rack units and they're a bit strapped for RAM (only 2GB per core
> right now) under load. Newer hardware with more ram per core would do
> better.

Unfortunately, there's always the difference between "average" and what one might be running...  Certain network-heavy or I/O-heavy workloads can still do nasty things to VMs.

That said, even the "nasty" I/O workloads of a few years ago are increasingly doable in VMs with correct tuning.  Unfortunately, "with correct tuning" may not be equivalent to "with default tuning".

Things are getting better as the years go on!