[Beowulf] Docker vs KVM paper by IBM

Gavin W. Burris bug at wharton.upenn.edu
Wed Jan 28 10:19:16 PST 2015


You've got it all wrong, Joe.  I repeat... Docker is a great enabler!
Don't all of your researchers yearn to be sysadmins?  Don't all
sysadmins yearn to be done with OS issues in order to free up more time
to chase amber lights on hardware?  

But seriously, more options are a good thing.  Docker is all about
options.  There are serious pitfalls, though, when application
developers are making distributions that have hard dependencies on an
entire OS container.  Sure, distribute your container, but also consider
writing code that can be compiled easily in a self-contained directory
like a home directory.

Cheers.

On 12:46PM Wed 01/28/15 -0500, Joe Landman wrote:
> On 1/28/15, 12:30 PM, Jason Riedy wrote:
> >And Andrew Holway writes:
> >>Docker was encouraging us to do very sloppy work.
> >Define "us."  My perspective is that Docker-ish things can let
> >system staff focus on the hardware and low-level interfaces
> >maintained exactly the way you intend.  Then "application
> >consultants," if you have those on staff, can help areas develop
> >their images in whatever way they need.
> >
> >And grad students can just throw any old crud together at the last
> >moment, use it, and not muck things up for people who spend (far
> >too much) time getting things set up just right (only to need
> >something else for the next project).
> >
> >The sloppy part of the problem can be addressed sloppily and the
> >fussy part can be addressed carefully when appropriate.
> 
> +1  ... beat me to the punch ...
> 
> The same arguments made by Andrew/Gavin could be made against VMs, and other
> things.  And those arguments are just as wrong there as they are here.
> 
> Docker has a concept and an implementation.  Complain about the
> implementation if you must, but the concept has been bought into by the
> major distro vendors and it will be interesting to see what emerges.  Note
> also that the implementation of other things (like, I dunno, SR-IOV and kvm
> pass through?) have also been "controversial" in the past (and may still be
> for all I know, though I've not heard much recently), as they poke through
> the "guaranteed" isolation of a VM.
> 
> Docker implementation will get better and more secure over time, and I
> suspect that the time scale for the major/minor bits getting better will be
> measurable in months.  Docker is a concept as well, and others may implement
> different versions of the concept.  I think there is at least one other from
> CoreOS folks (Rocket?).
> 
> What I'd like to see is a real API, and this is where things like what the
> Joyent team did with SmartOS are IMO a good base to work with.   Maybe not
> perfect (and it makes some assumptions about location of storage I don't
> like), but overall it is a very good concept and not hard to work with.  To
> plug them a bit, they've recently extended their bits to enable running
> Docker containers
> (https://www.joyent.com/blog/how-joyent-and-docker-are-working-together) .
> This is in part leveraging their SmartOS internals, but also, their recent
> ability to run Linux bits "natively" in containers on SmartOS
> (http://www.slideshare.net/bcantrill/illumos-lx).  If only ... infiniband
> were supported (hint hint two remaining IB vendors ... big hint hint).
> 
> Most of the people I've spoken to get why Docker is a good thing. Some are
> waiting for it to mature, some are not.  A few die hards may deride it.
> Their choice.
> 
> -- 
> Joseph Landman, Ph.D
> Founder and CEO
> Scalable Informatics, Inc.
> e: landman at scalableinformatics.com
> w: http://scalableinformatics.com
> t: @scalableinfo
> p: +1 734 786 8423 x121
> c: +1 734 612 4615
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

-- 
Gavin W. Burris
Senior Project Leader for Research Computing
The Wharton School
University of Pennsylvania


More information about the Beowulf mailing list