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

Re: [Publicity] Seems the VM vs container debate is heating up

Here's yet another article that just came out on the topic from Steven Vaughan Nichols, who did meet with Mark H. and Richard Phelan at Collab. Don't think this topic came up though. 

On Tue, Apr 8, 2014 at 2:24 PM, Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx> wrote:
Yes, there are: miniOS is similar to OSv but less complete and only
supports PV guests.
The miniOS approach allows one POSIX application to run within one PV
guest virtual machine with minimal operating system overhead. At the
same time PV (and the newer PVH) guests are low overhead compared to HVM
guests as they don't provide any emulated interfaces.
Overall miniOS (or OSv) on a PV guest is a great fit for

On Tue, 8 Apr 2014, Sarah Conway wrote:
> I like Stefano's point about Xen building bridges here -- seems like a new twist on the VM vs. container debate. The
> fact that Xen is evolving with the times and it's not a binary choice might play reasonably well.  
> Are there are any major technical differences between the miniOS approach and containers? 
> Thanks,
> On Tue, Apr 8, 2014 at 1:13 PM, Lars Kurth <lars.kurth@xxxxxxx> wrote:
>       I think it's a great post
>       Lars
>       On 08/04/2014 14:54, Russ Pavlicek wrote:
>             I've got the following blog post queued for open.citrix.com later
>             today (it was my turn and this topic seemed timely).
>             I am leaving for the airport in a couple hours, so if anyone has
>             concerns, please let me know in the next hour or so if you can.
>             Otherwise, I may have to edit after it is autopublished later tonight.
>             It isn't a Xen Project piece per se, but I wanted to highlight the
>             MirageOS/OSv/etc. solution to highly dense VM app farms.
>             ---
>             Are Containers the Right Answer to the Wrong Question?
>             Next-Gen, High Density App Servers Don't Require Scrapping Your Hypervisor
>             Recently, I sat in a conference session extolling the seemingly
>             endless virtues of Linux Containers.  I heard claims that hypervisors
>             were old hat: ancient bloated engines which rely on inefficient
>             replication of a large operating system stack in order to serve up
>             applications.  The speaker painted a picture of a future where
>             hundreds of applications are virtualized on each piece of hardware.
>             "What is really needed," glowed the speaker, "is a lightweight,
>             efficient means of serving up application: containers."
>             Containers are cool, but not a panacea
>             Containers share the same kernel as the host, so they are not burdened
>             with the extra memory and CPU cycles it costs to replicate a full
>             operating system stack in a hypervisor scenario.  Compared to
>             hypervisor-generated virtual machines, containers can be fast and
>             lean.  But they are also limited.
>             Since Linux containers share the same kernel as the host, it is
>             impossible to run Windows.  Or FreeBSD. Or NetBSD.   Or another
>             version of the Linux kernel.  Or another Linux distribution which
>             requires a different kernel.  All of those scenarios are best handled
>             by a real hypervisor.  And the security aspect of hypervisors is huge,
>             worthy of a separate blog entry of its own.  Still, if you need an
>             environment within your organization where many workloads can leverage
>             a single kernel environment, containers can be a viable solution.
>             However, some of the most vocal container advocates insist that these
>             problems relating to containers are really application problems in
>             disguise.  Issues about kernel support and security are the results of
>             improper application design, they claim.  When we raise the bar on
>             applications so that they are based solely on access to application
>             servers, then the objections to containers will melt away -- and so
>             will hypervisors, for the most part.  Or that's what some of these
>             advocates claim, at least.
>             The death of the hypervisor is greatly exaggerated
>             But is there another scenario which could answer the call for highly
>             responsive and lightweight virtual instances which does not use the
>             container solution?  Maybe one that can actually leverage the
>             flexibility and security which is part and parcel with most
>             hypervisors?
>             Yes, in fact, there is.  And the key is in the very
>             application-centric future which some container advocates believe will
>             cause the hypervisor to become obsolete.
>             Behold the birth of the Cloud Operating System
>             Instead of deriding hypervisors for the "bloat" of replicating a full
>             operating system, we can replicate ultra-light application-centric
>             operating systems which are meant to live in a VM.  A number of these
>             lean, mean virtual operating systems have arisen recently, including
>             the Xen Project team's MirageOS and Cloudius Systems' OSv.  These
>             lightweight operating systems, sometimes called "cloud operating
>             systems", lack the expensive drivers needed to talk to hardware --
>             because they aren't meant to run on hardware.  And they lack
>             multiprocess capabilities because they are not intended to be general
>             timesharing systems, but dedicated application server systems.  As a
>             result, they are very small, very lightweight, and can start up
>             amazingly quickly in a VM environment.  They enable the vision of
>             light and fast application servers, while preserving the superior
>             security and flexibility of true hypervisors.
>             Don't throw the hypervisor baby out with the dirty application bath water
>             So if you are intrigued by the vision of masses of small, efficient
>             application VMs packed into a minimum of number of servers, but
>             cringing at the notion of totally retooling your virtualization
>             infrastructure and rethinking your entire security mindset, don't
>             fret.  That vision is achievable while maintaining the flexibility and
>             security that mature hypervisors deliver.  Focus on a streamlined
>             application stack, and the vision of dense, efficient application VMs
>             is achievable -- regardless of whether you are using hypervisors,
>             containers, or both.
>             On Tue, Apr 8, 2014 at 9:42 AM, Stefano Stabellini
>             <stefano.stabellini@xxxxxxxxxxxxx> wrote:
>                   Also this tells us that whatever tecnology we choose to have upstream
>                   QEMU stubdoms, we should make sure it is easy to use for other POSIX
>                   applications too, so that we can advertise it as a generic solution for
>                   Xen micro-VMs and position it as a secure alternative to containers.
>                   On Tue, 8 Apr 2014, Stefano Stabellini wrote:
>                         Given that most people only use one VM for one task, the rise of
>                         "micro-VMs" makes sense, whether they are based on OSv (miniOS or rump
>                         kernels) on traditional hypervisors or on Docker, ZeroVM or other types
>                         of containers.
>                         However what these articles don't take into account is that Xen can
>                         already handle both kind of workloads extremely well: Xen can handle
>                         fat
>                         old-style VMs, with HVM guests, and new style lightweight micro-VMs,
>                         with miniOS (or rump kernels or OSv) inside a PV guest.
>                         If I manage to carve the time to do it I might be able to write a blog
>                         post about it.
>                         On Tue, 8 Apr 2014, Lars Kurth wrote:
>                               See:
>                               *
>                               http://blogs.bromium.com/2014/04/07/vms-the-new-infrastructure-anachronism/
>                               * On the other hand there is this article:
>                               http://www.zdnet.com/hypervisors-are-the-pillars-of-the-cloud-not-the-achilles-heel-7000027931/
>                               (both by Simon Crosby)
>                               Generally for the security discussion, which is related,
>                               the following
>                               articles may be relevant:
>                               *
>                               http://www.zdnet.com/hypervisors-the-clouds-potential-security-achilles-heel-7000027846/
>                               *
>                               http://www.economist.com/blogs/babbage/2014/03/computer-security
>                               The following two may also be useful (and contain quite a
>                               few useful links)
>                               despite also being partly a product pitch:
>                               *
>                               http://blogs.bromium.com/2013/03/27/micro-virtualization-for-the-security-architect/
>                               *
> http://blogs.bromium.com/2013/04/24/micro-virtualization-for-the-security-architect-2-of-2-isolation-%E2%89%A0-protect
>                               ion/
>                               Lars
>                               _______________________________________________
>                               Publicity mailing list
>                               Publicity@xxxxxxxxxxxxxxxxxxxx
>                               http://lists.xenproject.org/cgi-bin/mailman/listinfo/publicity
>                         _______________________________________________
>                         Publicity mailing list
>                         Publicity@xxxxxxxxxxxxxxxxxxxx
>                         http://lists.xenproject.org/cgi-bin/mailman/listinfo/publicity
>                   _______________________________________________
>                   Publicity mailing list
>                   Publicity@xxxxxxxxxxxxxxxxxxxx
>                   http://lists.xenproject.org/cgi-bin/mailman/listinfo/publicity
>             _______________________________________________
>             Publicity mailing list
>             Publicity@xxxxxxxxxxxxxxxxxxxx
>             http://lists.xenproject.org/cgi-bin/mailman/listinfo/publicity
>       _______________________________________________
>       Publicity mailing list
>       Publicity@xxxxxxxxxxxxxxxxxxxx
>       http://lists.xenproject.org/cgi-bin/mailman/listinfo/publicity
> --
> Sarah Conway
> PR Manager
> The Linux Foundation
> sconway@xxxxxxxxxxxxxxxxxxx
> (978) 578-5300  Cell
> Skype:  sarah.k.conway

Sarah Conway
PR Manager
The Linux Foundation
(978) 578-5300  Cell
Skype:  sarah.k.conway

Publicity mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.