That is the case anytime a hypervisor provides a VM with a "local" block device
which is backed by network based storage ("VM itself does not need to connect
to network storage"). This is not something new or unique provided by so called
"cloud storage" such as Ceph.
Cheers,
Brad
-----Original Message-----
From: Aldrin Isaac [[email protected]<mailto:[email protected]>]
Sent: Sunday, August 26, 2012 07:45 PM Central Standard Time
To: Black, David
Cc: [email protected]; [email protected]
Subject: Re: [nvo3] Storage (part of: Let's refocus on real world)
AFAIK, scale out cloud storage software such as Ceph do not rely on FC, FCoE,
NFS or iSCSI on the VM. Ceph storage appears to the VM as local storage and
does not depend on network virtualization. VM migration is not an issue for
Ceph since the VM itself does not need to connect to a storage server over the
network. So as far as real cloud storage is concerned nothing is being swept
under the rug.
-- aldrin
On Sunday, August 26, 2012, Black, David wrote:
Robert,
> Also as you have pointed out storage discussion can not be just swapped
> under carpet and addressed by quote: "storage issues are out of the scope".
I agree ... and that looks like my cue to say something ... e.g., see the
domain part of my email address ;-).
iSCSI and NFS use TCP/IP in the storage stack and hence will run fine over
all of the data encapsulations being discussed here. If the iSCSI initiator
or NFS client is in the VM, that's most of the discussion. That's not always
the case for a number of reasons - the obvious one is that a hypervisor
iSCSI initiator or NFS client is required if the VM's executable image is being
loaded and/or paged using one of those protocols. It's also the case that
many hypervisors simplify the storage interface presented to VMs so that it
looks like direct attached or internal disk drives), and map those disks to
networked storage using a hypervisor iSCSI initiator or NFS client. Ensuring
that the VM migration destination hypervisor has appropriate connectivity to
storage is mostly a configuration concern. The upshot is that iSCSI and NFS
run fine over nvo3-encapsulated networks.
In contrast, as I said at the microphone at the nvo3 BOF in Paris, I suggest
that the WG not initially consider FCoE, in order to defer spending time on
discussing how to deliver DCB Ethernet service/behavior (required by FCoE -
ordinary non-DCB Ethernet isn't sufficient for FCoE because FCoE is *very*
sensitive to drops) through the encapsulation(s).
Thanks,
--David
> -----Original Message-----
> From: [email protected]<javascript:;>
> [mailto:[email protected]<javascript:;>] On Behalf Of Robert
> Raszuk
> Sent: Saturday, August 25, 2012 11:55 AM
> To: Ivan Pepelnjak
> Cc: Black, David; [email protected]<javascript:;>; Linda Dunbar
> Subject: Re: [nvo3] Let's refocus on real world
>
> Ivan,
>
> > ... or I may be completely wrong.
>
> I think you are actually right on the spot correct.
>
> However I am afraid authors of this document are not likely to admit
> that TOR switches should be just basic IP nodes providing only transport
> between servers.
>
> Likewise they will not likely to admit that all logic of encapsulation
> should happen on the hyper-visors as they are simply not in that
> technology space.
>
> Similarly I very much agree and support providing clear distinction
> between "cold" and "hot" VM mobility cases and perhaps even further
> provide number of sub-classes hot VM mobility can be accomplish today -
> clearly there is more then one way.
>
> Also as you have pointed out storage discussion can not be just swapped
> under carpet and addressed by quote: "storage issues are out of the scope".
>
> While Linda was perhaps right to say that today most storage today is
> coming to servers via backend this is what I would call very inefficient
> and legacy way. If we are to think ahead one needs to observe how the
> industry advances in storage virtualization via front-end IP very often
> not co-located with the compute racks.
>
> In my view network related mobility discussion is not about TOR or about
> VLANs. It is about an IP layer above IP transport which would carry all
> necessary information of the actual location of the VMs and which in
> fact would play the main role in shortening or eliminating the
> triangular routing problem.
>
> Rgs,
> R.
>
>
>
> > On 8/24/12 11:11 PM, Linda Dunbar wrote:
> > [...]
> >
> >> But most, if not all, data centers today don't have the Hypervisors
> >> which can encapsulate the NVo3 defined header. The deployment to all
> > > 100% NVo3 header based servers won't happen overnight. One thing for
> > > sure that you will see data centers with mixed types of servers for
> > > very long time.
> >>
> >> If NVEs are in the ToR, you will see mixed scenario of blade servers,
> >> servers with simple virtual switches, or even IEEE802.1Qbg's VEPA. So
> >> it is necessary for NVo3 to deal with the "L2 Site" defined in this
> >> draft.
> >
> > There are two hypothetical ways of implementing NVO3: existing layer-2
> > technologies (with well-known scaling properties that prompted the
> > creation of NVO3 working group) or something-over-IP encapsulation.
> >
> > I might be myopic, but from what I see most data centers today (at least
> > based on market shares of individual vendors) don't have ToR switches
> > that would be able to encapsulate MAC frames or IP datagrams in UDP, GRE
> > or MPLS envelopes. I am not familiar enough with the commonly used
> > merchant silicon hardware to understand whether that's a software or
> > hardware limitation. In any case, I wouldn't expect switch vendors to
> > roll out NVO3-like something-over-IP solutions any time soon.
> >
> > On the hypervisor front, VXLAN is shipping for months, NVGRE is included
> > in the next version of Hyper-V and MAC-over-GRE is available (with Open
> > vSwitch) for both KVM and Xen. Open vSwitch is also part of standard
> > Linux kernel distribution and thus available to any other Linux-based
> > hypervisor product.
> >
> > So: all major hypervisors have MAC-over-IP solutions, each one using a
> > proprietary encapsulation because there's no standard way of doing it,
> > and yet we're spending time discussing and documenting the history of
> > evolution of virtual networking. Maybe we should be a bit more
> > forward-looking, acknowledge the world has changed, and come up with a
> > relevant hypervisor-based solution.
> >
> > Furthermore, performing something-in-IP encapsulation in the hypervisors
> > greatly simplifies the data center network, removes the need for
> > bridging (each ToR switch can be a L3 switch) and all associated
> > bridging kludges (including large-scale bridging solutions). Maybe we
> > should remember that "Perfection is achieved, not when there is nothing
> > more to add, but when there is nothing left to take away" along with a
> > few lessons from RFC 3439.
> >
> > I am positive a decade from now we'll see ancient servers still using
> > VLAN-only hypervisor switches (or untagged interfaces), so there might
> > definitely be an need for an NVO3-to-VLAN gateway, but we shouldn't
> > continuously focus our efforts on something that's probably going to be
> > a rare corner case a few years from now.
> >
> > ... or I may be completely wrong. Wouldn't be the first time.
> > Ivan
> > _______________________________________________
> > nvo3 mailing list
> > [email protected]<javascript:;>
> > https://www.ietf.org/mailman/listinfo/nvo3
> >
> >
>
> _______________________________________________
> nvo3 mailing list
> [email protected]<javascript:;>
> https://www.ietf.org/mailman/listinfo/nvo3
_______________________________________________
nvo3 mailing list
[email protected]<javascript:;>
https://www.ietf.org/mailman/listinfo/nvo3
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3
