Then I think we're in semi-violent agreement, as I strongly concur that those softswitches are important NVE locations. This is captured in the narten problem statement draft, but if you have specific text changes to suggest, please do.
I do want to see protocol interfaces on both sides of these softswitch NVEs - VM-facing for attach/detach, etc. and "oracle"-facing for address mapping lookup. Designing these as protocol interfaces results in the same structure of functional interactions when the NVE is offboard on a ToR switch or the like. Thanks, --David > -----Original Message----- > From: [email protected] [mailto:[email protected]] On Behalf Of Robert > Raszuk > Sent: Tuesday, June 26, 2012 3:48 PM > To: Black, David > Cc: [email protected] > Subject: Re: [nvo3] Softswitches > > Hi David, > > Honestly I think we are getting more into implementation then > standardization debate. > > What I call a softswitch can be either part of the linux kernel entirely > as a loadable module, can be partially in the kernel and partially in > the user space or if you follow NETMAP's direct memory mapping from > interface to user space could sit completely there. > > Examples of softswitches which I am experimenting with is as you said > one from VMWare, the other most popular is OVS, there is LINK released > recently and there is at least few more in the works which would be > partially sitting in the end-system kernel or the kernel and user space. > > At least this is what I meant by using the term "embedded". > > The main point I think is that the L2/L3 virtualization/separation will > be happening in all of the above cases which share one very important > common characteristic - thay are all residing on the end host. > > Best regards, > R. > > > > Robert, > > > > We may be in semi-violent agreement ... > > > >> All of current running production data centers and those which I am > >> working on to be shortly deployed assume NVE (Network Virtualization > >> Edge) like functionality is embedded within the end-host. > > > > So far, so good. As you're using the terminology, the Nexus 1000V is > > software that deployed in the end-host. > > > > Nonetheless, the Nexus 1000V a logically separate networking component > > (network admin views it as a layer 2 switch that is easily distinguishable > > from the hypervisor in which it runs), and the same is true to a lesser > > extent of the native softswitches in at least the VMware and Hyper-V > > hypervisors (they tend to be more oriented towards server admin management). > > > >> But as one requirement I am very interested in making sure the real use > >> case of embedded softswitch in the end system is captured and documented > >> in the problem statement document. > > > > Depending on what you mean by "embedded", we may be in semi-violent > > agreement. > > > > Please define "embedded" as you're using the term. > > > > Thanks, > > --David > > > > > >> -----Original Message----- > >> From: [email protected] [mailto:[email protected]] On Behalf Of > Robert > >> Raszuk > >> Sent: Tuesday, June 26, 2012 3:24 PM > >> To: Black, David > >> Cc: [email protected]; [email protected] > >> Subject: Re: [nvo3] call for adoption: draft-narten-nvo3-overlay-problem- > >> statement-02 > >> > >> David, > >> > >> Just as a data point .. > >> > >> All of current running production data centers and those which I am > >> working on to be shortly deployed assume NVE (Network Virtualization > >> Edge) like functionality is embedded within the end-host. > >> > >> It can be a form of prioprietary softswitch, opensource softswitch or > >> any other flavor of kernel module acting as such softswitch. > >> > >> I am not stating everyone does the same .. some may use TOR as a > >> softswitch or even furthere though. > >> > >> But as one requirement I am very interested in making sure the real use > >> case of embedded softswitch in the end system is captured and documented > >> in the problem statement document. > >> > >> Thx, > >> R. > >> > >> > >>>> Actually I do not see that those two are need to be decoupled. Linux > >>>> kernel with some additional enhancement module can act as NVE. Tenants > >>>> are just connected to the NVE over normal bridge interfaces by the > >>>> co-located hypervisor. > >>> > >>> The salient question is not what's possible, but what's actually being > >>> done ("running code"), as nvo3 can't succeed if it requires major > >>> hypervisor rewrites. > >>> > >>> Linux is not the system that supports hypervisors. The Cisco Nexus > >>> 1000V softswitches are examples where there is a strong decoupling > >>> and that softswitch is available for at least two non-Linux hypervisors, > >>> suggesting similar decoupling in the system architectures for those > >>> hypervisors. > >>> > >>> Thanks, > >>> --David > >> > >> > >> _______________________________________________ > >> nvo3 mailing list > >> [email protected] > >> https://www.ietf.org/mailman/listinfo/nvo3 > > > > _______________________________________________ > > nvo3 mailing list > > [email protected] > > https://www.ietf.org/mailman/listinfo/nvo3 > > > > > > > _______________________________________________ > nvo3 mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/nvo3 _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
