VMware buys Nicira: a Hypervisor Vendor Woke Up

Almost a year ago, I predicted that eventually the hypervisor vendors will wake up and realize it’s time to get rid of VLANs and decouple virtual networks from the physical world. We’ve got the first glimpse of the brave new world a few weeks after that post was published with the VXLAN launch, but that was still a Cisco’s solution running on top of VMware’s (and now everyone else’s) hypervisor. The recent VMware’s acquisition of Nicira proves that VMware finally woke up big time.

Why is this important?

A few days ago, Nicira was just another SDN startup (although one of the rare ones with an actual shipping product). Those that understood their Network Virtualization Platform (NVP) loved it, but everyone considered NVP to be a solution targeted at Rackspace-sized cloud providers.

We don’t know what VMware’s plans are, but they just might incorporate Open vSwitch into vSphere kernel, make NVP part of vCenter (or vCloud Director) and offer the whole solution as part of one of the standard products. Assuming VMware also buys a feature-rich high-performance VM-based L4-7 appliance vendor (vShield Edge is still a joke), you’d have a killer combo that would need nothing more than simple IP-based network transport to create billions of virtual networks.

Alternatively, I’m totally wrong and VMware might decide to use NVP as part of a multi-hypervisor orchestration tool. They have most of the components to do that.


MAC-over-IP virtual networking has arrived

Anyway, the times of large-scale VLANs are definitely coming to an end. We don’t know when the MAC-over-IP encapsulation will become part of a mainstream vSphere product (it will definitely be part of Hyper-V in Windows 2012 server), but that time will arrive sooner rather than later. TRILL (and related technologies) is truly nothing more than a tactical solution.

What this is not

As one would expect, some people view the VMware’s acquisition as validation of Software Defined Networking (SDN) approach. It might definitely look that way to a VC, but if you forget the marketing games, NVP’s architecture is not much different than Cisco’s Nexus 1000V or VMware’s VDS. They all use central control/management plane to download forwarding and policy information into distributed semi-autonomous soft switches. The “only” difference is that Nicira's architecture actually scales due to a number of design and implementation choices they made.

Other fine blog posts

You might want to read these great articles written by fellow bloggers:

More information

If you’re new to virtualized networking and would like to understand what this is all about, start with the Introduction to virtualized networking webinar. Various virtual networking technologies are described in Cloud Computing Networking webinar (which now includes a 1,5-hour long section on IaaS scalability). You can buy individual recordings or get access to both webinars (and numerous others) with the yearly subscription.

Finally, if you’d like me to review your data center/cloud network design or discuss various technology options, ExpertExpress just might be the best option.

10 comments:

  1. War is coming...

    Nicira haven't made a dime yet and the Cisco-VMware combo made a lot of business i guess. I'm not buying into into the 'we'll keep being friends' i hear from VMware over and over since the deal went public. see how it plays out...good luck to all.

    technology-wise i still think (and don't see why not) SPBM/TRILL will gain traction as the underlying transport for MAC-over-IP (btw, keep track of the IETF's NVO3 efforts). to me, next thing is SDN/OF vs. Control-Plane for these MAC-over-IP implementations.

    ReplyDelete
    Replies
    1. > technology-wise i still think (and don't see why not) SPBM/TRILL
      > will gain traction as the underlying transport for MAC-over-IP

      Mac-over-IP does not need Layer2 fabric as an underlying transport. It works just fine over standard and inexpensive IP networks. Mac-over-IP obsoletes TRILL before it even had a chance to gain traction.

      Delete
    2. Glad I'm not the only one preaching this ;)

      Delete
    3. don't get me wrong. i do agree L3 is the way to go and scales best. i just argue we'll still need scalable L2 to some degree inside those subnets. STP/MCLAG/ETC won't cut it.

      Delete
    4. You don't need scalable L2 if every single ToR switch is a L3 switch ... and all those cheap pizza boxes with Broadcom chipset are.

      Delete
    5. but do you honestly think this is going to happen? managing this much subnets (on servers/switches/dhcp etc)?

      In a larger perspective, we are always trading configuration for efficiency. *Usually* the best result would be the average.

      Delete
    6. I would absolutely prefer the "pain" of managing many small IP subnets over the adrenaline thrill of deploying TRILL or SPBM. But maybe it's just me being old.

      Delete
  2. Looks like a direct swing at openstack/quantum. Especially considering how much they paid to keep these technologies away from other companies

    ReplyDelete
  3. It's actually much simpler to configure L3 tor's than to expand out vlan's everywhere. In particular, if you fully standardize configurations (vlans etc) automating tor config is hugely simplified.

    ReplyDelete
  4. So now that Nicira is part of VMware, will we see OVS in ESXi. Nicira has been preaching Openflow for the last 5 years or so. It is only interesting to see if they will make the ESXi vswitch an Openflow switch as well (and thus enable an ecosystem of network controllers).

    ReplyDelete

You don't have to log in to post a comment, but please do provide your real name/URL. Anonymous comments might get deleted.

Ivan Pepelnjak, CCIE#1354, is the chief technology advisor for NIL Data Communications. He has been designing and implementing large-scale data communications networks as well as teaching and writing books about advanced technologies since 1990. See his full profile, contact him or follow @ioshints on Twitter.