Outburst: VMware Networking Blog: vSphere 5 New Networking Features – Score D Minus – Must Improv

The VMware networking blog talks about the “new” networking features in vSphere 5. Well, talk would be overstating it. Mention, maybe. Post It Note, perhaps.

In fact, you could probably burp out the list on a single beer.

There are two broad types of networking capabilities that are new or enhanced in the VMware vSphere 5 release. The first type improves the network administrator’s ability to monitor and troubleshoot virtual infrastructure traffic by introducing features such as

  • NetFlow V5
  • Port mirror (SPAN)

The following are the key NIOC enhancements:

  • Ability to create User-defined resource pool
  • Support for vSphere replication traffic type
  • Support for IEEE 802.1p tagging

Colour me dubious but a 300% price hike in vSphere licensing for this ? They are definitely taking the mickey. Even with the backdown that delays the price hike to a couple of years in the future (when everyone has forgotten about it), we still don’t have proper networking in VMware.

vSphere Report Card: D-Minus. Shows no sign of improvement in Networking Class.

VMware: VMware Networking Blog: vSphere 5 New Networking Features – Introduction: “”

  • Benjamin Story

    Of course Cisco does still have a vested interest in EMC/VMWare and happens to sell a virtual switch (Nexus 1k) that improves VMWare’s networking capabilities.  I wonder why VMWare isn’t quick to improve their vSwitch? 

  • http://blogstu.wordpress.com stu

    There is also Open FCoE (I wrote about here: http://wikibon.org/wiki/v/Ethernet_Giants_Intel_and_Broadcom_Make_FCoE_Moves), but since you’re not a fan of FC/FCoE, I don’t expect that it will raise your grade.
    Last year around VMworld, there was concern from many that VMware was trying to take over too much of the networking ecosystem, so maybe less is more :)

    • http://etherealmind.com Etherealmind

      VMware implements someone else free code ? Wow, impressive. Thats worth lots of extra licensing fees.

  • http://pktmaniac.info Yandy Ramirez

    I would much rather have Cisco/Juniper/HP/Arista… and so on, do the networking for VMware than, VMware themselves. I’m only familiar with the Nexus 1000v, but that greatly improves on whatever VMware comes with, which by any other words is just junk.

    • http://etherealmind.com Etherealmind

      Which begs the question — why isn’t VMware setting up initiative that deliver better networking ? What is VMware doing about promoting IEEE 802.1Qbg/Qbh ? What about a statement of intent on OpenFlow ? Are they having another go at Software Networking using L3 Tunnels this time ?

      • http://pktmaniac.info Yandy Ramirez

        While I don’t know the answer to those questions, not exactly sure who does. It may be a better idea for VMware to work with networking vendors and create the framework/api for those things, rather than for them to try and implement. I’m not a VMware defender, just a networking guy that would like to see things improve, but with more integration from true networking companies.

      • Art Fewell

        Do you think qbg/qbh is the right path to go down anyway? One popular topic at ONS has been that the networking industry must move from the way it standardizes and interoperates today to a model more like software. QBH/QBG/BR seem to be based in the notion that 1. We need something other than CPU to improve I/O performance (Some at VMware debate this, also networkheresy has a lot of info that would indicate otherwise), or 2. We need these to preserve operational silos, the networking group must maintain control over I/O. 

        Moving towards these hardware-centric technologies pose a LOT of challenges for workload portability … of course from a hybrid cloud perspective there is no good way to address nic policy portability today IF you go with a hardware-centric approach. But if you leave policy not contingent on hardware it is completely portable. I would tend to think extending on things like the netqueue framework so that a NIC can provide technology agnostic hardware offload from the CPU to the NIC. QBH/BR and QBG will continue to face complexities as different vendors support different aspects and have different focus. If these become more popular, they will get more complicated as we try to extend into multi-vendor environments as well as hybrid and community clouds. Not to mention they will only work on complete greenfield hardware with the exception of a tiny percent of nics deployed today. 

        In my opinion, the access layer switch has always been the place where we filter, enforce and mark. As long as we can ensure performance works, this seems to me to be clearly the logical place to keep these types of policies. This way the server admin absorbs the VM Specific policies and they reside in the software layer where they are more manipulatable by vcenter or other orchestration frameworks. The Physical network is no longer the access layer, allowing the physical network to focus on implementing a security and performance framework that enforces the markings that are instantiated by the application or hypervisor switch. 

        • Art Fewell

          That is until we move towards controller-centric architectures where we define policy through a central point and both the vswitch and the physical network absorbs policy from the same central point. 

  • Mrs. Y.

    Isn’t the current version of Netflow V9?! V5 doesn’t even support IPv6. They couldn’t bother to do V9/IPFIX? Bad, bad VMWare.

  • Chris Young

    Hey Greg,

    I actually can understand to some extent the licensing hike. Just 2 years ago, we were talking 10:1 consolidation ratio, which means for 100 servers, that’s 10 license units ( more or less). With the new ultra high density platforms coming out with 2tb of ram, it’s nit unheard of for the ratios go up to 50:1, which would net VMWare just two licenses units. I can understand the linkage to how much ram a system has since that’s a pretty good correlation to how many servers will be virtualized on that host, not perfect, but at least understandable.

    On the other hand, this upgrade reminds me a little of windows 98 when it comes to the networking features. It’s not better, it just sucks less. :)

    On the bright side, we did get net flow back finally. Even if it is v5. :)

    • http://etherealmind.com Etherealmind

      Couple of things. One, there is no constraint in supply of software, therefore attempting to charge like it’s a limited commodity doesn’t make sense. The customer can make as many copies as needed — unless the developer is trying to exploit the market position. Which is fair enough in capitalist context……

      But competitors have already developed other software and are just waiting to eat your lunch. Given that there are dozens of hypervisors out there, it’s _not_ very hard to make one. Ok, so we have KVM, Xen, Hyper-V, Virtual Box, Oracle, IBM LPAR, QEMU ….. the list goes on and on. And many of these are free, as in open source, or have much better pricing models.

      VMware missed the point that their product is not unique, and they exist in a commodity market. They have a leading position by being first to market but that is already being eroded by other platforms such as OpenStack.

      • Chris Young

        Hi Greg,

        Sorry for not been clearer, I guess my point was that with the increase in server resources, VM is losing the licenses they would have sold for a customer to provide the same amount of virtualized hosts. It’s understandable to me from that perspective, they are watching their revenue stream get edited and are trying to protect it.

        For the record; not my lunch. :)

        I do agree that they missed the point and there are a lot of other optins out there and I hope e industry as a whole starts to embrace them. At least customers will be looking at alternatives given that the wizard of Gartner told them it was safe to do so (citrix ms & vm all in MQ)

        Whenever a single vendor has control over the whole market segment, it’s bad for everyone.

  • Ian Bowers

    Maybe I’m short sighted…  but I have a couple sad old 2600s at home in my lab.  They support netflow too.  And version 9 at that.  I have a trooper of a 2950, the very first piece of cisco gear I acquired back in the day, and it SPANs like a pro.  Even the NM-16ESW in one of the old 2600s does SPAN.  These features aren’t exactly new to networking gear.  It’s like if I were to say to my wife “And for my next trick I will upgrade to Ian V3.0…  I can now make breakfast and take out the trash!”.  You know what she’d say?  I should have been doing that in the first place, and the fact that she overlooked it and went with me anyways doesn’t mean she’ll be impressed when I come out with those new features.

    I don’t know a ton about the pricing hikes, but it reminds me a lot of how the music industry treats music.  Humans got clever, we built some internets, and now music is just bits.  There’s no longer a physical resource attached to the music…  no CDs, no materials to ship and press, etc.  but they feel entitled to get the same amount per album/track that they always did.  It sounds like VMware is trying to do something similar, where they’re penalizing humans for advancing in hardware and being more clever with how and where they run the VMs.  They were making essentially X per VM on average, and they feel entitled to make at least that much (preferably more) now.  

  • Leo Song

    The future of 1000v isn’t clear either.

  • Pingback: Top Five Things About VXLAN – And why it’s full of Fail — My Etherealmind()