Show 26 – Virtual Connect – The Conversion ?

This week we are joined by Ken Henault from HP. Ken is a leader in the VirtualConnect product from HP. In the past, Greg has ranted about the the VirtualConnect product and how crap it looks so we got in touch to ‘go at it’ and see if he can be convinced that VC has value.

How did it go ? ….. you be the judge.

Thanks also to the people who tweeted their questions during the recording of the podcast.

  • Why did HP Create Virtual Connect? ( define the business problems)
  • Physical – Too many Cable or Too many switches
  • Organizational – 3 teams managing one blade enclosure
  • Agility & resiliency – Separate workload from physical server

mezzanine card = blade

What is Virtual Connect? What’s needed to use it?

  • two minutes on virtual network adapters and how it impacts server and network design.

History of Virtual Connect

  • 1Gb
  • 10Gb w/Flex-10
  • FCoE – FlexFabric

What’s Good About Virtual Connect ?

  • Fit into any environment
  • Reduce cables without adding “switches”
  • Enable stateless servers
  • Bridge between Server team & LAN/SAN teams

What’s the difference between Flex-10, FlexFabric, etc etc.

How to understand the differences, because I find the plethora of products to be confusing and a lot of the terminology drives me mad.

Links to relevant HP documenation

Virtual Connect Homepage

Virtual Connect FlexFabric homepage

HP BladeSystem Technical resources
Link to the “Virtual Connect for Dummies”
Virtual Connect for the Cisco Network Administrator
Virtual Connect Ethernet CookBook
Eye on Blades blog

In addition to full iSCSI offload, the Virtual Connect FlexFabric module supports and manages iSCSI boot. The iSCSI boot parameters can be managed as part of the Virtual Connect server profile, creating a stateless server using a cost effective iSCSI boot device. This capability was only available on Native Fibre Channel in previous versions of Virtual Connect, now it’s available on all storage protocols supported by the Virtual Connect FlexFabric module; Fibre Channel, FCoE and iSCSI. Much as I’m confused by Virtual Connect for its complexity and stupid design parameters, at least they have delivered a viable iSCSI solution for blade / chassis servers. I’ll be taking a closer look at this.

Feedback

Follow the Packet Pushers on Twitter (@packetpushers | Greg @etherealmind | Ethan @ecbanks), and send your queries & comments about the show to [email protected].  We want to hear from you!

Subscribe in iTunes and RSS

You can subscribe to Packet Pushers in iTunes by clicking on the logo here.

Media Player and MP3 Download

You can subscribe to the RSS feed or head over to the Packet Pushers website to download the MP3 file directly from the blog post for that episode. Also, subscription options for Zune, Boxee and a range of other podcatchers are listed on the website.

Greg Ferro
Greg Ferro is a Network Engineer/Architect, mostly focussed on Data Centre, Security Infrastructure, and recently Virtualization. He has over 20 years in IT, in wide range of employers working as a freelance consultant including Finance, Service Providers and Online Companies. He is CCIE#6920 and has a few ideas about the world, but not enough to really count. He is a host on the Packet Pushers Podcast, blogger at EtherealMind.com and on Twitter @etherealmind and Google Plus.
Greg Ferro
Greg Ferro
Greg Ferro

Latest posts by Greg Ferro (see all)

  • http://www.workingfrommyshed.co.uk Stuart Howlette

    Hey, I was mentioned again!

    In response to the answer about how you'd rather be doing the network management, looking at future technologies and the like, all I can say is, so would I, but I aint being given the opportunity in my current role. A lot of my interaction with the network (and I'm guessing a lot of other people) is adminning up switchports and assigning VLANs, I've not been in the industry long enough to be doing the fun stuff.

    It's not just about protecting my domain, its protecting my job role :(

  • http://blog.michaelfmcnamara.com Michael McNamara

    I thought I would provide some real world feedback around HP's Virtual Connect and your discussion points.

    With regard to your cabling comment.. the cabling issues are real… I've had multiple pass-thru modules on the older P-class HP enclosures and the cabling created all sorts of issues including air flow restrictions. If you place 3 or 4 enclosures in a single rack you'll need every bit of cooling and air flow.

    With regard to the pass-thru modules I believe they only support 10/100Mbps (I'm not sure if that has changed since the older P-class) so I wouldn't think they are really an option in this day and age. Last year we deployed the HP Virtual Connect Flex/1 for a small VMware cluster and then this past May we deployed the HP Virtual Connect Flex/10 (two interconnects – an A side and B side). We uplinked the HP Virtual Connect Flex/1 interconnects independently via a LAG/LACP bundle to two Avaya ERS 8600 switches running IST/SMLT. We uplinked the HP Virtual Connect Flex/10 interconnects independently via a LAG/LACP bundle to two Cisco Nexus 7010 switches running vPC. So you can run NIC teaming to your blades but they can't be LAG/LACP bundles running active/active, they must run active/passive in the traditional sense.

    The real drawback with Virtual Connect is that you can't create a LAG/LACP bundle across interconnects, the endpoints much terminate on a single interconnect (switch). So you can't utilize multiple VC interconnects in an active/active configuration, you must settle for an active/passive uplink configuration between two interconnects – the A side and B side.

    In short the HP Virtual Connect was worked well for us in our VM enclosures and provides use 10Gbps to all 16 blades in our VM cluster.

    Cheers!

  • Paul Hyde

    Not being able to LACP bundle accross interconnects is the same for all interconnects not just Flex-10. The way round this from my experience is to define two of everything from shared uplink sets to networks this gives you redundancy all the way too you switch core.

    I have installed Flex modules of different types since they were introduced and they are great to deploy and maintain.

    • http://blog.michaelfmcnamara.com Michael McNamara

      You can certainly do that Paul… the problem then is that every broadcast packet will be received on each physical server NIC (every server will seen the broadcast frame twice). I actually have a configuration running just like that right now… leaves me uneasy to see all those duplicate broadcast packets. I'm curious if it's possible that 16 blades could push more than 20Gbps of bandwidth out of an enclosure.

      Just as a side note, you can stack the Cisco 3120 interconnects and then create a LACP bundle across both interconnects. I'm currently using this configuration for my non-VM enclosures and it works great. The Cisco 3120 just behaves like a stackable 3750 so you can create port-channels from Te1/0/1 and Te2/0/1.

      Cheers!

  • Markku L.

    Note that there are some ridicuously low VLAN limits in Flex-10 if you have got used to normal Cisco DC switching. Prepare to run dedicated uplinks to your ESX servers and separate general-purpose uplinks for your other servers in the same chassis due to the VLAN restrictions. I don't even recall the details anymore. Please correct me if any post-2.12 firmware have raised the limits.

  • http://www.hp.com/go/virtualconnect Chris Lynch

    (Full Disclosure: I work for HP and am on the same team as Ken Henault)

    @Markku:

    Keep in mind that there are plenty of Cisco switches that are incapable of supporting greater than 512 VLANs per switch. Virtual Connect currently has a limitation of 320 per physical module; 128 defined Ethernet Networks/VLANs per Shared Uplink Set (802.1Q trunk); 28 Mapped Networks per Network Connection. If you require additional VLAN support, it is recommended to use Tunnelling instead of Mapped. HP continues to improve the capabilities of Virtual Connect with customer input. Please provide any of your comments or input to your HP Sales Team.

  • Anders M. Jør

    (Full disclosure: works for a cisco partner)

    I think this excellent podcast illustrate an important problem related to Virtual Connect from my (somewhat biased) point-of-view, namely unclear responsibilities.

    Despite HP marketing tries to position Virtual Connect as an advanced network adapter, VC is so network centric you need alignment with the network team.
    (Shared vs Tunneled, VLAN limitations, OS support for Smartlink,. missing LACP across modules, QoS, etc).

    The Network team, however, does not like VC since it is not a 'proper' networking device and the server team has.. lets say… a slightly different focus.

    In other words; the network team ends up configuring the VC adapter, they didn't want in the first place. And even worse will have to troubleshoot the adapter.

    Ethan stated that few times and tried to get a discussing rolling but did not get an answer.

    In contrast; using regular blades switches or pass-through modules provide a clean cut.

  • http://www.mplsvpn.info shivlu jain

    I like the comment of "Video Poadcast"…Hope you are planning to start that also…

    • http://packetpushers.net Ethan Banks

      LOL. Have you SEEN us? Faces made for podcasting, just sayin'. Seriously, would be pretty tough to do a video podcast. Greg and I live thousands of miles apart, and the guests are scattered all over the world, literally. Video is a whole different editing world as well. Editing audio is plenty challenging enough, and that alone eats a good bit of Greg's time…I've produced a few of the shows, but Greg does the vast majority. I can't imagine what video would do the production timescale, never mind the bandwidth requirements…

7ads6x98y