HP c-Class Training

I’m in training this week in Canada for the HP c-Class BladeSystem.  That means two things:  1) I was unable to attend VMware TSX 2007 in Las Vegas (bummer!), and 2) blog posts may be a bit less frequent than normal.  (Of course, I don’t really have a “normal” blog posting routine, but you get the idea.)

The HP c-Class BladeSystem is HP’s latest blade offering, replacing the earlier p-Class enclosure and blades with a new enclosure (10U instead of 6U), new blades (both half-height and full-height blades, up to 16 half-height blades per enclosure), new interconnects (both Ethernet and SAN), and an entirely redesigned internal architecture.  In my opinion, it’s a pretty impressive offering.

With this new BladeSystem, HP offers the following:

  • Up to 16 half-height blades (the BL460c or BL465c) in 10U of rack space; each half-height blade can go up to dual socket/dual core configurations
  • Up to 8 full-height blades (the BL480c, the BL685c, or the BL860c) in 10U of rack space; the BL480c offers dual socket/dual core (and quad core) configurations; the BL685c offers quad socket/dual core configurations.  (The BL860c is an Itanium-based blade.)
  • Mix and match full-height/half-height blades within an enclosure as needs dictate (with a few caveats, of course)
  • A variety of interconnect options for both Ethernet (including a Cisco switch that runs IOS—great for guys like me who are familiar with IOS) and Fibre Channel (both Cisco MDS and Brocade SAN switches are available)
  • A variety of mezzanine expansion options that add capabilities to the blades, like additional NICs or Fibre Channel connectivity

With all these options and all the various ways these pieces can be put together, planning and deploying one or more of these enclosures really demands some professional services assistance.  That’s great for guys like me who work in the professional services arena, but it’s going to take some work to get customers to realize that this isn’t the typical server deployment.  Yes, the blade servers themselves are very similar to your typical rack-mount server, but when you combine the blade servers with the interconnects and the mezzanine cards—the sum of the whole becomes more complex than the individual parts.

It’s also going to place more of a burden on the sales team to properly position solutions based around the c-Class BladeSystem to customers.  In some ways, that will be a greater challenge than dealing with the customers.  (Those of you that work in technical pre-sales roles will appreciate that comment.)

This generation of the blade enclosures and the blade servers themselves also addresses one key concern I had with earlier generations and using them as a virtualization platform:  a lack of NICs.  With the c-Class, I can now provision quad socket/dual core AMD Opteron-based server blades (the BL685c) with up to 12 NICs (four onboard plus two quad-port NICs in the mezzanine slots), and still have room for Fibre Channel connectivity.  That’s pretty impressive and makes for quite a platform for VMware ESX Server, if you ask me.

I’ll post more information here as the class progresses.  If anyone else has any experience (good or bad) with the c-Class enclosures, I’d love to hear about it in the comments.

Tags: , ,

  1. Nico’s avatar

    I am currently using two c-Class enclosures populated with BL460C blades, no special connectivity options, only GbE passthrough.

    Overall it is a very nice piece of Hardware; easy to use and impressive fast. I was a little surprised the GbE passthrough is only operating at GbE switchports; but that’s by design.

    The two links are showing problems we had. The “fan redundancy lost” is not resolved yet. That’s a little bit nasty when occurring at night….

    http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c00871795

    http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c00805127

  2. slowe’s avatar

    Nico,

    Thanks for the links; there’s good information there.

    I’m curious–what led you to choose the GbE passthrough modules instead of the GbE2c or Cisco blade switches?

  3. Nico’s avatar

    Our network infrastructure is maintained by another divison an currently they favor another brand for ethernet switching rather than HP or Cisco; so the GbE passthroug was not my choice.

    If it had been my decision i had choosen the GbE2c; it supposed to fullfill all our needs. Also, compared to the GbE2c and its functionality the “dumb” passthroug modules are a little bit pricy.

  4. slowe’s avatar

    That seems to be the prevailing reason that people choose the passthrough module. Thanks for the feedback!

  5. vmzare’s avatar

    This was nice piece of article. Could you please let me know more details on the blade or website where I can. I’m still confused as to how many NICS each blades sees, since this becomes major design consideration from virtualization point of view.

  6. slowe’s avatar

    VMzare,

    The full-height BL685c blade (the blade equivalent of the DL585 rack-mounted server) has four onboard NICs and three expansion slots, known as mezzanine slots. We can mount two more quad-port NICs in two of the mezzanine slots for a total of 12 Gigabit Ethernet ports per blade. At eight blades per enclosure, that’s 96 NICs in 10U of rack space. With the third mezzanine slot, we can add a dual-port Fibre Channel HBA for SAN connectivity. Try this link for more information:

    http://h18004.www1.hp.com/products/servers/proliant-bl/c-class/685c/specifications.html

    That should give you the information you are seeking. Hope this helps!

  7. vmzare’s avatar

    Need less to say, it has cleared my doubt. Thanks slowe.
    Link is more close representation what are c-class benefits.

  8. Tom’s avatar

    The C-Class has been great for us so far. I’m just about to finish my VI3 rollout. We started with 2xBL 685C full height blades. They are sporting 2xAMD Opteron 2400 Dual Core and 16Gb RAM. We’re using Qlogic dual port HBA’s and HP Quad port NIC mezz’s. We’re using 4xHP GB Eth pass through and 2xHP 4gb fibre pass through. The backend storage is 3Par e200 InServ with about 4TB raw capacity. Just finished configuring my VCB proxy with ESX Ranger so completely no network needed for backups.

  9. slowe’s avatar

    Tom,

    I have to ask you the same question I asked Nico: why pass-through? Placing the Ethernet swtiching and/or the fabric switching in the enclosure seems, to me at least, the best way to go, yet you are the second person to comment on this article that is using pass-through. What decisions led to your choice of pass-through?

  10. Mark’s avatar

    Forget pass-thru or switches, the new virtual connect modules give you far more flexibility and reduce your cabling requirements, when you see a demo of a blade failover ‘between’ blade enclosures themselves, now that is cool.

  11. slowe’s avatar

    Mark,

    I completely agree that pass-through is DEFINITELY not the way to go. I haven’t had the opportunity to real get any in-depth experience with the VirtualConnect modules yet, but I am impressed with what I have seen. In my view, one big advantage to using the switch modules is that organizations that have standardized on Cisco (or Nortel, although you’re not supposed to call it a Nortel) can continue to leverage the same knowledge and tools to manage the enclosure switches that they use to manage all the rest of the switching infrastructure. In some organizations, that’s a big deal.

  12. Brian Saunier’s avatar

    When you say there are a few caveats when mixing full height and half height blades, what exactly are the caveats?

    I am looking into buying the C-class chassis and we plan on having a few full height blades clustered for our SQL server and then some less expensive half height blades for other applications.

  13. slowe’s avatar

    Brian,

    The caveats really center around the mapping of the mezzanine cards to the interconnect bays, and understanding how the port mapping works. In addition, you’ll want to make sure you provision the interconnect bays properly with enough ports, or you could run short depending upon your mix of full-height and half-height. There aren’t any big show-stoppers, that’s for sure. You just need to be sure to do some diligent planning up front.

    There are some requirements for the half-height storage blades and their placement, but that may not be applicable in your environment.

    If you want or need more information, feel free to contact me directly.