BladeCenter basics

Question

Every manufacturer seems to have different standards for what ships with their hardware; mounting rails, cable arms, power cables etc. I've got used to that for standard rack-mounted gear, but now I've been asked to spec out an IBM BladeCenter installation, and I'm not sure what parts I need.

Assuming an H model chassis and (say) 6 blades, maybe something like:

• BladeCenter H chassis
• Power cable (what kind?) x 2
• Power supply x 2
• HS22 x 6
• SAS HDD (Hot swap) x 12

Now some questions:

• Do I need a power distribution unit of some sort?
• Are the twin GbE connectors on each HS22 used as the interconnect between blades?
• Do I need to add a networking module for external connectivity?
• What about a management module?

Basically, I'm open to any suggestions that'll help me understand what I need to be buying.

Answer 1

The options can be a bit overwhelming to newcomers.

To start with a new H-series will come bundled with the following items:

• Chassis populated with media tray and blowers.
• PSUs (pack of two).
• Advanced Management Module (AMM).

Besides your actual blades you will need the following as a minimum:

• Connectivity - Ethernet

  Unless you have some very specific requirements from pre-existing switches then avoid Copper Passthru Modules (CPM). They aren't that much cheaper, can be fumblesome and you won't benefit from the blade simplicity. Go for an Ethernet Switch Module (ESM). This will provide internal Ethernet switching between all blades and a handful of external ports to take connectivity out of the chassis.

  Each blade's onboard NIC is hard-wired to an I/O module in the chassis, so if you wish to use both NICs then you'll need to purchase two ESMs. The choice is between Cisco or Nortel (aka Blade Networks). If you don't have any ties to Cisco then I'd thoroughly recommend the Nortels. If your requirements are simple then the L2/3 model will do fine.

• Power

  There are five choices of power cables. The most suitable here in the UK is the 3x 16A IEC (25R5785). For each power connector on the back of the chassis, the cable splits out, in the form of two IECs for blades (allow 14A each) and one IEC for blowers (allow 5.5A).

  You'll need to purchase two of these cables for redundancy and a C19 PDU. Ideally you should be running two PDUs from two separate power sources and then connect the Bladecenter in a fashion whereby the load of the IECs is both distributed and redundant.

  Be sure to spec your power requirements carefully. Our PDUs have ammeters built-in so that we can observe the load and they all terminate to 32A Ceeform feeds. IBM provide a free utility called Power Configurator which can help you calculate your power requirements.

All of the above assumes that you have 6 blades with no additional I/O requirements. You may then wish to consider the following:

• Redundant AMM

  There is space for two Management Modules in each chassis. When two are installed, they will run in a primary and secondary arrangement, so that the secondary takes over in the event of the primary's failure. It's always worth having two.

• PSUs

  All BladeCenters are divided into two power domains. With the two bundled PSUs you will be able to run blades in slots 1-7 and I/O modules in bays 1-4 and 7-10. If you wish to expand and run blades or I/O modules in the remaining spaces then you will need to purchase an additional pack of two PSUs. Thankfully there is only one 2900W rating to choose from for the H-series.

• Connectivity

  If you want to expand on the connectivity of your blades, such as additional NICs, FibreChannel or InfiniBand, then you can purchase expansion cards for the blades and I/O modules for the chassis.

Finally the IBM Redbook entitled "IBM BladeCenter Products and Technology" is absolutely essential reading. It contains details all the available options, compatibility matrices and detailed descriptions about I/O use.

Answer 2

As I recall it (correct me if I'm wrong cause it's been a while) you can add "cheap" copper pass-through modules if you don't want to take the networking module cost up-front. That way you can access the blade NICs from the center backside - but it kinda defeats the purpose with a bladecenter.

There should be a couple of different networking module options, cheaper and more expensive depending on whether you want - say Cisco manageable level 3 routing in them or not and so on, see the tech sheets list.

The chassis I'm used to hard-routed one blade NIC to one networking module bay and the other blade NIC to the other networking bay - so if you needed both NICs on a blade you needed two networking modules. You could also add another two networking modules (or FC modules and so on) and the corresponding module to a blade to get two additional interconnects of your choosing.

There are loads to regard, the power modules power specific blade and expansion module slots (first two included power modules power blade bay 1-7 and expansions bay 1-4 - to use blade bay 8-14 and expansion module 5-10 you need two additional power modules).

The center should come with at least one management module, otherwise you need to add one. The blowers are fun to play with (rebooting management module ;)

Most questions should be answered by your hardware partner or vendor imho as it gives you the chance to ask more intricate questions about interoperability, licensing and additional costs.

Here's a list of tech sheets for different modules and the center itself and this is the installation user guide which is fairly detailed.

_{(source: keenzo.com)}

About the power cords, they're most likely not what you're used to and may need some local help from your electrician - order the specific BladeCenter H power cables. They each need three "standard" UPS power outlets which powers different parts of the chassis (so some can be left unplugged if you've only filled half the center with blades as I recall). During startup or when the management module is rebooting/offline - the blowers are uncontrolled, creating a decently sized storm and drawing a lot of power.