Every Data Center has its unique requirements and (usually) set of challenges to address, but, at the same time, there are common requirements that just seem all too difficult to solve because of the deluge of information and equipment choices available.  Here are a couple I've found in the past year to be indispensable.

10 Gigabit Ethernet

Fibre Channel is easy to hate:  The equipment is largely proprietary, requires long training cycles, is perpetually outmoded, and is ridiculously expensive to upgrade (scaling up or scaling out), plus you have to deal with all that crazy fiber running around the data center to get to your SAN.  It has been a necessary evil as long as it's been around, commanding 3-4 times the per-GB/per-Host cost of direct-attached-storage with little competition from the paltry bandwidth limitation (1Gb) of iSCSI... until now.

Yes, friends, 10GbE has reached prime time, and I love it.  With a per-port cost that's one-fifth of Fibre and per-GB cost at one-third (or less), 10Gb (and much higher in a few years) iSCSI has surely numbered the oligopoly days of the Fibre vendors.  You don't need a highly-overpriced SAN controller any more (your favorite OS will do) and fan-out is limited only by IP addresses.

I highly recommend Neterion's Xframe family of 10GbE cards: they have drivers for virtually every OS (even Mac & Solaris) and have proven the most capable cards I've used.  They're also the only 10GbE cards I could find capable of teaming in Windows Server 2008 Hyper-V.

Don't forget to use CX-4 (copper) connectors to avoid paying the Fibre overhead if you don't need longer than 50' runs.

I use HP's 10GbE ProCurve equipment, notably the 5400zl line of modular switches:  Great bang for the buck.

The cost of one system-wide SAN controller upgrade can probably outfit (or retrofit) most fibre installs with a full 10GbE replacement.

Pervasive Redundant Power

Part of the IT job is continually mitigating single points of failure (SPoF) to ensure uptime.  Let's follow the path for power:

• Diverse street power. Check.
• Redundant generators. Check.
• Redundant UPSes. Check.
• Redundant power supplies in equipment.  Mostly.

The fact is that lots of equipment doesn't afford redundant power; this is especially true of "smaller" networking equipment or specialty one-offs (like environmental monitors)... usually all you get is one 120V (NEMA) connector.  I'd always struggled with this, since there had to be a better way; but, no matter who I asked, the answer was either that it was an acceptable risk or you bought two (which isn't always possible for the application).  Then, late last year and by happenstance, I found the answer: Server Technologies' Fail-Safe Transfer Switch.

The idea is so simple, it's obvious: Take all the failover wiring that goes into redundant equipment power and put it at the rack.  Ta da! Now one UPS can be serviced without requiring a macabre lesson in quick-plugging equipment to another UPS.  It's like magic and very inexpensive, especially for the peace-of-mind it affords.  They're available in a variety of input and output power configurations.

What about you?

That's it for today.  Do you have any must-have gadgetry that's made your life better?  If so, let me know.

This is just a quick post to help anyone else out there looking for the same information.  OEMs sometimes get creative with their WMI implementations for some reason, particularly when it comes to providing CIM_TemperatureSensor (which is pre-v2.2 CIM schema).

After much research and browsing WMI hierarchies, I've landed on a set of queries that properly gets the info from most servers (all in our environment).   YMMV.

Dell and HP both inherit the CIM_NumericSensor definition from the CIM (Common Information Model) 2.2+ schema

HP

CIM standard schema (query in \root\HPQ):
SELECT * FROM CIM_NumericSensor WHERE SensorType=2

HP's MOF extension (query in \root\HPQ):
SELECT * FROM HP_WinNumericSensor WHERE SensorType=2

Most HPs report for every sensor in the system; although the guide is specified for HP-UX, their definitions are available on their site on pages 3-6.

Oddball HP_Health extension to WMI (query in \root\WMI):
SELECT * FROM TempSensorInfo

Dell

Dell's MOF extension (query in \root\CIMV2\DELL):
SELECT * FROM CIM_NumericSensor WHERE SensorType=2

I can only seem to get "Ambient" temperature readings from the Dells via WMI, although they do provide more detailed information in DRAC.  Note also that their readings are multiplied by 10.