I’m sure this will change as attitudes shift, but for now the biggest block I find myself facing is people, and the $ are too low to provide ammunition to force the issue.

I find this topic quite interesting – possibly for the wrong reasons. I find it interesting that folks seem to have such strong opinions on it(!) DPM is one of many features provided within vSphere that together provide a hugely powerful solution for customers. Is DPM in itself the be-all and end-all of datacentre efficiency? Of course not. Can it be useful in some/most scenarios? Absolutely. But the benefits will naturally vary according to use-case.

I completely accept that (particularly older) datacentre PDUs and HVAC systems are designed and optimised to run at full load, and therefore don’t offer linear savings when power draw is reduced. However, there’s no doubting that taking away the power consumption for devices (blades, servers, whatever) that are essentially idle for many hours at a time has to have a positive effect on the TCO for the majority of deployments.

There are some great use-cases I can think of just off the top of my head:
– Small to mid-sized dedicated Lab Manager clusters sized to cope with peak demand but often left under-utilised. The power of LM is in all the configurations/VMs you can hold – extremely efficiency – on shared storage without requiring dozens of racks of equipment. But much of the time they may not be powered on. I completely agree this would be better served as a resource pool within a larger multi-use cloud, but many orgs just aren’t there yet with this kind of model.
– SRM failover sites which may host small non-essential workloads but are sized to cope with a full site failover. Are you really telling me it’s not beneficial to have, say, 5 out of 8 blades in each cluster shut down if they’re not being used?
– VDI clusters used by 8-6 workers that aren’t utilised during the night. (again, I fully agree that it’d be great if a use could be found for these compute nodes during the night – but most orgs just aren’t there yet – so why pay for the leccy?)

As I say, it’s not the be-all and end-all. It’s not as game-changing as HA or DRS and it’s not going to cut your IT spend by 15% or anything radical – but it all helps. We’re all used to TVs that are now designed to use 1W in standby rather than 15W… we have CPUs that enter power saving mode and hard drives that spin down after 5 mins of inactivity to save a few Watts here and there… etc, etc. I could provide dozens of examples of technologies that have a tiny power-saving effect in isolation, but together add up to a lot.

“Every little helps” as they say.

But does it really warrant arguing about? And a TCO calculator?! Really?! To me that would only ever make sense if DPM were a $500/CPU add-on feature.

“TCO Evidence” is where I have the biggest issue with DPM right now. At a high level, I question if the $ saved by turning off random servers in a rack while leaving most still running impacts the bottom line. You’re not adjusting HVAC that much (servers that get powered off will be randomly located so you can’t just “switch this row from 100% to 20% cooling power to maintain room temp”, you’re not spinning down any shared disk or additional peripherals.

So it comes down to the “hourly cost to run a server” * “the number you could power down”. Unless that number is pretty damn high, I’m not convinced it’s worth the effort to implement/manage. I think I’d rather focus on different green areas (even if it’s just pushing harder to achieve 100% virtualization).