I think you'd need to identify specific reasons to choose hibernation over a regular system shutdown before it would make sense. Hibernation was introduced primarily for laptops, where being able to power down yet keep the user session state has significant value. But ideally servers should be able to shut down fully, then start up on a fresh boot without having lost anything of value.
An alternative approach to conserving power is the power efficiency of your servers. There are big potential gains here. Servers made by Dell, IBM and HP in the last 2 years give vastly greater power efficiency over their predecessors, because of a general shift to 'green' datacenter technologies. One of our HP G5 servers (2007) draws 400W under regular load to run a single 2008 instance, whereas one of our G6 servers (2009) is running 70 virtual machines and barely drawing 180W.
Another Alternative - If you're virtualized, you can also benefit from cluster power management solutions like Insight Power Manager. There are plugins available that can dynamically power on/off VM host servers to meet demand, so when you systems go quiet in the evening, a bunch of the servers that run them will shut down. When things pick up in the morning the servers will come back online. This is a much more elegant solution if you're looking to save power across more than a handful of machines.
On the subject of whether daily power cycles of your servers might be bad for the hardware - I couldn't find too much data on the subject but there are good reasons to suspect you will have problems. Disks tend to fail during spin-up, and it seems to be due to the material it's made of expanding and contracting under changing temperatures (powered on vs powered off). Repeat this expansion and contraction enough times and the drive fails. I suspect this is why laptop drives fail most regularly, not because they're moved around, but because they're frequently power-cycled, and tend to run closer to hot components.
The reliability issue is another reason why the virtualization solution may be a better fit - host servers can boot from an SD card and have no disks, therefore won't suffer from this problem.
Here's the G6 server I mentioned:
Here's the G5 (rebooted this evening, so only partial graph):
Not necessarily harmful, but huge performance speed-bump when awaking from hibernation. You might take a look at spinning down the hard drives or maybe even placing the system into Sleep as possible alternatives.
I think you'd need to identify specific reasons to choose hibernation over a regular system shutdown before it would make sense. Hibernation was introduced primarily for laptops, where being able to power down yet keep the user session state has significant value. But ideally servers should be able to shut down fully, then start up on a fresh boot without having lost anything of value.
An alternative approach to conserving power is the power efficiency of your servers. There are big potential gains here. Servers made by Dell, IBM and HP in the last 2 years give vastly greater power efficiency over their predecessors, because of a general shift to 'green' datacenter technologies. One of our HP G5 servers (2007) draws 400W under regular load to run a single 2008 instance, whereas one of our G6 servers (2009) is running 70 virtual machines and barely drawing 180W.
Another Alternative - If you're virtualized, you can also benefit from cluster power management solutions like Insight Power Manager. There are plugins available that can dynamically power on/off VM host servers to meet demand, so when you systems go quiet in the evening, a bunch of the servers that run them will shut down. When things pick up in the morning the servers will come back online. This is a much more elegant solution if you're looking to save power across more than a handful of machines.
On the subject of whether daily power cycles of your servers might be bad for the hardware - I couldn't find too much data on the subject but there are good reasons to suspect you will have problems. Disks tend to fail during spin-up, and it seems to be due to the material it's made of expanding and contracting under changing temperatures (powered on vs powered off). Repeat this expansion and contraction enough times and the drive fails. I suspect this is why laptop drives fail most regularly, not because they're moved around, but because they're frequently power-cycled, and tend to run closer to hot components.
The reliability issue is another reason why the virtualization solution may be a better fit - host servers can boot from an SD card and have no disks, therefore won't suffer from this problem.
Here's the G6 server I mentioned:
Here's the G5 (rebooted this evening, so only partial graph):
Not necessarily harmful, but huge performance speed-bump when awaking from hibernation. You might take a look at spinning down the hard drives or maybe even placing the system into Sleep as possible alternatives.