The practice saves energy and this cannot be refuted. However, many think that frequently starting and stopping hard drives causes more wear and tear than leaving them spun up continuously.
Personally, my server is always on and the drives are not spun down. I have drives in it that have been continuously spinning for more than eight years, minus perhaps a few hours at most of off time.
The practice saves energy and this cannot be refuted. However, many think that frequently starting and stopping hard drives causes more wear and tear than leaving them spun up continuously.
Personally, my server is always on and the drives are not spun down. I have drives in it that have been continuously spinning for more than eight years, minus perhaps a few hours at most of off time.
I don't spin mine down either. I do have drives that have been spinning for a number of years too, but as a rule I aim for replacement every 5 to 6 years, but sometimes it is a bit longer. I just replaced a pair of 3TB with a pair of 10TB, and I would estimate those 3TB dives were 6 or 7 years old. the 3TB still work but will now be used for dead storage in an HDD storage toolbox type of case.
Interesting... With these HPE Proliant's, I am shutting them down every night to "save energy" but am I doing more damage in the long run, ya think?
Putting on my electronics technician's hat, as a general rule for solder joints, repeated heating and cooling can cause expansion and contraction that will, over time, cause the solder to crack and form a bad solder joint. In general electronics, this is usually most common in areas that run higher voltages, like power supplies, but as you are probably aware by now, computers in general do produce a lot of heat, so this rule can apply to most computer components.
On the flip side, a spinning disk has a spindle/platter motor. In most motors, the bushings or bearings can wear faster from continuous spinning but HDD spindle/platter motors (unless they are really really old), use what's known as a fluid bearing where the spindle is surrounded by a thin layer of oil that allows it to spin freely without excess friction, there is very little excess room for the spindle to tilt, and the gyroscope effect of the spinning platter aids to keep it from tilting, so the worry is not so much the bearing wearing out, but the heating and cooling cycles can still mess with the electronics boards.
Well, I spent a fortune on one of these so I suppose I would rather pay the higher electric bill then have drives and solder joints go bad.
My system is built on an amd 5600G cpu, no pcie gpu, 1 x 128GB ssd for the OS (smallest I could find), 8 x 5400rpm NAS HDD for data, 5 slow moving fans, and the cpu fan. I used an 80+ gold or platinum supply for extra efficiency
Average CPU loads range between 9% to 22%
Estimated power consumption is about 230W maximum, and about a 180W expected average.
This equates to it costing me about $2.00 to $3.00 a month ($22 to $32 a year) based on our electric rates here.
If you want to save power, a low power system is the answer, not turning them on and off.
If you want to estimate your usage/cost here is a good calculator:
If you want to save power, a low power system is the answer, not turning them on and off.
That is the key. Totally agree with this.
Personally I prefer not to worry about those €2 per month. A hard drive costs much more than that.
Putting on my electronics technician's hat, as a general rule for solder joints, repeated heating and cooling can cause expansion and contraction that will, over time, cause the solder to crack and form a bad solder joint. In general electronics, this is usually most common in areas that run higher voltages, like power supplies, but as you are probably aware by now, computers in general do produce a lot of heat, so this rule can apply to most computer components.
That was the FIRST thing I was told back in 1984 when I was at the SUN Microsystems Natwork Administration School. Computers have two modes that they like: Permanent OFF or permanent ON. Thermal stress is the main cause of computer breakdowns.
I'm in the group with letting drives spin. The greatest stress on any electric motor is during start up. The current surge on startup (very high compared to simply letting the drive spin) offsets the energy savings of times where the spindle is not turning. Depending on how many times drives are started, in a 24 hour period, there may be no savings at all.
Don't think you have to access data to spin up drives either. There are a number of processes (like SMART tests, rsync, etc.) that will spin up a drive without user intervention.
I'd say: it depends. I have a low power NAS in the office and when I am longer than 6 hours not using them the harddrives do spin down. I also have a SSD in the NAS as principal hdd therefore it happens that the disks are not spun up sometimes even for two weeks.
Some of my disks are older than five years and work flawless. At least from the specs of the disk wear leveling should not be an issue but I do know that a lot of people do let their disks spinning even when not accessed.
I'd say: it depends. I have a low power NAS in the office and when I am longer than 6 hours not using them the harddrives do spin down. I also have a SSD in the NAS as principal hdd therefore it happens that the disks are not spun up sometimes even for two weeks.
Some of my disks are older than five years and work flawless. At least from the specs of the disk wear leveling should not be an issue but I do know that a lot of people do let their disks spinning even when not accessed.
It would only make sense to stop the drives in a NAS that does not have access to the drives for a long time as you say. Actually, the only one who can know when a disk access occurs is the NAS administrator himself. So this decision must be made by each administrator.
In my case, the disks have continuous access during the day and night, during the day several local and remote clients access through different containers. A few scheduled tasks are executed overnight via containers or via the OMV GUI. So in my case it would be foolish to try to stop the disks.
It will also depend on the file system used. If you use a file system like EXT4 with mergerfs it may make more sense, since accessing a file means accessing a single disk. But if you use a file system like ZFS where the files are spread across multiple disks, accessing one file means accessing all disks.
All this means that this decision depends on many factors that each administrator must analyze.
Regardless of all this, in many specialized internet publications the recommendation is always to NEVER stop the drives.
I also have a SSD in the NAS as principal hdd
SSD's are in another category, when compared to spinning drives. In comparison, SSD's are super power efficient.
SSD's are in another category, when compared to spinning drives. In comparison, SSD's are super power efficient.
Yes, I just wanted to point out that sometimes it depends on the setup itself... that's all 
I'm by no means an expert at all, but I've not been able to find any research that draws a conclusion either for or against letting drives spin constantly. Obviously it seems to make sense that starting and stopping wouldn't be good for them and your spin down time would significantly impact any data points, but does anyone have any links to research on the topic?
For my use case, my drives will have 6-10 hours between activity so I do have them spin down and my snapraid parity drive is only in use for a few minutes per week so that's spun down pretty quickly. However, as mentioned by others, their disks might be in use constantly and need to have minimal latency when accessing so it makes sense to keep them running.
I've not been able to find any research that draws a conclusion either for or against letting drives spin constantly.
I've looked for data as well and didn't find anything. In speculating on the reasons why; it would take a LOT of hardware and testing several different scenarios. 5.25 versus 2.5" drives, 5400rpm versus 7200rpm [some enterprise drives run at 10,000 and 15,000rpm], hybrid drives [with a small SSD cache] versus standard spinning drives, SMR drives versus PMR, etc., etc.
All of these hardware differences and various OEM implementations affect power consumption. With all the factors previously mentioned, we have yet to look at use cases such as a commercial server, versus a small business server, versus a home server.
Conclusions from testing, that I can envision, still wouldn't be empirical. They'd be anecdotal at best. Costs would be prohibitive, especially for a Computer Science thesis and the commercial world is likely to just let drives spin.
So,, I'm guessing no one is willing to take it on.
I have seen many drives in the enterprise that are working fine for months and then the server is shutdown for maintenance. When it is powered back on, the drive fails to start. This is just one more reason that leads me to believe that drives will last longer with less startups.
Obviously it seems to make sense that starting and stopping wouldn't be good for them and your spin down time would significantly impact any data points, but does anyone have any links to research on the topic?
Any one that administered network servers, "mission critical" data will tell you that stopping, spinning down HD`is something that you do not do.
Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!
