Server Hard-drive

[frozenfire75i]

I got an IBM Netvista with a 40GB HDD in it (My Server), not vintage but still old and I wanna prolong it's life as long as I can, So I wondering what's better.. Leave it run 24/7 or set it so the HDD spins down after 3 mins of non use?

[DreadStorm]

I have mine spin down after 30 minutes.

[Chuck(G)]

I got an IBM Netvista with a 40GB HDD in it (My Server), not vintage but still old and I wanna prolong it's life as long as I can, So I wondering what's better.. Leave it run 24/7 or set it so the HDD spins down after 3 mins of non use?

Depends on what your duty cycle is. If it's 4 minutes typically between requests, then a spindown after 3 is going to do more harm than good.

I'm assuming that you have long periods of no access at all. Figure out how much inactivity you'll have before you are fairly sure that you're entering one of these periods and set the spindown time to that.

Of course, if this is a Netvista with a Deathstar drive, then all bets may be off... :)

[frozenfire75i]

Yes I do have long idle times I would assume, But I have heard it spin up a few times in the last few hours so maybe it gets used more then I think it dose!

It is an IBM Deskstar 40GB HDD, are those the bad ones ;-)




Depends on what your duty cycle is. If it's 4 minutes typically between requests, then a spindown after 3 is going to do more harm than good.

I'm assuming that you have long periods of no access at all. Figure out how much inactivity you'll have before you are fairly sure that you're entering one of these periods and set the spindown time to that.

Of course, if this is a Netvista with a Deathstar drive, then all bets may be off... :)

[barythrin]

There was an interesting article I think my dad actually pointed me to years back (quite unfortunately I can't recall where it was to back up the claim) but it said if you were going to turn the computer on again within 4 hours it was better (certainly couldn't be less electricity?) to leave it on instead of turn it off and back on afterwards.

I'm not 100% sure why it was, but just accepted it as making some sense as far as moving parts go and historically when I leave my new system off for a weekend (or unplugged) it tends to never turn on again (bad (crappy) power supply/caps).

[QuantumII]

Last year I heard about this issue at work;

One site was doing a shutdown of a NetApp diskshelf, which had been running 24/7 for a long time with no errors. When they applied power to it again, 3 of the disks did not spin back up because of wear.

Server hdd's are designed to run 24/7. Start-stop cycles is not always a good thing for the bearings and the heads.

[Chuck(G)]

Yes I do have long idle times I would assume, But I have heard it spin up a few times in the last few hours so maybe it gets used more then I think it dose!

It is an IBM Deskstar 40GB HDD, are those the bad ones ;-)

I'd leave it spinning then. If it's a server, spin-up time will have an effect on response time, no?

There's got to be some way of estimating the power used in spin-up and head retraction (to park) and the CPU sitting and waiting for the drive to come ready with the power used to keep the drive spinning instead. I don't know what the magic number is.

Note that when you're turning the drive spindle motor off, it isn't the same thing as turning the whole drive off. But, like an airplane taking off and landing, you do have to get those heads up in the air when you spin up again. You know what they say about the two most dangerous times when flying an airplane...
----------------------------
Google "deskstar deathstar" and draw your own conclusions.

[IBMMuseum]

Last year I heard about this issue at work;
...Server hdd's are designed to run 24/7. Start-stop cycles is not always a good thing for the bearings and the heads.

But this drive is IDE. A drive operated 24/7 (whether there is timespans that they spin down or not) will eventually fail. That is what RAID on a server (Netfinity instead of Netvista) is designed to counter.

[Chuck(G)]

Re: Deathstar drive. Check out this firmware upgrade from IBM (http://www-947.ibm.com/systems/support/supportsite.wss/docdisplay?brandind=5000008&lndocid=MIGR-60241).

[Lorne]

There was an interesting article I think my dad actually pointed me to years back (quite unfortunately I can't recall where it was to back up the claim) but it said if you were going to turn the computer on again within 4 hours it was better (certainly couldn't be less electricity?) to leave it on instead of turn it off and back on afterwards.

I'm not 100% sure why it was, but just accepted it as making some sense as far as moving parts go and historically when I leave my new system off for a weekend (or unplugged) it tends to never turn on again (bad (crappy) power supply/caps).


And I remeber way back when in the 80's, someone telling me that you were better off leaving the computer (PC XT compatible type) running all the time. I was told that the constant heating up and cooling down (= expansion & contraction) could be worse for "cold soldered" (?) joints, than just leaving it running all the time.
Then again, that may just have been a good story from someone selling hard drives at the time, and wanting to sell more after they wear out.
I left mine running 24/7, and a hard drive quit after about 18 months, which probably wasn't too bad with the hard drive technology that was available in the 80's.

[Yzzerdd]

I had read for PCs without a hard drive, it was best to leave them on 24/7 unless you wouldn't use it for 2 or more days. This was due to heating/cooling being bad on ICs, solder joints, and PSU. The article said if you have an external HDD drive, you should park it and turn it off when you are done for the day on the computer. It is bad on the bearings to leave it on 24/7. This may have changed with new technology, this article was written in the PC 5150 days when MFM ruled all the lands.

--Ryan

[frozenfire75i]

drifting off subject a bit,But while we are on the on the subject of heating and cooling and such. I only turn my 5150 once a month or so such to both keep it alive and in running order and minimize the effect's of heating and cooling. all thought I might try to space it out even more to two months or so.

My room stays at 70 degrees all the time, so I would think when it dose heat and cool it don't that much or dose that matter?

On the subject of the server HDD, I think I might just clone the Deathstar HDD to one of my 20GB HDD's and leave it run for 24/7 as I have a whole box full of the 20GB's.

[Chuckster_in_Jax]

Actually, there is a split on which is harder on the electronics. A problem that hasn't been discussed yet is the amount of current that rushes through circuits when they are first powered up due to the capacitors charging. Capacitors take a brief period of time to charge up. While they are charging, the current they draw is very high and they essentially act like a short circuit. When they are fully charged the current drawn by them is negligible. Look at the graphs that show charging characteristics (current vs. voltage) of a capacitor and you will see what I mean. Everytime electronics are powered up they are subjected to these current surges. In fact, I believe that most failures occur at power-up. Anybody had experience with tantalums?
Anyway, this is one school of thought.

[Terry Yager]

What we're dealing with here is two separate issues; mechanical vs electronics. I'm basically a mechanic, not an ee, but AFAIK, the consensus seems to be that on/off cycles are harder on electronic parts than 24/7, for all of the reasons mentioned above (especially the surges at powerup), and mebbe some others as well. For some mechanical parts, especially bearings, 24/7 is also better, but for some other mechanical parts, commutators & brushes, etc, shutting down is better, as the more the motor spins, the more wear on these parts (parts which make physical contact with each other). Spindle bearings are a different issue, because that when the motor runs up or down, there is a slight wobble, no matter how close the tolerance is. After spinning up, however, centrifugal force takes over, and the drive spins on the level, like a gyroscope, so there is less wear & tear on those bearings. Remember, spindles 'float' in a layer of lube, so the only time they actually make intimate contact with the bearing is during the wobble periods at spin-up/down, whereas other parts make actual contact at all times. (Armature shafts also ride in bearings, so the same principals apply). QC in designing motors, etc, is a process of figgering out a good balance between the various opposing stresses, so that they all wear out at pretty much the same rate (including the physical fatigue from heat/cool cycles, (expansion/contraction), which can lead to micro-stress fractures in solder joints, etc).

Basically, you get to decide whether you'd rather have the drive seize-up from wear, or broke down from electronic failure.

--T