Array Power Management

Array Power Management

[jahammonds prost]

I've been becoming interested in the power consumption of the arrays I have on a couple of servers here that are basically used as media servers. Since they're not being used when I'm at work, I was looking at the possibility of spinning the disks down during certain times, and having them either spin up at a set time, or (ideally), when there is disk activity.

I can do this on single drives using hdparm -S to set the spindown timeout, and the disks will spin up on activity as needed. Is there something similar I can do with an md array? I can see there's a /sys/block/md0/power/wakeup file, but I can't seem to find any documentation on it. I have thought about doing an hdparm -S on the array disks, but I suspect that would be A Bad Thing (tm).

Does anyone have any advice/pointers? The servers are currently running fc9 - I'd like them to be Centos, but had to go with fc9, as I needed Port Multiplier support.


Graham


      
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Array Power Management

[Majed B.]

If you're going to spin down the disks, you'd need to unmount the
array first to ensure your data is fine, and after a spin-up then you
mount the array again.

I would suggest you buy a meter to measure how much power your storage
is consuming before you decide on your next action.

Personally, I'd turn off the machine altogether if it's not needed
during working hours (assuming no one accesses through VPN). Check the
BIOS, it may have the capability to wake up the machine on a certain
date & time (So you can specify to turn itself on daily @ 8 AM).
If not, you should have Wake-up On LAN support, so if you turn off the
machine, another machine (a monitoring server) can turn it on for you
(just put the wake-up on LAN command as a cron job to execute daily --
perhaps exclude holidays).

On Mon, Sep 7, 2009 at 6:48 AM, jahammonds prost<gmitch64@yahoo.com> wrote:
> I've been becoming interested in the power consumption of the arrays I have on a couple of servers here that are basically used as media servers. Since they're not being used when I'm at work, I was looking at the possibility of spinning the disks down during certain times, and having them either spin up at a set time, or (ideally), when there is disk activity.
>
> I can do this on single drives using hdparm -S to set the spindown timeout, and the disks will spin up on activity as needed. Is there something similar I can do with an md array? I can see there's a /sys/block/md0/power/wakeup file, but I can't seem to find any documentation on it. I have thought about doing an hdparm -S on the array disks, but I suspect that would be A Bad Thing (tm).
>
> Does anyone have any advice/pointers? The servers are currently running fc9 - I'd like them to be Centos, but had to go with fc9, as I needed Port Multiplier support.
>
>
> Graham
>
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>



-- 
       Majed B.
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Array Power Management

[Jon Hardcastle]

--- On Mon, 7/9/09, Majed B. <majedb@gmail.com> wrote:

> From: Majed B. <majedb@gmail.com>
> Subject: Re: Array Power Management
> To: "jahammonds prost" <gmitch64@yahoo.com>
> Cc: linux-raid@vger.kernel.org
> Date: Monday, 7 September, 2009, 10:57 AM
> If you're going to spin down the
> disks, you'd need to unmount the
> array first to ensure your data is fine, and after a
> spin-up then you
> mount the array again.
> 
> I would suggest you buy a meter to measure how much power
> your storage
> is consuming before you decide on your next action.
> 
> Personally, I'd turn off the machine altogether if it's not
> needed
> during working hours (assuming no one accesses through
> VPN). Check the
> BIOS, it may have the capability to wake up the machine on
> a certain
> date & time (So you can specify to turn itself on daily
> @ 8 AM).
> If not, you should have Wake-up On LAN support, so if you
> turn off the
> machine, another machine (a monitoring server) can turn it
> on for you
> (just put the wake-up on LAN command as a cron job to
> execute daily --
> perhaps exclude holidays).
> 
> On Mon, Sep 7, 2009 at 6:48 AM, jahammonds prost<gmitch64@yahoo.com>
> wrote:
> > I've been becoming interested in the power consumption
> of the arrays I have on a couple of servers here that are
> basically used as media servers. Since they're not being
> used when I'm at work, I was looking at the possibility of
> spinning the disks down during certain times, and having
> them either spin up at a set time, or (ideally), when there
> is disk activity.
> >
> > I can do this on single drives using hdparm -S to set
> the spindown timeout, and the disks will spin up on activity
> as needed. Is there something similar I can do with an md
> array? I can see there's a /sys/block/md0/power/wakeup file,
> but I can't seem to find any documentation on it. I have
> thought about doing an hdparm -S on the array disks, but I
> suspect that would be A Bad Thing (tm).
> >
> > Does anyone have any advice/pointers? The servers are
> currently running fc9 - I'd like them to be Centos, but had
> to go with fc9, as I needed Port Multiplier support.
> >
> >
> > Graham
> >
> >
> >
> > --
> > To unsubscribe from this list: send the line
> "unsubscribe linux-raid" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >
> 
> 
> 
> -- 
>        Majed B.
> --
> To unsubscribe from this list: send the line "unsubscribe
> linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

This is interesting.. I have the -S set for all my 6 drives in the array and have for some time and I do not unmount the array - access wakes the drives up - usually most of them, sometimes only a few of them. The only part of this that has caused me any problems are scheduled smartd checks - these cause timeout errors.

The timeout is between 30mins and 1hr as if it is over 1 hour smart will keep them awake forever. In my case, if they are not used for 30mins; they wont be used for several hours.

-----------------------
N: Jon Hardcastle
E: Jon@eHardcastle.com
'Do not worry about tomorrow, for tomorrow will bring worries of its own.'
-----------------------


      
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Array Power Management

[jahammonds prost]

> This is interesting.. I have the -S set for all my 6 drives in the array and have for some time and I do not unmount the array - access wakes the drives up - usually most of them, sometimes only a few of them.

I wasn't sure that that would be a safe thing to do, and not finding anything via google, I though it was best to ask. I've also been looking at sdparm, which seems to be more in tune for SATA drives, rather than PATA ones.


> The only part of this that has caused me any problems are scheduled smartd checks - these cause timeout errors.

How are you scheduling the checks? via cron? It's not something I do at the moment, but probably should - I'd rather get some notification that a drive is going to fail, rather than have to deal with a failed one. I'd think about putting an sdparm --command=start /dev/sd[c-p] or whatever to start the disks up before doing a check (and maybe a command=stop when done). I'm guessing that it would be OK to use hdparm to set the spin down timeout, and then sdparm to start the disk spinning again (I can't find an sdparm option to set the timeout)

> In my case, if they are not used for 30mins; they wont be used for several hours.

This is usually the situation I am in as well, certainly my media server... It usually get used for 3 or 4 hours, then nothing much till the next day.


Graham



      

Re: Array Power Management

[Majed B.]

Well if that's the case, then why not just enable Wake-up On LAN and
keep one machine running to be able to power-on those that are off?

When you're not using the machine, power it off. You could keep it
running on one certain day where you schedule your smartd's tests.

On Mon, Sep 7, 2009 at 5:02 PM, jahammonds prost<gmitch64@yahoo.com> wrote:
>> In my case, if they are not used for 30mins; they wont be used for several hours.
>
> This is usually the situation I am in as well, certainly my media server... It usually get used for 3 or 4 hours, then nothing much till the next day.
>
>
> Graham
-- 
       Majed B.

Re: Array Power Management

[Tapani Tarvainen]

On Mon, Sep 07, 2009 at 12:57:33PM +0300, Majed B. (majedb@gmail.com) wrote:

> If you're going to spin down the disks, you'd need to unmount the
> array first to ensure your data is fine, and after a spin-up then you
> mount the array again.

You don't needo to unmount it: just leave it mounted, and it will
automatically spin up when accessed.
I've been doing just that with a backup (rsnapshot) server for
a some time, works just fine.

> I would suggest you buy a meter to measure how much power your storage
> is consuming before you decide on your next action.

In my case the total power consumption went down by over 50%
(the disks stay powered down by about 85% of the time).

> On Mon, Sep 7, 2009 at 6:48 AM, jahammonds prost<gmitch64@yahoo.com> wrote:
> > I've been becoming interested in the power consumption of the arrays I have on a couple of servers here that are basically used as media servers. Since they're not being used when I'm at work, I was looking at the possibility of spinning the disks down during certain times, and having them either spin up at a set time, or (ideally), when there is disk activity.
> >
> > I can do this on single drives using hdparm -S to set the spindown timeout, and the disks will spin up on activity as needed. Is there something similar I can do with an md array?

I simply do it with every individual disk of the array.

-- 
Tapani Tarvainen

Re: Array Power Management

[jahammonds prost]

> You don't needo to unmount it: just leave it mounted, and it will automatically spin up when accessed.
> I've been doing just that with a backup (rsnapshot) server for a some time, works just fine.

That's great news, I was hoping there would be something simple.

> In my case the total power consumption went down by over 50%
(th> e disks stay powered down by about 85% of the time).

I don't have my figures here at the moment, but I'm hoping to save a big chunk like that too. I'm spinning 12 x 320GB disks on my file server, 15 x 500GB disk on one media server, and 7 x 1TB disks (and climbing - Got to lovr array expansion) on my new media server. Since it's really only mainly in use between 5pm and midnight, being able to spin down the disks would be a huge plus (or rather, minus on the electric bill). I had considered just shutting down the server, but I'm probably going to be (re)installing MythTV when .22 is released, so one (or more) of the servers will end up being storage for that, so I would never really know when the array is going to be needed.

> I simply do it with every individual disk of the array.

I was tempted to try that, but I wasn't sure if that would cause issues with md or not. I did some googling - found quite a bit on spinning down single drives, but noone had really anything to say about arrays (maybe I should have binged instead...:) )


Thanks to all for the suggestions.

-- 
Tapani Tarvainen
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html



      
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Array Power Management

[Matt Garman]

On Sun, Sep 06, 2009 at 08:48:03PM -0700, jahammonds prost wrote:
> I can do this on single drives using hdparm -S to set the spindown
> timeout, and the disks will spin up on activity as needed. Is
> there something similar I can do with an md array? I can see
> there's a /sys/block/md0/power/wakeup file, but I can't seem to
> find any documentation on it. I have thought about doing an hdparm
> -S on the array disks, but I suspect that would be A Bad Thing
> (tm).

On the same note, does anyone have any thoughts on the wear-and-tear
caused by frequent spinup/spindown cycles?

My fileserver has the Western Digital RE2 "enterprise" grade drives.
I remember reading (years ago) that, in general, "enterprise" grade
drives were designed to be always running (think 24/7 server), and
rarely spun down.  As such, they did not tolerate a "consumer
desktop" usage pattern very well, and trying to save power in this
way would actually cause them to fail prematurely.

Is/was such a thing true?  Is it worth worrying about?

Thanks,
Matt

Re: Array Power Management

[Ryan Wagoner]

I remember this used to be the case, but it seems to be the opposite
now. What is interesting is in the lower priced drives there seems to
be a trade off for the non recoverable error rate and load / unload
cycles.

When I looked at drives over year ago for my storage I went with 1TB
Hitachi 7K1000 as they were a good price, rated for 24/7 operation,
and had the 1 per 1.0 E15 bits error rate. Recently prices have come
down and when I put together my ESXi box I went with 500GB WD RE3
drives which offer the best rating all around.

Ryan

WD RE3
http://products.wdc.com/library/specsheet/eng/2879-701281.pdf
Load/Unload Cycles 300,000
Error Rate: 1 per 1.0 E15 bits

WD Caviar Black
http://products.wdc.com/library/specsheet/eng/2879-701276.pdf
Load/Unload Cycles 300,000
Error Rate: 1 per 1.0 E14 bits

WD Caviar Blue
http://products.wdc.com/library/specsheet/eng/2879-701277.pdf
Load/Unload Cycles 50,000 minimum
Error Rate: 1 per 1.0 E15 bits

Seagate ES.2
http://www.seagate.com/docs/pdf/datasheet/disc/ds_barracuda_es_2.pdf
Load/Unload Cycles: Not Published
Error Rate: 1 per 1.0 E15 bits

Seagate Barracuda 7200.12
http://www.seagate.com/docs/pdf/datasheet/disc/ds_barracuda_7200_12.pdf
Load/Unload Cycles 50,000
Error Rate: 1 per 1.0 E14 bits

Hitachi 7K1000
http://www.hitachigst.com/tech/techlib.nsf/techdocs/67A68C59B27368FC862572570080FC70/$file/Deskstar7K1000_010307_final.pdf
Load/Unload Cycles 50,000
Error Rate: 1 per 1.0 E15 bits

Hitachi 7K1000.B
http://www.hitachigst.com/tech/techlib.nsf/techdocs/D70FC3A0F32161868625747B00832876/$file/Deskstar_7K1000.B_DS.pdf
Load/Unload Cycles 300,000
Error Rate: 1 per 1.0 E14 bits

On Thu, Sep 10, 2009 at 12:51 PM, Matt Garman <matthew.garman@gmail.com> wrote:
> On Sun, Sep 06, 2009 at 08:48:03PM -0700, jahammonds prost wrote:
>> I can do this on single drives using hdparm -S to set the spindown
>> timeout, and the disks will spin up on activity as needed. Is
>> there something similar I can do with an md array? I can see
>> there's a /sys/block/md0/power/wakeup file, but I can't seem to
>> find any documentation on it. I have thought about doing an hdparm
>> -S on the array disks, but I suspect that would be A Bad Thing
>> (tm).
>
> On the same note, does anyone have any thoughts on the wear-and-tear
> caused by frequent spinup/spindown cycles?
>
> My fileserver has the Western Digital RE2 "enterprise" grade drives.
> I remember reading (years ago) that, in general, "enterprise" grade
> drives were designed to be always running (think 24/7 server), and
> rarely spun down.  As such, they did not tolerate a "consumer
> desktop" usage pattern very well, and trying to save power in this
> way would actually cause them to fail prematurely.
>
> Is/was such a thing true?  Is it worth worrying about?
>
> Thanks,
> Matt
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html