Question Protecting the Equipment - UPS, Surge Protection or Nothing?

Monster900

Established Member
Joined
Dec 30, 2011
Messages
410
Reaction score
103
Points
83
Just an open question really. The focus here seems to be, quite rightly, on protecting the data. There seems to be little discussion on protecting the equipment that stores the data.

I have recently re-purposed an old, low power consumption PC to act as a dual purpose machine NAS (RAID 1) and PC which is located in the garage. The now superseded Zyxel NSA310 was protected by an APC Back-UPS in the office, but I was concerned about storing data on the unprotected machine in the garage. I was pretty unsure what to do, but eventually decided to buy a UPS for the the NAS/PC in the garage. The only thing that bothers me is that the UPS probably uses at least as much electricity as the NAS/PC, equivalent to about £27 per year.

On the other hand, in the month that the UPS has been installed, there have been two power cuts each lasting about 20s, the first within 24 hours of putting it in and the second about two weeks later. Where I live this is not an uncommon event. In the winter longer power cuts often happen as well, so maybe £27 p.a. is just the price I have to pay for data security where I live. It may be worth mentioning, before the 'RAID is not backup' cries kick in, that critical data is backed up to a re-purposed thin client machine back in the office.

So the question is, do others here protect their equipment with UPS, surge protectors or just not bother? I'm just interested to know and welcome the input. Thanks in advance.
 
Last edited:
The only thing that bothers me is that the UPS probably uses at least as much electricity as the NAS/PC, equivalent to about £27 per year.

Firstly, I absolutely agree with a UPS. I run x4 to support my kit across the property inc Routers, core switches, Ap’s, servers, x2 Synology nas etc. Sone UPS’s support extended runtime in case of power outages, (router, switches, ap’s (so we still have network, internet)), some just to support safe shutdown of non essential kit like the NAS’s etc.

Re the above quote, not sure how you arrive at that conclusion. If battery is charged, the UPS is just a pass through (perhaps minus a tiny overhead). It’s certainly not “using as least as much as the PC”.
 
The only thing that bothers me is that the UPS probably uses at least as much electricity as the NAS/PC, equivalent to about £27 per year.

Thanks for the quick response. I'm glad to see that someone else thinks that UPS should form part of the data protection strategy for home NAS.

As far as the electricity cost estimate, I emailed PowerWalker and asked them what the fully charged power consumption is for my UPS (PowerWalker Basic VI 1000 SB UK). They responded to say that it is 22 watts. By my calculation at about 14p per KwH that's just under £27 a year for 24/7 operation.

I guess even fully charged it takes some power to drive the electronics that monitor the utility power supply and send event data or shutdown instructions to the NAS. Does seem a bit on the high side though.
 
The batteries will be float charged all the time, as well as the overhead for the monitoring and even the relays that change the power over.

I would certainly use a UPS in your situation, but I would be less concerned about surge protection, as many UPSs detect over voltage as well as brownouts and will switch over to batteries in this case as well.
 
As you might imagine "in business" we use UPS all the time to protect critical infrastructure. It's not just about protecting data loss (due to the servers having the legs pulled out from under them), but about availing continued operation in the event of power failure.

A lot of businesses these days use "VOIP" telephones, so in the event of a power outage when the phones still "need to work" (for safety reasons for example) we need to ensure the data network that carries the VOIP traffic keeps on running, so we UPS all the network switches etc. too.

Some places have the entire computer room on UPS. Back in my days of working on mainframes, the entire site was on UPS with a deisel generator on standby for when the batteries ran out.

How much protection one might avail is a value judgement which then informs what we want to protect and how much "run time" it needs whilst on battery. APC used to have a tool on their web site to help you size a UPS based on the load, required run time and so on.

Even when "on mains" most UPS's "condition" the power to provide a "cleaner" supply to the kit (including surge protection and so on.) (Hi-Fi buffs used to be into this sort of thing years ago when Hi-Fi was all analogue.)

A facet of UPS use often overlooked is that the batteries don't last forever and reduce capacity over time and eventually need to be replaced. One should factor that in to OpEx costs, though whenever I've done it the cost of replacing the batteries was about 90% of the cost of a new UPS so it was easier to "just buy a new one" rather then replace the batteries. A lot of UPS's will do a periodic "self test" and use that as the basis for their predicted run time - if one monitors that over time and looks for the trend, we can start to predict when it's time to change the batteries/replace (though we are usually talking a time frame of years.)
 
Last edited:
I agree that UPS is important to protect equipment and data. I live in an area that has frequent power cuts and thunderstorms, and earlier this year there was a serious fault in the power supply coming into the village. All the houses received double voltage, and lost appliances (5000 euros worth in our case). However, computers, router, NASs, etc were all on UPS and survived.
Some of the worst cuts are when the power flickers on and off several times for a few seconds. I have lost back up drives when this happens.
Last time the power was cut, i discovered that the battery of one UPS had failed, so I have ordered new ones (generic, and cheap, so I hope they are OK.)
I definitely recommend UPS if your electricity supply is at all unreliable.
 
Also +1 for a UPS as I had issues with an RCD becoming far too sensitive and tripping. It can be quite easy to lose a disk\data under these conditions.

I have my server on it, plus networking gear and now my Office PC. If there is a power failure my server will do a controlled shutdown when a given battery percentage is reached. This still allows the network gear to run longer.

One thing to watch is if you have converted a PC then how is it writing to the disk. You can have write through (to the cache and disk) and write back (to only the cache). If you have a power failure on the latter it will cause data loss not to mention possible corruption. A UPS helps, as does a battery backup for a RAID controller which you get on high-end cards.

I am on my third set of batteries now and stick with APC branded ones as last time they were around the £140 mark or so this was for an APC 1500 unit which produces a pure sine wave. Some of the cheaper models do not so it is worth checking if you have anything that needs a proper AC signal (e.g. some motors)

My house uses quite a lot of power in "standby" even though I have reduced this recently. I reckon on every 100W of power costing £100 a year if on 24/7 but even that is a bit low these days.
 
Yup, I have an APC UPS that covers my NAS and backup server, along with critical networking infrastructure. It's safely shut down NAS/Server several times over the past 10 years or so. Even with off-line backups I'd hate to have to re-populate 30TB odd of data (mostly media rips, but there's several TB or my own photos, videos and data). The only PITA is having to replace the batteries every 3 years or so, but in my case that only comes to £35 or so (I use branded 3rd party batteries, not the APC kit).
 
Thanks for all the input, it would seem that the consensus is that it was the right thing to do to buy an UPS for the NAS/PC in the garage, which is a useful confirmation as I wasn't sure it was worthwhile.

...............................................
One thing to watch is if you have converted a PC then how is it writing to the disk. You can have write through (to the cache and disk) and write back (to only the cache). If you have a power failure on the latter it will cause data loss not to mention possible corruption. A UPS helps, as does a battery backup for a RAID controller which you get on high-end cards.

Thanks for this bit of info,I had forgotten all about disk caching. This is not a sophisticated machine with a RAID controller, it's just using Windows 10 disk mirroring, network folder sharing and the built in Win 10 DLNA to provide the NAS/media server functionality . It is now protected by an UPS with automatic shutdown after 10 minutes of utility power failure, but I think you are suggesting that I should turn disk write caching off in 'device manager' for the two mirrored disks to be sure. Is that correct?
 
You could turn off write back caching for the discs in the OS, however that's not the only caching going on and not usually the source of the "RAID write hole" you may have read about.

Most modern "consumer" discs have their own internal "on board" caching. They use this to give the illusion of faster performance to the host operating system by accepting a write operation, then reporting it back as "completed" even though the data hasn't actually hit the platters yet and is still sat on the discs internal buffers.

Ideally, if we wanted to built a robust RAID implementation, we'd use discs with no on board caching, or at least the ability to turn it off with either a jumper or firmware change. As discussed above, the reason "big boys" RAID controller cards for servers used to cost so much (500GBP or there abouts) is because the controller not only knew how to do al the RAID'ing, but also managed all the caching including using battery backups. (HP used to sell a card that would maintain it's state for a couple of days if the power went off.) We'd then use discs with no caching at all.

If you are relying on "software" RAID, and commercial "desktop" discs, (even things aimed at NAS's like WD Red's) then it's drivers and/or the host OS doing all the RAID and may not be "aware" of any caching going on in the disc drives themselves, and thus be vulnerable to the "write hole" issue.

"Silent corruption" AKA "bit rot" is something else than can trip some people up also. As rare as it is, data on a disc might "go bad" when you're not looking. Again, in business, in olden days we'd copy the data away to tape every night to make a backup. That means we've read every file stored every day and if there's a block gone bad and thus rendered a file unreadable, the backup process would discover it, we'd spot it in the next days report then fix the disc and/or restore the damaged file from the last good backup.

As a lot of backup solutions have moved to delta-blocking techniques, the backup process is no longer reading every file every night, just the stuff that's changed, so we run the risk of not spotting such silent corruption so quickly, if at all. And of course, if one never makes backups and/or never checks the logs, you never spot it anyway.

And of course, this is somewhat overkill for a media tank where the data rarely changes.

When I built my tank, I used a file system available under UNIX/Linus called ZFS. (Some of the turnkey software NAS solutions are also built on it.) ZFS design mantra seems to be "don't loose data" and it implements techniques to deliver this. One of them is called "scrubbing." When ZFS scrubs, it goes and reads all the data stored and verifys it against various checksums etc. and tests whether what is stored is still good. If it can, it'll try to fix anything that's gone bad.

Thanks to ZFS scrubs, I found one of my discs "going bad" long before SMART or anything else reported any issues, simply because the number of times scrub was repairing data, so I replaced the offending unit "early" and everything has been good since. (Incidentally, it was a WD Red only a year or so old - sometimes it's just unlucky and you get a "bad" one.)

If you are really concerned about data integrity, you might care to check out ZFS and see if you can find a platform based on it.
 
Last edited:
You could turn off write back caching for the discs in the OS, however that's not the only caching going on and not usually the source of the "RAID write hole" you may have read about.

Most modern "consumer" discs have their own internal "on board" caching. They use this to give the illusion of faster performance to the host operating system by accepting a write operation, then reporting it back as "completed" even though the data hasn't actually hit the platters yet and is still sat on the discs internal buffers.

Ideally, if we wanted to built a robust RAID implementation, we'd use discs with no on board caching, or at least the ability to turn it off with either a jumper or firmware change. As discussed above, the reason "big boys" RAID controller cards for servers used to cost so much (500GBP or there abouts) is because the controller not only knew how to do al the RAID'ing, but also managed all the caching including using battery backups. (HP used to sell a card that would maintain it's state for a couple of days if the power went off.) We'd then use discs with no caching at all.

If you are relying on "software" RAID, and commercial "desktop" discs, (even things aimed at NAS's like WD Red's) then it's drivers in the host OS doing all the RAID may not be "aware" of any caching going on in the disc drives themselves, and thus be vulnerable to the "write hole" issue.

"Silent corruption" AKA "bit rot" is something else than can trip some people up also. As rare as it is, data on a disc might "go bad" when you're not looking. Again, in business, in olden days we'd copy the data away to tape every night to make a backup. That means we've read every file stored every day and if there's a block gone bad and thus rendered a file unreadable, the backup process would discover it, we'd spot it in the next days report then fix the disc and/or restore the damaged file from the last good backup.

As a lot of backup solutions have moved to delta-blocking techniques, the backup process is no longer reading every file every night, just the stuff that's changed, so we run the rusk of not spotting such silent corruption so quickly, if at all. And of course, if one never makes backups and/or never checks the logs, you never spot it anyway.

And of course, this is somewhat overkill for a media tank where the data rarely changes.

When I built my tank, I used a file system available under UNIX/Linus called ZFS. (Some of the turnkey software NAS solutions are also built on it.) ZFS design mantra seems to be "don't loose data" and it implements techniques to deliver this. One of them is called "scrubbing." When ZFS scrubs, it goes and reads all the data stored and verifys it against various checksums etc. and tests whether what is stored is still good. If it can, it'll try to fix anything that's gone bad.

Thanks to ZFS scrubs, I found one of my discs "going bad" long before SMART or anything else reported any issues, simply because the number of times scrub was repairing data, so I replaced the offending unit "early" and everything has been good since.

If you are really concerned about data integrity, you might care to check out ZFS and see if you can find a platform based on it.

Wow and many thanks for that really thorough explanation of RAID and disk caching potential pitfalls. I will think about all of that for any future NAS build or purchase.

For now, I think I'm reasonably content with what I have set up with everything on UPS and back up to a different machine in a different building. This is far more than anyone else in my family does, but then they lose stuff all the time but it doesn't seem to bother them as much as it would bother me.
 
Last edited:
Something else occurred to me regarding "spent" UPS batteries - of course in the "good old days" we'd have just chucked them in the bin (or piled them up in the corner of the server room until we got sick of tripping over them - on one site I used to use one as a seat!) These days they need to be responsibly disposed of and/or recycled which might be a further cost to the full lifecycle to keep in mind.

APC, for example, used to have schemes to return spent batteries to them as part of the replacement cost.
 
Last edited:
Something else occurred to me regarding "spent" UPS batteries - of course in the "good old days" we'd have just chucked them in the bin (or piled them up in the corner of the server room until we got sick of tripping over them - on one site I used to use one as a seat!) These days they need to be responsibly disposed of and/or recycled which might be a further cost to the full lifecycle to keep in mind.

APC, for example, used to have schemes to return spent batteries to them as part of the replacement cost.

At the moment my Local Authority will take them without charging at the local recycling centre as long as you don't turn up in a van. Not sure if other LAs are the same though.
 

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom