Hi All, spending some time pondering how/if to improve my servers and the way that shares have filled it up over the years.    I should probably start by mentioning that i have watched all of the super-useful spaceinvader yourtube videos, and specifically the one about getting the pest performance out of shares.    truth be told, i have no need to change anything, so this is a thought exercise and i would love your thoughts.

 

I have 8 HDD's in my array.   I have SDD mirror cache drives, doing their thing, and i have a separate HDD being bludgeoned by my security cameras.    as part of my recent renewed interest, i have been looking at my family shares (all my family documents, pictures, etc) and our super precious photos directory and some low value stuff like an isos directory and a temp folder (for my content creator) kid and a busy timemachine share for a few people with macs etc.   anyway, i had always used the default high-water "spread it all over the place" default option, which ultimately meant more spindles and (possibly incorrectly assumed by me) more IO speed.     the issue is, that i started to think that maybe i was just ageing all 8 of the drives for a very marginal performance increase in a house-setting.  I mean, these poor disks literally never spin down.  while they are all WD Red drives, i am not in a corporate environment and dont want to unnecessarily increase the time between drive failure, etc.

 

So, I discovered unbalance, and thought "I Know, i will consolidate all my family profiles onto one disk, all my photos and videos onto another disk, any rarely-access stuff onto another disk and leave one disk spinning all the time for timemachine, temp files an dbackups etc.   the goal here is to have as many drives sleeping as i can, unless i need them.    unless i go on to discover that this has substantially impacted performance, or that i will get annoyed by having to wake up disks etc...

 

i am sure that a lot of people have got to this point.   the issue is that if you try any google a common sense discussion on this topic, you end up with a deep-dive into nerdery and an immediate obsession to post links to the same unraid youtube videos that i have watched countless time and learnt a lot from.    what i want is some real-world discussion and guidance on how other people manage this.   

 

Am i better off prolonging the life of my drives, by consolidating the shares and letting them sleep?

- of am i totally misunderstanding this concept?

is there a good-enough performance bump for me to spread shares across all the spindles?

- (noting that i realise that this also reduces a single point of failure for the data in that share)

 

actually i am somewhat out of questions.....   rightly or wrongly, i have done a bit of consolidating recently in my shares and everything is running nicely now, however, i am happy to have someone talk common sense into me if i have done something daft.

 

I cant wait to hear your advice!

cheers!

Floyd

I'm in the consolidate and let them sleep camp, in fact I keep most of my "always on" stuff on the cache to help. As long as they're not constantly spinning up and down all the time, It's the best way to go IMO - if you don't mind the delay when waking.

 

Set a decent timeout, 60-120 mins is my suggestion.

Thanks Michael, it is reassuring to hear that at least one other person is thinking like me    it would be great if this lights up a bit with ideas, views and details that i had not considered.   it is interesting that some of that topics that are considered to be uninteresting core functionality, often have the least diversity of information, disucssion and understanding.     eg, there is typically the singular 'best' video on the topic and it becomes the default answer for any variation of question.    

 

thanks for sharing your individual view on the topic.  good suggestion on the timeout setting too.

The choice to sleep or don't sleep drives is often pretty binary - mechanical drives are most likely to encounter an issue on power on, or on spin up, so if you are doing it often it's adding additional "wear" to the drive increasing (by some amount) the likelihood of component failure.

 

Conversely, while a drive is happier to keep spinning than not (easier to keep moving than to start moving), there is wear on the bearings - so it's not a panacea, either.

 

You need to take your personal use case into consideration, and add whatever variables to weight the decision on one side or the other (lower power consumption = plus, longer wait times to access files = con).

 

There really is no Right Way™️

 

 

1 hour ago, TheFloyd said:

Am i better off prolonging the life of my drives, by consolidating the shares and letting them sleep?

- of am i totally misunderstanding this concept?

It is debatable whether spinning them down prolongs life.  The spin-up process is that tends to be hardest on drives.

 

1 hour ago, TheFloyd said:

is there a good-enough performance bump for me to spread shares across all the spindles?

You are always limited by the performance of a drive as Unraid does not split files across drives or employ striping.  Whether you will see any performance gain it therefore highly dependent on whether concurrent access is to different drives.

 

1 hour ago, TheFloyd said:

(noting that i realise that this also reduces a single point of failure for the data in that share)

Ideally you mitigate this risk by having backups of anything important.

 

I must admit I personally decide on a per-share basis whether I want it to be split.  The main reason for me to constrain the drives to be used is to simplify backup and recovery.  I have a good backup strategy so I know I can always recover that way if needed.

Thanks Michael and itimpi, i have to say...   i am pretty new to the unraid forum, and so far i would have to rank it as the best environment ever for asking people for help and a general chat.    super impressed with people offering solid advice!

 

itimpi, i had not considered the fact that it doesnt split files across drives.   I think that is a remnant of too many years playing with other raid solutions in corporate...  good call.

regarding backups, i have that covered.    in addition to raid resilience with my parity, i have it continuously syncing from a crashplan docker to an off-site encrypted cloud storage.   I actually have a spare drive floating around asking for trouble, so i might work out how i can cron a nightly task to nightly rsync my most valuable data to it, such that it is out of the array entirely and in a caddy that can be pulled and chucked in a grab-bag.     that will be the next thing i look at....

 

the other thing i was wondering is if there is much of a performance hit from using the drive i just mentioned as a secondary parity drive?   seems to me from what i have read that if you are not protecting mission-critical data, the 2nd parity disk is perhaps not that valuable, given the even lower chance of a second drive failure.   I currently like my grab-bag idea instead of Parity2.