2
I am planning to build a small server for personal use, utilising a relatively large RAID-5/-6 (or RAID-Z/-Z2) array as main storage.
The vast majority of workloads I envision for this would have very predictable disk access patterns (e.g. movie streaming), and a priori I don't think there would be a need for the RAID array to be powered up most of the time. In principle it would be possible to e.g. cache an entire video file on a SSD upon start of playback and then spin down the disks. Most other workloads I imagine would have similar patterns (e.g. filesharing client would only ever seed a small number of files at a time, remote backup solution could easily cache to SSD first, then write back to disk in one go). Needless to say, minimising disk uptime for five or six hard disk might make a noticeable difference to power usage.
My question is thus: Is there any software solution (OS, file system, etc.) that would allow me to use a SSD as a cache for a RAID array in a way that minimises disk uptime, rather than maximising performance?
I know zfs for instance allows one to specify separate read and write caches, but as far as I know this is only used to improve performance of the filesystem. The same goes for all other SSD caching / tiered storage solutions I have found so far.
Not sure if this questions is most appopriate here or in serverfault - Since it is for personal use I've posted here, but let me know if I should consult serverfault instead.
matt
Posted 2013-02-20T11:38:33.257
Reputation: 21
1Just a thought - is having a large amount of RAM and using a RAMdisk an option? – Karan – 2013-02-21T03:59:15.203
1Hadn't thought of that. In principle it would be possible, though I think most mainboards will only support 32GB of RAM. Still the issue would remain the same though: Is there any OS or filesystem that could use such a (RAM or SSD) cache / 1st-level storage tier in order to keep disks spun down as much as possible? – matt – 2013-02-21T14:01:36.577