Hello!
I am getting the parts together for a tower server build. I plan on running Jellyfin, maybe dive into arrs and nextcloud for 2 users total, wireguard only for external access as it’s not the main focus for now.
Situation: if I have access to refurb/used 4TB enterprise HDDs at the same price as 1.9ish TB enterprise SSDs.
I’d take lower capacity as it is not that big of a concern for me rn. I want to have somewhat redundant storage of my documents, photos, but otherwise it’s not gonna be a giant media vault overflowing with movies.
Question: In terms of noise, shipping concerns and longevity, would you go with SSDs instead of HDDs? Is it lower maintenance?
I can of course buy spinners later if I find flash only to be restricting in any way, and add to the rig as needed.
Speed would not be an issue in any case. This is for TrueNAS scale, so zfs. I am planning to buy 3-4 disks now, and add more if needed in 6 months time or later.
I am eager to hear others opininons on this. Thanks!
I’m curious on where did you find this. Maybe they have lower DOA rates and decreased chances to fail in the first year, but SSDs have a limited usage lifetime / limited writes, so even if they don’t fail quickly, they wear out over time and at first they have degraded performance, but finally succumb in 5 years or less, even when lightly used (as in as OS drives).
To avoid DOA / first year issues with HDDs, just have the patience to fully scan them before using with a good disk testing app.
Backblaze and other “bulk storage providers” tend to release reports every-now-and-then about the lifetime of their drives (it’s important for them to understand the costs). Here’s one such Backblaze Report.
Holy shit, I stand corrected, those graphs speak for themselves. Bookmarked for future stats.
LE: Well, there’s also the section about average age of failure in their newest report: 2 years and 7 months for HDDs, 14 months for SSDs.
Yeah there are a lot of ways to cut the data. Average age of failure doesn’t mean more things fail, just that if they do fail they’re likely to be around that age.
In general the reliability seems to be “close enough” between the two that it won’t matter for a home user who doesn’t have 10,000 storage units running in a server room. 😀
The write cycles shouldn’t really be an issue for a home NAS because you’re not erasing and rewriting over and over. For commercial projects, where logs, security video, or rotating data needs to be stored and erased hundreds of thousands of times.
True, but it depends from person to person and it counts if you have a small or big drive, how often you watch and rotate your media, how large the media is. If you only have a 1TB SSD, and often download and watch blue-ray quality, 20 movies will fill it. It won’t be long until the same blocks get erased, no matter how much the SSDs firmware tries to spread the usage and avoid reusing the same blocks.
Anyway, my point is, aside from noise and lower power consumption advantages, I wouldn’t use SSDs for a NAS, I regard them as consumables. Speed isn’t really an issue in HDDs.
Good point, but let’s say you download 20 new movies, meaning rewrite to every block on the drive each week. That’s barely 1,000 write cycles a year, and we’re still talking about a hundred thousand write cycles, which would take 100 years. Even if you start seeing bad blocks at 10,000 write cycles, by the time the drives are wearing out, the cost of replacement drives should be considerably lower.
even consumer SSDs have around 1500 TBW (Terrabytes written) per TB until warrenty excludes any failure
which means you could write for example every day for 10 years 400 GB on a 1 TB SSD
this is already a very low estimate, most SSDs do better
anyway OP mentioned enterprise SSDs which can write 1.0x or 2.0x it’s own size every day for 10 years
Would you suggest some good testing app? I just use crystal disk info to see the stats. Never ran a test as such.
I use Hard Disk Sentinel, it’s not free, but it also monitors drives in Windows so you have an early warning at the first sign of issues. Also logs historic data (writes, temperature, etc) and displays them as graphs.