3

I am getting ready to implement a RAID-10 config with 6 Samsung 860 EVO 500gb SSDs. They will be attached to an LSI 9260-8i controller. The specs for the SSDs lifespan read 300TBW.

Q1 - Can anyone verify if I would be correct to say that the total approx lifespan would be 900TB? (300TBW x 3 span volumes = 900TBW)

Q2 - Would writing be pretty much spread across all dives before writing over previous written cells keeping cell wear balanced?

Any feedback would be greatly appreciated.

crb314
  • 33
  • 3
  • Why are you using consumer SSDs in a server? – Chopper3 Oct 17 '19 at 14:20
  • I see a lot of comments indicating that; "today's consumer level is kind of like yesterday's enterprise level". So, I pretty much go by workload (in this case it is pretty low to medium), specs of the SSD and price. We are OK in replacing them in a couple of years. I have been using Seagate Constellation SAS Enterprise level on some of my other servers. – crb314 Oct 17 '19 at 15:53
  • If your load is low then I can see why you've gone this way, still not a very 'pro' thing to do to be honest. – Chopper3 Oct 17 '19 at 16:04
  • 1
    Thank you for your comment. I changed my mind and ordered Enterprise SSDs. I think your help made me decide on a better choice. – crb314 Oct 17 '19 at 20:51

1 Answers1

0

Mirror pairs double the writes, and there are three pairs. So yes, the total rated endurance is 900 TBW. However, this is a coarse and fuzzy wear indicator, actual life can be less or more.

RAID stripe size will be relatively small. However, those chunks are not what is written to the physical medium. SSDs virtualize a storage interface but do what they want at the page level.

Wear leveling is mostly a function of the SSD controller. It attempts to fight write amplification and uneven wear. Although it only can erase blocks at a time, and has a limited amount of spare over-provisioned cells.

John Mahowald
  • 32,050
  • 2
  • 19
  • 34