Hacker News Reader

Comment by mrb | Hacker News Reader

I get annoyed when people make generic claims that SSDs from vendor X are not as reliable as vendor Y.

The truth is that each vendor manufactures multiple lines of SSDs, each running different major firmware versions, with different hardware revisions. So reliability variance between models from a single vendor is often much more significant that variance between different vendors.

In your experience with OCZ, which model do you have experience with? Vertex 2? 3? Agility? Enterprise-class Talos? What about the Intel ones? Were they 50nm NAND-based SSDs? 34nm? You need to share some of these details. Claiming that Intel > OCZ is devoid of significance.

View on HN

The article makes those claims, based mostly on anecdotal evidence rather than a peer-reviewed scientifically rigorous study. It concludes with: "Giving credit where it is due, many of the IT managers we interviewed reiterated that Intel's SLC-based SSDs are the shining standard by which others are measured."

While an itemized breakdown by datacenter isn't available one datacenter (InterServer.net, with fewer than 100 SSDs) says: "Intel SSD's are night and day in failure rates when it comes to some other drives. For example the SuperTalent SSD drives have had an extremely high failure rate including model FTM32GL25H, FTM32G225H, and FTM32GX25H. I estimate about two-thirds of these drives have failed since being put into service. With these failures however, the drives were not recoverable at all. They generally disappeared completely, no longer being readable. Spinners die much more gracefully with an easier disk recovery. I cannot compare this to the Intel's SSDs yet since I have not experienced any failures.". They apparently use Intel's X25-E (SSDSA2SH032G1GN).

They generally disappeared completely, no longer being readable. Spinners die much more gracefully with an easier disk recovery.

That's a weird thing to say. In a datacenter you want your disks to be fail-fast. I.e. you want them to drop out of the array early and cleanly.

Nobody "recovers" harddrives in a datacenter, you just swap them.

The worst failure-mode (that is still way too commonly seen) is that of the "lame drive" that grows bad sectors, predictive failures and all that jazz over time, causing bus-resets, confused controllers and bringing the whole array to a crawl until a human steps in.

That said, my experience with SSDs in that regard isn't too good either. I've seen them fail cleanly but I've also seen a single failing Vertex lock up a LSI-controller hard...

I agree - you can't just look at brand, as there are often a lot of issues that are specific to individual models. Having said that, my experience selling OCZ drives (100s of drives) is that they all have a high failure rate :( The only drives I haven't had experience with are the Enterprise level drives, so I can't comment on them.

I think the reason for this is that OCZ largely uses Indilinx and Sandforce controllers, and while very quick, they have had more than their fair share of issues.

Reply on news.ycombinator.com