Boston Linux & UNIX was originally founded in 1994 as part of The Boston Computer Society. We meet on the third Wednesday of each month at the Massachusetts Institute of Technology, in Building E51.

BLU Discuss list archive


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Discuss] SSD drives vs. Mechanical drives



Bill Bogstad wrote:
> Any hardware that isn't redundant can cause corruption, but proper
> use of redundant hardware (appropriate RAID levels) can protect
> against corruption caused by those devices.

Write holes. Parity corruption. Rebuild errors. Cache inconsistency.
These can and will damage your data despite RAID systems being operated
properly. RAID can prevent data loss but it can't do anything about data
integrity. RAID plus checksums might be able to detect and correct
errors, or it might not. This depends on the RAID controller in question
and the cause of the errors.


> Only if you turn the storage device off when you are done with it. 
> With rotating rust that has traditionally been considered the best
> way to kill a hard drive.

Much of what we traditionally believe about hard drive life and death is
incorrect. In my -- and quite a few others' -- experience, the best way
to kill a rotating drive is vibration. A bad bearing in a chassis fan
can kill perfectly good drives within a few hours. Impacts against a
rack will be transmitted to operational drives and can almost instantly
damage them. If something like this damages one drive in a set then
chances are that all of them will fail within short spans of each other.


> Also, it isn't clear to me what you are saying about the power usage
> for SSDs vs. hard disks. My impression is that with SSDs and hard
> disk with capacities in the range of 500G-1000G, that SSDs are faster
> and use less power across a wide range of use cases.

It's not that simple.

These figures are usually based on peak performance benchmarks rather
than live operation, and they typically don't blow through the caches on
the devices. In other words, you're getting the 20%-200% write
performance benefit only as long as the DRAM cache holds out, then
you're down to whatever the flash chips can sustain directly. But then,
rotating media has always had the same problem. But then, rotating media
has more consistent performance than flash.

Once you hit that bottleneck the watt-hours start to matter. If a
process that takes 1 hour is extended to 4 hours because of a bottleneck
then that's a 4x increase in power consumption for the device. That's
not factoring the additional 3 hours worth of power that the hardware
around the device consumes. A HDD with more consistent performance under
load may be more efficient.

-- 
Rich P.



BLU is a member of BostonUserGroups
BLU is a member of BostonUserGroups
We also thank MIT for the use of their facilities.

Valid HTML 4.01! Valid CSS!



Boston Linux & Unix / webmaster@blu.org