SSDs Today - Any Reason to Fear Writes?
-
I understand that it was once appropriate to fear excessive writes to a SSD. Is that still the case today?
I am running an Intel 320 SSD on my Supermicro 5018A-FTN4.
At the moment, it appears that I log about 100 packets a minute, although since I've newly installed pfsense on this machine, I am logging more packets than is really necessary.Should I be concerned about the SSD? I just want pfsense run uninterrupted. I've already enabled TRIM on pfsense.
-
If it's a top-tier manufacturer like Intel, there is not much need to worry.
And not that it has a direct relation, but this story came out yesterday as well:
http://hothardware.com/news/google-data-center-ssd-research-report-offers-surprising-results-slc-not-more-reliable-than-mlc-flash
-
Dig out your calculator, figure out how much space your 100 packets a minute use and divide the free space on your ssd by that amount. That will give you the number of minutes it would take to fill your drive ONCE. Divide by 60 for hours, then hours by 24 for days, then days by 365 for years. The current gen ssd's can be overwritten daily for years without being worn out.
The 5 computers in my house are 100% ssd. The gamer I'm writing on has 4tb of ssd in it. The only mechanicals in use are in the NAS boxes. The ssd's in these machines will easily outlast the life of the computer. Being an old geek that likes to play, all the ssd's are in carrierless removable drive bays. It takes 15 seconds to change the operating system on the gamer to another version of windows, linux, freebsd, or what ever the hardware can handle.
I'm putting pfsense on a Shuttle with an older 250gb ssd from the cabinet and expect to never bother it again for the life of the machine.
With the new memory Intel and Micron are working on a quantum leap in memory and ssd speed is coming. Ssd's are already moving from overloaded sata busses directly to the pcie lanes and getting faster all the time.
Enjoy your silent, fast ssd.
-
No you shouldn't be concerned.
We have 320's being brutalized as zfs-cache drives in "el-cheapo" archive servers and even with that workload they're handling that very well in the scope of being a consumer drive.
@ddaniel51:
Always take the erase-block-size of 256 pages (of 8k) into account. A block write on a ssd is not simply a written block as it is on a harddrive.
https://en.wikipedia.org/wiki/Write_amplification