Welcome, Guest. Please login or register.
Did you miss your activation email?
November 29, 2024, 03:56:55 am *

Login with username, password and session length
Poll
Question: Does READING from a SSD affect the life of it?
Yes
No

Pages: [1]
Print
Author Topic: SSD Question  (Read 7443 times)
Hunter
EZ Server GM
Legend
*******
Posts: 8100


EZ Server GM


View Profile
« on: April 02, 2013, 05:37:42 pm »

I've done some Google search and seems to be a 10k - 100k limit of how many times each part of the SSD can be read/write, and then some websites saying its only the writes that are limited to 10-100k, not the reads.

For example, if the same file is going to be opened (read) about 10,000 per day, would that affect the SSD? Or only if the file is edited/updated/saved aka written to?

Logged

Hunter - EZ Server GM
Crabthewall
Sr. Member
****
Posts: 293


View Profile
« Reply #1 on: April 02, 2013, 05:57:26 pm »

Check email for answers specific to EZ Server hardware.
Logged
hateborne
Legend
*******
Posts: 2282


Don't nerf me bro!


View Profile
« Reply #2 on: April 02, 2013, 05:58:53 pm »

Going to sound presumptuous, but are you considering it for server? I ask because I've heard quite a few horror stories of SSDs being used for highly active databases and the drives keeling over extremely fast. I have also heard stories of people popping a bone for SSDs as DB drives. It seems to be case by case thing at the moment. The main difference that I do remember from the two is that a standard HDD will slowly die down (decreasing in performance and slowly increasing chance for data corruption1) whereas the SSD will work one and simply quit when it's dead.

There are also a few mentions of SSDs slowing down as they fill up. How true this is remains to be seen (as this was a common hearsay ~10 years ago with current hard drive types). The only substance seems to be enterprise level SSDs having much more "hidden" (for lack of a better word) space to allow for 'pinch hitters' when certain sections start to slow/fail than a standard consumer drive (numbers from a quick search seem to be 15-25% hidden for enterprise and 0-7% for consumer).

Finally, the one comment that I have seen repeatedly is avoid cheap drives and avoid OCZ's lower end. I hear nothing but horror stories about off-brand or inexpensive drives (being used for critical things ...dumbasses). It would also be a HIGHLY advisable suggestion to try and RAID the drives. RAID1 for small number (2-3) or RAID5 for three and above. The mirroring or drive replacement allows you to pop another drive in when one fails and not have to "OHSHI" restore from back up.


-Hate



1 - This is not to imply that standard HDD is bad or that this is currently happening. Simply stating that data corruption occurs and slightly increases as the drive is pushed past it's rated life span. This should be nothing new though.
Logged

I'm so sorry Hunter, I tried...
Crabthewall
Sr. Member
****
Posts: 293


View Profile
« Reply #3 on: April 02, 2013, 06:03:48 pm »

Going to sound presumptuous, but are you considering it for server? I ask because I've heard quite a few horror stories of SSDs being used for highly active databases and the drives keeling over extremely fast. I have also heard stories of people popping a bone for SSDs as DB drives. It seems to be case by case thing at the moment. The main difference that I do remember from the two is that a standard HDD will slowly die down (decreasing in performance and slowly increasing chance for data corruption1) whereas the SSD will work one and simply quit when it's dead.

There are also a few mentions of SSDs slowing down as they fill up. How true this is remains to be seen (as this was a common hearsay ~10 years ago with current hard drive types). The only substance seems to be enterprise level SSDs having much more "hidden" (for lack of a better word) space to allow for 'pinch hitters' when certain sections start to slow/fail than a standard consumer drive (numbers from a quick search seem to be 15-25% hidden for enterprise and 0-7% for consumer).

Finally, the one comment that I have seen repeatedly is avoid cheap drives and avoid OCZ's lower end. I hear nothing but horror stories about off-brand or inexpensive drives (being used for critical things ...dumbasses). It would also be a HIGHLY advisable suggestion to try and RAID the drives. RAID1 for small number (2-3) or RAID5 for three and above. The mirroring or drive replacement allows you to pop another drive in when one fails and not have to "OHSHI" restore from back up.


-Hate



1 - This is not to imply that standard HDD is bad or that this is currently happening. Simply stating that data corruption occurs and slightly increases as the drive is pushed past it's rated life span. This should be nothing new though.

EZ Server in it's current generation has been running on a SSD for the database. That's what helps keep it so fast - the os and rest of the data are in a raid 1. It is not a cheap SSD however, that isn't to say it doesn't need maintenance from time to time, but it will show it's usable life cycle if the proper tools are loaded and used. Instructions have been sent for that in particular.
« Last Edit: April 02, 2013, 06:18:14 pm by Crabthewall » Logged
Noel
Newbie
*
Posts: 48


View Profile
« Reply #4 on: April 02, 2013, 06:26:38 pm »

The answer depends on the type of the flash chips used by the maker of the drive. Furthermore, beyond the technical limit on number of reads and writes one can do before exhausting the resource life, it truly matters how the controller manages the data. Entire companies have been started around algorithms for writing data to SSDs. Example, if your controller is "dumb" - it could write data into the same cells, as long as they are marked as free (so you write data, you delete data, you write data, repeat many times). If it does that to the same cell over and over - the cell may go out. You do that on a big portion of cells and your drive is toast. Now, the new generation of controllers is data/space aware and use algorithms to minimize the chance of such activity by spreading reads and writes across all cells.

So the answer your question - it depends! To know the true answer, one would need to know your exact hardware, chip used on that drive, what algorithm it uses, how data is changing and why type of memory chips are installed. Enterprise grade single SSDs are MLC type, yet new gen of controllers combined with SLC appear to be showing comparable life expectancy.

Logged
Hunter
EZ Server GM
Legend
*******
Posts: 8100


EZ Server GM


View Profile
« Reply #5 on: April 02, 2013, 06:32:24 pm »

So what if you read the same file located in the same cell 10,000x
Logged

Hunter - EZ Server GM
Crabthewall
Sr. Member
****
Posts: 293


View Profile
« Reply #6 on: April 02, 2013, 06:36:24 pm »

So what if you read the same file located in the same cell 10,000x

Simple answer - no impact to the cell.

If you wanted to get super technical no appreciable impact to the cell which is slightly different.
« Last Edit: April 02, 2013, 06:39:14 pm by Crabthewall » Logged
Hunter
EZ Server GM
Legend
*******
Posts: 8100


EZ Server GM


View Profile
« Reply #7 on: April 02, 2013, 07:02:08 pm »

That helps a ton, thanks much!
Logged

Hunter - EZ Server GM
Noel
Newbie
*
Posts: 48


View Profile
« Reply #8 on: April 03, 2013, 01:47:53 am »

reading is not a problem until the cells are very "worn" and voltages read start to look like "noise", but even then error correction can help quite a bit.
Logged
Pages: [1]
Print
Jump to:  

Recent

Stats

Members
  • Total Members: 6156
  • Latest: Rin1
Stats
  • Total Posts: 65127
  • Total Topics: 5066
  • Online Today: 298
  • Online Ever: 8678
  • (December 19, 2022, 02:32:09 pm)
Users Online
Users: 0
Guests: 363
Total: 363
TinyPortal v1.0 beta 4 © Bloc