Hi,
Quote:
Originally Posted by ajohn
I'm in the process of upgrading my desktop and am interested in using an ssd disc for system and program files but I never want to write to it except for updates. My home directory and it's associated temp will be on a hard drive as will swap but I am not clear what temp files the os and various programs make use of.
The reason for never writing to it is based on using the technology in the automotive field. Generally data retention time for reads is enormous but have to admit that very low nm technology generally isn't use there. As to writes there is some info here
https://wiki.archlinux.org/index.php/Solid_State_Drives
It struck me that as linux can be booted from a usb stick there must be a mechanism for arranging all temp files where ever they go to or come from to end up in ram. In some ways I would prefer the option of a hard disc but am not sure how to set up either. If they go to ram there would be a need to clear them out automatically really. I don't reboot my machine very often. Currently the systems memory of files in temp seem to be shorter than their lifetime on disc. I'm think particularly of something I often do - click and open a pdf from the web. These go to root tmp. Afraid I am not aware if something ever removes them. There are also many many folders and other bits and pieces in it.
Any ideas as to what would need to be done? I will probably be running opensuse 12.3 KDE and am not much of a kernel person even thought I have used linux for a long long time.
John
-
|
The linked Arch wiki information is based on older technology and should be brought up to date. I suggest that you look at
SSDWiki;
Quote:
I selected Force as an example;
Force Series
The Corsair Force Series is based on SandForce SF-1200 controllers with a maximum sequential read speed of 285MB/s and write speed of 275MB/s, and 50,000 IOPS (4K). It is available with both 34nm and 25nm NAND flash modules, the latter of which have a decreased user-accessible capacity (over-provisioning) but a lower price due to durability issues with flash memory produced with the 25nm production process. It is available in capacities ranging from 60GB to 240GB, with no discernible difference in performance between the different capacities, which is a common problem in other drives.
Specification:
- Sequential Read: Up to 280 MB/second
- Sequential Write: Up to 270 MB/second
- NAND type: MLC
- Interface: SATA 2
- Controller: SandForce SF-1200
- Form Factor: 2.5 inch
- DRAM Cache: N/A
- Power Consumption (Active): 2W
- Power Consumption (Idle): 0.5W
- TRIM support: Yes
- SMART support: Yes
- Warranty: 3 years
|
Depending on the device manufacture and your configuration for operation to get the max life out of the modern SSD you should still create backups. Newer SSD controllers along with density will dictate life. Wear level for writes for the life is around 100,000,000, or (the warranty of 3-5 years for most new SSD) to me that will never exceed the life of the system since it will be replaced long before the 50+ years for the typical expected life of the SSD in a normal operated system. Sure older SSD technology had a shorter life span as compared to modern MLC or SLC based drives with modern SSD controllers. Be sure to compare
IOPS for the SSD to get the max I/O.
Life for a modern SSD is not a real concern for me since the newer technology has extended the span if configured properly for a system.
Be sure to use a newer kernel that supports 'TRIM' along with configuring the scheduler for the drive, I prefer [noop] scheduler for the SSD since it is a FIFO and you will not have the latency that a spinning disk will have so the FIFO [noop] will perform better than [deadline] or [cfq]. Just assign [noop] for the SSD not system wide.
Check out this LQ post: #
26 for references & links.
HTH!