« Back to articles
Solid-state drives are storage devices using memory modules. By emulating a hard disk interface, SSD devices can be used as if they were a normal mechanical drive. In the last years solid-state drive are becoming more common in notebooks, desktop and server systems.
Although SSD technology seems to be fairly new, it already exists for more than 20 years. The first generation of SSD's where used in supercomputers and of course technology has been improved over the years. Current SSD vendors use often non-volatile flash memory, which do not require batteries and are better protected for power outages.
When comparing traditonal HDD technology with SSD technology, the differences are big in favor of SSD. But when it comes to comparing prices per GB, SSD suddenly becomes the big loser.
- No moving parts
- Less fragile
- Very low access time (also for random access)
- Low latency
- Performance relatively constant (file fragmentation is less of an issue as well)
- Low amount of energy needed
- No spin-up time needed
- Larger range of temperature to operate
- High price (per GB)
- Alignment issues with older operating systems
- Limited write cycles (per block)
SLC vs MLC
The cells used in flash technology are usually SLC (single-level cell) or MLC (multi-level cell), with the biggest difference the amount of bits stored per cell. Single-level cells store only one bit, multi-level usually 3 or more.
When comparing SLC and MLC, they have the following characteristics:
- SLC is faster and has around 100,000 up to 1 million write cycles. This is pretty similar to hard disks.
- MLC is slower but also cheaper. The reliability is about 1/10th of SLC (sometimes close to only 10,000 write cycles).
Flash and the RIP effect
Flash based memory chips are becoming faster and faster, but they will have one big advantage, a limited amount of write cycles. Depending on the cell type, the quality and the used conditions, flash cells will die.
Even though dying cells are a risk for the data integrity, this happens to individual cells and does not apply to the whole disk at once. Manufacturers minimize the risk by using a term called wear leveling, nothing more than spreading data writes, instead of using the same cells again and again. Wear leveling has a speed penalty for MLC writes, since data which could be stored together will be spreaded instead. Another protection measure is adding spare cells, which replace 'dying' cells.
Last updated by Michael Boelen at 29 June 2009
» FreeBSD Easy Installation Generator
» Rootkit Hunter
» Architecture Best Practices
» OpenLDAP book (LPIC-3, 117-301)
» SNIA SCSE (S10-201)
» About me
"A master piece of software and a must for every server
admin." - Jose
Tip: Want to know everything about the latest changes and additions?
» Join the notification list