[GRLUG] NOT LINUX - hybrid drives

[Adam Tauno Williams]

Bob Kline <bob.kline at gmail.com> wrote:
>http://www.amazon.com/gp/product/ref=pe_172730_26630760_pe_b2/?ASIN=B003NSBF32
>This seems like a promising concept.
>Keep everything needed to get up and
>going, and the most often used software, in
>the faster SSD - faster read that is.

This isn't new.  Higher end SAN and NAS devices have offered this feature for some time.  Hot-block tracking keeps more frequently used blockes on faster storage and migrates cold blocks to lower performance / higher capacity tiers.

>4GB might be a bit small, but at some
>point this could speed up a system.

Or it is plenty.  If the redistribution of blocks is transparent this might be sufficient.  In the life cycle of even very large applications it is very common to see concentrated I/O;  mostly this is about write performance (as read performance for real I/O setups is easily accelerated by huge buffers).   This is much the same as the *BIG* performance payoff one sees by simply putting a database's journal on separate spindles (or an SSD) from the data (which remains on spinning rust in the SSD case).

>Conceptually, the "sticky bit" in file
>permissions was intended for just that
>purpose, 

I think the sticky bit was more about trying to control caching.

>but the process could, and
>apparently has been, automated to a
>large extent so that the system does
>the job.



-- 
Adam Tauno Williams, LPI1 / NCLA