Seagate ST3250620AS - Barracuda 250GB 7200 RPM 16MB Cache SATA 3.0Gb/s Perpendicular Recording Hard Drive, Руководство пользователя

Страница: 3 / 12

Serial ATA Native Command Queuing
sides of the discs, is collectively called a cylinder. Thus, data is laid out across the discs
sequentially in cylinders starting from the outer diameter of the drive.
One of the major mechanical challenges is that applications rarely request data in the order that it
is written to the disc. Rather, applications tend to request data scattered throughout all portions
of the drive. The mechanical movement required to position the appropriate read/write head to
the right track in the right rotational position is non-trivial.
The mechanical overheads that affect drive performance most are seek latencies and rotational
latencies. Both seek and rotational latencies need to be addressed in a cohesive optimization
algorithm.
The best known algorithm to minimize both seek and rotational latencies is called Rotational
Position Ordering. Rotational Position Ordering (or Sorting) allows the drive to select the order of
command execution at the media in a manner that minimizes access time to maximize
performance. Access time consists of both seek time to position the actuator and latency time to
wait for the data to rotate under the head. Both seek time and rotational latency time can be
several milliseconds in duration.
Earlier algorithms simply minimized seek distance to minimize seek time. However, a short seek
may result in a longer overall access time if the target location requires a significant rotational
latency period to wait for the location to rotate under the head. Rotational Position Ordering
considers the rotational position of the disk as well as the seek distance when considering the
order to execute commands. Commands are executed in an order that results in the shortest
overall access time, the combined seek and rotational latency time, to increase performance.
Native Command Queuing allows a drive to take advantage of Rotational Position Ordering to
optimally re-order commands to maximize performance.
Seek Latency Optimization
Seek latencies are caused by the time it takes the read/write head to position and settle over the
correct track containing the target Logical Block Addressing (LBA). To satisfy several commands,
the drive will need to access all target LBAs. Without queuing, the drive will have to access the
target LBAs in the order that the commands are issued. However, if all of the commands are
outstanding to the drive at the same time, the drive can satisfy the commands in the optimal
order. The optimal order to reduce seek latencies would be the order that minimizes the amount
of mechanical movement.
One rather simplistic analogy would be an elevator. If all stops were approached in the order in
which the buttons were pressed, the elevator would operate in a very inefficient manner and
waste an enormous amount of time going back and forth between the different target locations.
As trivial as it may sound, most of today’s hard drives in the desktop environment still operate
exactly in this fashion. Elevators have evolved to understand that re-ordering the targets will
result in a more economic and, by extension, faster mode of operation. With Serial ATA, not only
is re-ordering from a specific starting point possible but the re-ordering scheme is dynamic,
meaning that at any given time, additional commands can be added to the queue. These new
commands are either incorporated into an ongoing thread or postponed for the next series of
command execution, depending on how well they fit into the outstanding workload.
To translate this into HDD technology, reducing mechanical overhead in a drive can be
accomplished by accepting the queued commands (floor buttons pushed) and re-ordering them to
efficiently deliver the data the host is asking for. While the drive is executing one command, a
new command may enter the queue and be integrated in the outstanding workload. If the new
command happens to be the most mechanically efficient to process, it will then be next in line to
complete.
3