Architecture

Figure 1: (a) Mover and microactuators. The device is shown from above; the tip array lies below the mover surface. (b) A probe-based storage device that includes one mover (the shaded areas) above a tip array. The mover is divided into four clusters, each of which has 12 dedicated tips. One tip per cluster is active at a time (the shaded tips), accessing the tip areas depicted by the dark rectangles. The mover has limited range of movement over the tip array in the $X$ and $Y$ directions.

(a)

(b)

Figure 1a is a top view of a typical probe-based storage device. In this figure, the shaded parts move and the unshaded parts are stationary. The media mover is suspended above a surface on which a grid of many probe tips are embedded. Collectively, the tip array is the logical equivalent of the read/write heads of a traditional disk drive. Voltage applied to the fingers of the microactuator combs exerts electrostatic forces on the mover that cause it to move in the $X$ and $Y$ directions, overcoming the forces exerted by the anchors and beams that keep it in place. To service a read or write request, the mover first repositions itself so that the tip array can access the required data. This repositioning time is called seek time. The mover then accesses the data while moving at a constant velocity in the $Y$ direction, incurring transfer time.

Figure 1b magnifies the mover area from Figure 1a. This shows that a mover is divided into one or more clusters. Each cluster can read data independently of the others, which provides higher parallelism to the device. As an example, the mover in Figure 1b is subdivided into four clusters. Each cluster is a media area that is accessed by many tips, only one of which can be active at a time. Using several tips in parallel, one from each cluster, compensates for the low data rate of each individual tip, which is on the order of 1Mbit/sec. The number of bits accessed simultaneously is equivalent to the number of clusters per mover times the number of movers in the device. We call this the number of active tips.

The mover's range of movement and the bit size determine the amount of data that can be manipulated by one tip, or tip area. Because several tips are active at a time, different tip areas of the mover are manipulated simultaneously, as depicted by the shaded rectangles in Figure 1b. Different areas of the mover are accessed by switching between sets of active tips.

Table 1: Configurable and physical probe-based storage architecture parameters.

Configurable	Default	Symbol
parameter	value
Movement range
in $X$ (5-80$\mu$m)	40$\mu$m	$\delta_x$
in $Y$ (5-80$\mu$m)	40$\mu$m	$\delta_y$
Active tips (10-2560)	320	$T_{active}$
Tips in $Y$
per cluster(1-20)	10	$T_y$
Physical
parameter
Settle time	200$\mu$s	$t_{settle}$
X-move time	1ms	$t_{XM}$
Velocity	0.05m/s	$v_0$
Turn around time	400$\mu$s	$t_{TA}$
Tip change time	0	$t_{TS}$
Bit width	50nm	$d_b$
Acceleration	250m/${s^2}$	$a_0$
Active tips per cluster	1	$T_{cluster}$
Workload
parameter
Average request size	calculated	$r$
	from workload
Runlength	calculated	$r_l$
	from workload

Many architectural configurations are possible for probe-based storage. For example, we might vary the number of active tips, the mover's movement range, the media density, and so on. The values summarized in Table 1 are reasonable defaults for probe-based storage devices [25,6,1], and are the parameters we used in simulations unless otherwise specified.