IBM C4C 4.51 3.5-inch Small FF 38-pin Unitized manual Comments, Data Transfer to/from SSA Link

			U S E R RESPONSIBLE F O R V E R I F Y I N G VERSION A N D COMPLETENESS
O E M F U N C T I O N A L S P E C I F I C A T I O N						U L T R A S T A R X P ( D F H C ) SSA M O D E L S 1.12/2.25 G B - 1.0" H I G H
	Model Type			All			C4x		C2x		C1x
	Notch #		Instant.		Track		Theoretical	Typical	Theoretical	Typical	Theoretical	Typical
	Average		12.07		7.91		7.17	7.13	7.13	7.10	7.06	7.03
	1		12.58		8.30		7.52	7.48	7.48	7.44	7.40	7.37
	2		12.58		8.30		7.52	7.48	7.48	7.44	7.40	7.37
	3		12.51		7.99		7.22	7.18	7.18	7.15	7.11	7.08
	4		11.96		7.74		7.02	6.99	6.99	6.95	6.92	6.89
	5		11.26		7.38		6.66	6.63	6.63	6.60	6.57	6.54
	6		11.05		7.07		6.41	6.38	6.38	6.35	6.31	6.28
	7		10.64		6.88		6.23	6.20	6.19	6.16	6.13	6.10
	8		10.29		6.64		6.03	6.00	6.00	5.97	5.94	5.91
	9		10.01		6.45		5.85	5.83	5.83	5.80	5.77	5.74
	10		9.59		6.15		5.55	5.53	5.53	5.50	5.48	5.45
	Note: The values for Typical Data Sector Transfer Rates assume a typically worst case value of 3.16 errors in 109 bits read at
	nominal conditions for soft error rate.
	Note: Contact an IBM Customer Representative for values when formatted at other block lengths.
	Note:	"Average" values are sums of the individual notch values weighted by the number of LBAs in the associated notches.
	Table	7. D a t a Sector Transfer Rates. (All rates are in MB/sec)

Data Transfer to/from SSA Link

The time required to transfer data between the SSA link and the drive's internal data buffer, that is not overlapped with the time for the Seek, Latency or Data Transfer to/from Disk.

When the drive is reading, data is transferred from the medium to its data buffer and from the buffer across the SSA link simultaneously. However, data transfer to the link from the data buffer buffer lags transfer from the medium to the buffer by one block. At the end of the transfer from the medium, one block still has to be transferred across the link.

For a write operation, the data is normally transferred to the data buffer during the seek and latency time. In the rare case that these are both zero, the write cannot begin until one sector is transferred, and the time to do this becomes part of the overhead.

Each block of data is transferred as one or more frames on the SSA Link. Each frame requires 10 bytes of overhead and may contain up to 128 bytes of data. The time to transfer one block depends on the number of frames required. For example, a 744 byte block needs 6 frames (5 x 128 byte, 1 x 104). This adds 60 bytes of overhead making 804 bytes total. At an instantaneous transfer rate of 20MB/s, that is 40 microseconds per block (17.7MB/s sustained).

3.3.2 Comments

Overlap has been removed from the Command Execution Time calculations. The components of the Command Execution Times are truly additive times to the entire operation. For example,

The Post Command Processing times are not components of the Command Execution time therefore they are not included in the calculation of environments where the re-instruct period exceeds the Post Command Processing time.

The effects of idle time functions are not included in the above examples. The 3.2.1, “Sequential” on page 40 and 3.2.2, “Random” on page 40 both define environments where the effects due to increased command overhead of Idle Time Functions upon Command Execution time are less than 0.15%.

Page 42 of 87 I B M Corporation

Source filename=PERFORM