JPS6132704B2 - - Google Patents

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention provides an access control method for a magnetic disk device that reduces the burden on hardware on the control device side and that does not cause discontinuity due to head switching during continuous processing. It is related to.

[Technology background]

A sector format is generally used as a file recording format in a small computer system.

One problem with this sector format is how to allocate record addresses.

Generally, when allocating n records on a magnetic disk, a method of sequentially allocating records as shown in FIG. 1 is used.

In the diagram, R ₀ to _{R o-1} are record numbers, G1 to G3
is a gap, AA is an address area, DA is a data area, and S is a sector mark.

The most common length for one record is 128 bytes or 512 bytes, because if it is too long, a large amount of space will be required on the main memory, and the loss will be large when handling a small amount of data. used.

One record consists of an address, as shown in Figure 1b.
Gap G1 between Area AA and Data Area DA
~ It is placed between G3.

This method of allocating magnetic records is extremely simple and generally has the least disadvantages, but when connecting a high-speed disk drive to a small computer system, access to the main storage device is occupied by high-speed data transfer, causing the computer to Program execution stops and the overall efficiency of the computer decreases.

When reading the address area of the record following the record in which data has been written, it is necessary to provide a large gap G1 to avoid the influence of the writing process, resulting in a reduction in storage capacity. (See Figure 3a.) It is necessary to switch the head to read or write a track other than the track currently being read or written on the same cylinder, but after switching the head, when reading the first record, the head In order to avoid the influence of the switching time, it is necessary to make the gap G1 of the record near the sector mark S large. (See FIG. 3b.) There are problems such as these, and a method to solve these problems is desired.

[Conventional technology and its problems]

Conventionally, in order to solve the above problems, the second
The method shown in the figure was adopted. That is, this is a method of arranging record numbers every other record as shown in FIG.

As shown in the figure, after allocating every other record from R ₀ to Rn/2-1, the remaining records R
Allocate a total of n records from n/2 to R _o-1 .

According to this method, in the case of continuous processing of records that are completed only within one track, there is always one
Since there is a gap for each record, the problem ~ mentioned above that was a problem with the method shown in Figure 1 is solved, but there are cases where continuous processing of records is not completed in one track, and in this case, In order to prevent head movement, it is common practice to continuously process records placed on different tracks in the same cylinder following the above process. The problem is not resolved. That is, when switching heads, the next record is not processed every other record, but the next record is processed continuously.

Another problem that arises is that when the number of records is even, every other record from record _R0 to record Rn/2-1 is processed, and this record R
When moving the process from record n/2-1 to record Rn/2, it becomes necessary to skip two records for control purposes, and 1
There are cases where control is performed by skipping one track, and cases where processing is performed by skipping two tracks, and this process is performed each time each track is read or written, which not only complicates control, but also reduces data transfer per unit time. The disadvantage is that the rate is reduced. (See Figure 3c.) [Means for Solving the Problems] The present invention aims to solve the above problems. Address information consisting of a head and a sector address is instructed by the host device, and the magnetic head is moved and positioned to select one cylinder from the plurality of cylinders, and one head is selected from the plurality of heads on the cylinder and tracked. The system is configured to select at least one sector from a plurality of even sectors on the track and perform access for reading and writing data, and sequentially input sector addresses and head addresses to the same cylinder. - An access control method for magnetic disk devices in which each track on the same cylinder and each sector on that track is controlled to read and write sequentially and continuously without moving the magnetic head by specifying an address. In the device, two logical cylinder addresses are apparently assigned to one physical cylinder of the magnetic disk device, and an even number of physical sector addresses arranged on one physical track are assigned. Apparently, half of the physical sector address is allocated, and the lower bits of the cylinder address information made up of multiple bits of the address information instructed by the higher-level device are used as the least significant bits of the sector address made up of multiple bits. Add as,
By accessing the magnetic disk device, every other sector on each track physically located on the same cylinder is sequentially accessed, and the head addresses are sequentially switched to access each sector on the same cylinder. Accesses every other sector, and then accesses the remaining sectors on each track physically located in the same cylinder as the cylinder without moving the magnetic head based on instructions from the higher-level device that arrive after that. This is achieved by sequentially accessing every other sector and sequentially switching head addresses so that the remaining sectors physically on the same cylinder are accessed every other sector.

〔Example〕

Hereinafter, an access control method for a magnetic disk device according to the present invention will be explained using an embodiment of the present invention.

First, before explaining the present invention, the configuration of a general magnetic disk device will be explained with reference to FIG. 4 in order to facilitate understanding of the present invention.

Generally, in a magnetic disk device, as shown in FIG. 4, a plurality of magnetic disk disks 2 are arranged on the same rotating shaft 1, and this rotating shaft is rotated at a predetermined speed by a rotary drive motor (not shown). The recording surface of the magnetic disk disk 2 is composed of a recording surface 3 on which servo information for moving and positioning the magnetic head to a desired track is recorded, and information recording surfaces 4 to 7 on which actual information is recorded. .

A magnetic head (not shown) is mounted on the same carriage (not shown) and arranged on each of the recording surfaces 3 to 7, and the magnetic heads (not shown) are arranged in accordance with the positioning information read from the recording surface on which the servo information is recorded. They are moved and positioned to a desired track by being commonly driven by an access motor.

Therefore, the magnetic heads arranged on recording surfaces 4 to 7 on which information is recorded are arranged at the same track position on each recording surface, and a set of identical track numbers on each surface is a cylinder. This cylinder number (SYL) is selected by the host device.

Each track in the same cylinder is selected by a head number (HD).

Each track has a plurality of records recorded in the form of a plurality of sectors, and these are selected by sector (SEC) numbers.

In order to read and write desired information from the host device, a cylinder number, head number and sector number are given.

First, the magnetic head is moved and positioned to a desired cylinder based on the cylinder number, and then read and written to the sector of the desired information surface is performed based on the head number and sector number.

On the other hand, when reading and writing information from a plurality of sectors in succession, the information is arranged so that the movement time of the magnetic head is minimized to reduce access time as much as possible.

That is, the cylinder number, head number, and sector number are first given, and after the desired cylinder and head are moved and positioned, the head number is first fixed, and the sector number of a given track is sequentially changed to make the same track. Read and write the upper sectors sequentially,
Next, the head number is switched and sectors arranged on tracks on different recording surfaces of the same cylinder are sequentially read and written.

To explain this with reference to FIG. 4, for tracks a to d belonging to the same cylinder, first read and write sectors of track a on the recording surface 4, then read and write sectors of track b, and then read and write sectors in the order of c and d. Controlled to read and write.

The present invention provides an access control method suitable for simplifying control when continuously reading and writing to a plurality of sectors as described above. FIG. 5 is a diagram for explaining an example of reading and writing when sequentially processing sectors according to the present invention.

In the figure, FIGS. 4a to 4d represent each track a of interest on the same cylinder shown in FIG.
~d are shown correspondingly.

In this example, track a on recording surface 4 arranged on the same cylinder shown in FIG. 4, track b on recording surface 5, track c on recording surface 6,
It is assumed that processing is performed in the order of recording surface 7 tracks d.

In this case, first read and write every other sector on the track shown in Figure 5a starting from _R0 , and when reading and writing up to _R0-2 are completed, the head is switched and the sector on the track B located on the next recording surface is read and written. Read and write from R ₀ to _{R o-2} .

Then, the heads of the sectors on the track shown in _FIG .
Reading and writing are performed up to R _{o -2} , and when reading and writing are completed, the head is switched and reading and writing are performed from R _o to _{R o -2} of track d located on the next recording surface.

After processing every other sector on each track on the same cylinder in this way, return to the first track a and process one sector from R ₁ to R _o-1 for the remaining sectors that were not initially read or written. Then, in the same way as described above, the head is switched and read and write sectors are sequentially read and written every other sector.

In this way, by processing each sector on each track on the same cylinder, the record will always jump by one even at the time of head switching, so there is no need to increase the gap G due to the head switching time. Furthermore, when the number of sectors arranged on the same track is even, it becomes extremely rare for two sectors to be skipped for control reasons, and the amount of data transferred per unit can be increased.

FIG. 6 is a diagram illustrating the address allocation method according to the present invention. In the present invention, FIG.
The conventional layout method shown in FIG. 6 is apparently changed to the layout method shown in FIG. 6b.

That is, a plurality of records that physically exist on the same track are apparently stored in the same track as shown in FIG. 6b.
The controllers are controlled by a higher-level controller as if they were placed on alternately adjacent separate tracks.

For this purpose, the command given from the higher-level device is
The configuration is such that the number of tracks is doubled and the number of records existing in one track is halved to receive commands.

Therefore, A in FIG. 6b is assigned, for example, with cylinder address CYL=0 and head address HD=0, and B in FIG. 6b is assigned with cylinder address CYL=0 and head address HD=0.
It will be allocated with CYL=1 and head address HD=0. In this way, since the address actually accessed (physical address) and the virtually allocated address (logical address) are different, it is necessary to convert them in the control device DPC.

Now, as shown in Figure 7, for example, the cylinder address is sent from the CPU to the control device DPC.
If you specify CYL1628, head address HD1 and sector address SEC30,
Since these are logical addresses, the control device
In DPC, the cylinder address CYL is halved and the sector address SEC is doubled to access the magnetic disk unit DPU using a physical address.

The operation on the circuit in the control device DPC is
As shown in the figure, after transferring the least significant bit of the cylinder address CYL instructed by the CPU as the least significant bit of the sector address SEC,
Just send it to the device DPU side along with the head address HD.

Therefore, the hardware configuration on the control device DPC side can be extremely easily realized by providing a gate circuit that only transfers bits as shown in FIG.

FIG. 9 is an example of an address conversion circuit in the control device DPC.

In the figure, the circuit is constructed in units of 8-bit registers, and head address registers HD and REG are omitted.

As shown in Figure 8, the cylinder address
When CYL is up to 2048, 11 bits are required, so two 8-bit registers are used, and when sector address SEC is up to 32,
Uses 5 bits of an 8-bit register.

First, the cylinder address CYL is sent to the device bus side using eight buses twice.

That is, after opening the NAND gate with the cylinder address sending timing signal TM and sending out a total of 8 bits, the 0th to 6th bits of the cylinder register CYL・REG and the 7th bit of the cylinder register CYL・REG, the cylinder register CYL・
The 5th and 6th bits of REG are sent twice. As a result, among the 11 bits of the cylinder address CYL obtained from the host device, the least significant bit (7th bit of the cylinder register CYL REG) is excluded, and the cylinder address CYL is sent to the device DPU side.

Next, as the sector address SEC, the 7th bit, which is the least significant bit of the cylinder address, and the 3rd to 7th bits of the sector register SEC REG are set to 6 in synchronization with the sector address sending timing signal TM. Device through book bus
Send to DPU.

In this way, the cylinder address
CYL2048, each logical address up to sector address SEC32 is assigned to CYL1024,
After being converted to each physical address up to SEC64, it is sent to the device DPU side, sharing eight buses.

〔effect〕

As explained above, according to the present invention, records on the same track appear to have two cylinders.
Since it is divided into addresses and logically assigned sector addresses to each, when processing each sector continuously, in most cases it is possible to process every other record, making control extremely simple. At the same time, it is possible to improve data transfer efficiency per unit time.

[Brief explanation of the drawing]

FIG. 1 and FIG. 2 are explanatory diagrams showing the conventional record address allocation method, FIG. 3 is an explanatory diagram showing the problems in FIGS. 1 and 2, and FIG. FIG. 5 is an explanatory diagram showing an overview of such a magnetic disk device, FIG. 5 is an explanatory diagram showing the concept of the magnetic disk access control method of the present invention, and FIG. 6 is an explanatory diagram showing an example of the record address allocation method of the present invention. FIG. 7 is a system diagram of addressing signal transmission and reception showing an embodiment of the present invention. FIG. 8 is an explanatory diagram of address conversion showing an embodiment of the present invention.
The figures each show a connection diagram within the control device having the address translation means of FIG. 8. DPC: Magnetic disk control unit, DPU: Magnetic disk unit, CYL: Cylinder address, HD:
head address, SEC: sector address,
AA: address area, DA: data area, S: sector mark, G1, G2, G3: gap, 1: rotation axis, 2: magnetic disk disk,
3, 4, 5, 6, 7: magnetic recording surface.

Claims (1)

[Claims] 1 Address information consisting of a cylinder address, a head address, and a sector address each consisting of a plurality of bits is instructed by a host device, and the magnetic head is operated to select one cylinder from a plurality of cylinders. move and position, select one head from a plurality of heads on the cylinder to select a track, select at least one sector from an even number of sectors on the track, and read/write data. By sequentially specifying the sector address and head address for the same cylinder, each track on the same cylinder and each sector on the track can be accessed sequentially without moving the magnetic head. In an access control method for a magnetic disk device that is controlled to read and write continuously, the host device apparently assigns two logical cylinder addresses to one physical cylinder of the magnetic disk device, and , apparently allocating half logical sector addresses to an even number of physical sector addresses arranged on one physical track, and assigning the above address information consisting of a plurality of bits among the address information instructed by the host device. Adding the lower bit of the cylinder address information as the lowest bit of the sector address made up of the plurality of bits,
By accessing the magnetic disk device, every other sector on each track physically located on the same cylinder is sequentially accessed, and the head addresses are sequentially switched to access each sector on the same cylinder. Accesses every other sector, and then allocates the remaining sectors on each track physically located in the same cylinder as the cylinder without moving the magnetic head based on instructions from the higher-level device that arrive after that. An access control method for a magnetic disk device characterized by sequentially accessing every other sector and sequentially switching head addresses to access every other remaining sector physically on the same cylinder.