EP0438556A4

EP0438556A4 - Variable capacity cache memory

Info

Publication number: EP0438556A4
Application number: EP19900911237
Authority: EP
Inventors: Daniel Esbensen
Original assignee: TOUCH TECHNOLOGIES Inc
Current assignee: TOUCH TECHNOLOGIES Inc
Priority date: 1989-07-14
Filing date: 1990-07-13
Publication date: 1993-03-17
Also published as: WO1991001525A1; CA2036388A1; EP0438556A1; AU6075790A

Abstract

A variable length cache system (58, 60) keeps track of the amount of available space on an output device (10). The capacity of the cache system (60) is continuously increased so long as it is less than the available output space on the output unit (10). Once the size of the cache system exceeds the available output space on the output unit (10), which is less than the total space available on the output unit (10) by a predetermined amount, the contents of the cache memory (60) are flushed or written to the output device (10) and the size of the cache memory (60) is reduced to zero.

Description

VARIABLE CAPACITY CACHE MEMORY

FΪELD

This invention relates in general to methods for maximizing the efficiency of bulk data transfers from computer operating systems to peripheral devices, i.e., write operations, which store data serially such as tape devices, and particularly to such methods based on caching the write operations so they can be performed concurrently with other system operations, such as application program execution.

BACKGROUND OF THE INVENTION

The use of magnetic tape drives for bulk storage in computer systems is well known. While useful, such devices have undesirable characteristics that must be minimized. For purposes of this specification, the terms "tape device" and "tape devices" shall refer to and mean data storage devices which store and access data serially, whether in "streaming" fashion or "start/stop" fashion, on or in a storage media of finite known length, such as magnetic tape devices

(e.g. reel-to-reel and cartridge types), optical disks configured as tape devices, helical scan drives, and the like; and the terms "concurrent" and "concurrently" in the context of describing the timing of events controlled by a computer operating system shall also mean the overlapping or time multiplexing of such events, such as multiple programs being executed by the operating system apparently simultaneously. For purposes of this specification, the term "write head" shall refer to and mean the instrument by which data is actually recorded upon a tape or tape-like storage medium. For example, magnetic tapes can be recorded with a magnetic write head and optical disks configured as tape devices can be recorded with an optical write head. In a computer system, the operational speed that data can be read, processed, and then written to a tape device is normally limited by the speed of the tape device. Advances in the pertinent art have increased many fold the access speed of disk devices and the processing speed of computers. Unfortunately increases in tape device speed have not kept pace. With faster disk reads and processing, whole operations are bottlenecked by the speed of the tape devices. It has been found that in prior art computer systems when data is to be written to a tape or tape¬ like storage medium, the input/output (I/O) operation requests made by an application program to the system involving the writing of data to the medium are sent as fast as possible by the operating system to the tape device controlling the medium. There is no attempt to regulate the rate at which such write requests are sent to the device driver. This causes the data and I/O reguests to back up and wait in a queue to be written to tape. This queue at some point causes the operating system or application program to stop and wait for the completion of one or more of the queued operations. This creates a sort of traffic jam for the data and an irregular write request rate. When this happens, the tape starts and stops over and over. This start and stop process is an agonizingly slow operation since the tape has to decelerate, rewind to the place where it last wrote data before decelerating (since it cannot write data while decelerating or accelerating) , and then accelerate again in order to write data at maximum speed. This start and stop process is expensive in terms of overall or elapsed time and therefore a costly process. In order to work at their peak throughout capacity, tape devices need a controlled flow, i.e., a steady flow of data and I/O requests to the tape sub¬ system.

In order to speed up the whole operation, the tape devices must be able to write data concurrently with the reading of disk data and the processing. For concurrency to occur, tlϊe data to be written must be put into a high-speed memory cache. It has been known in the prior art to use cache memories with output devices, such as tape drives, in both conventional memory management systems as well as in virtual memory systems.

A problem occurs when the tape reaches an end-of-media indicator (typically a reflector on the tape itself used as an end-of-tape marker) . At that point the cache must be flushed out or write to the media. If the cache contains more data than can fit on a length (e.g. 10 feet) of writable media beyond the end-of-tape marker (past-end-of-tape media) , some of the data will be lost or the tape device will signal a fatal drive error. Therefore the current art limits the size of the tape I/O data cache to the amount of data that can be written onto the past-end- of-tape media. This has the effect of limiting the number of concurrent tape write operations, which in turn limits the speed of the whole system.

It would be desirable to be able to use a cache memory larger than such a system has heretofore been able to use, i.e., larger than the data capacity of the past-end-of-tape media. It would also be desirable to be able to detect a near end-of-tape condition sufficiently in advance of the conventional end-of-tape indicator to allow the system to flush all of the cached I/O requests before the end of the writable portion of the medium is passed. The net result would be more concurrency than heretofore available, which in turn speeds up the whole system.

SUMMARY OF THE INVENTION

In accordance with the invention, a method and an apparatus are provided for dynamically changing the storage capacity, or size, of a cache memory used in the transfer of data to an output device which has a predetermined maximum storage capacity. The apparatus utilizes a maximum storage parameter which has a value a predetermined amount less the actual maximum storage capacity of the output device.

The apparatus determines the presently available storage capacity of the output device. This corresponds to the maximum storage capacity parameter reduced by the amount of information previously stored on or written to the device. The current capacity of the cache memory is compared to the available storage capacity of the output device. If the current capacity of the cache memory exceeds the available storage capacity of the output device, as described above, the contents of the cache memory are transferred to the output device.

If however the currently available storage capacity of the output device exceeds the current capacity of the cache memory, the apparatus increases the capacity of the cache memory by an amount corresponding to the next received block or file of information to be written to the output device. That information is then stored in the cache memory and the process is repeated. The cache memory could be implemented as a contiguous group of memory locations which are located either in or coupled to a central processor or which can be separately located in a displaced cabinet. Alternately, the cache memory can be implemented as a group of pages in a virtual memory management system. In this implementation, the size of the cache memory is limited only by the available pages which can be acquired via the virtual memory management system. Hence, in accordance with the virtual memory implementation, the practical size of the cache memory is limited only by the amount of space still available for storage on the output device. The output device can be any form of bulk storage medium such as magnetic tape drive or optical storage device.

The present method includes the steps of calculating the amount of output medium necessary to store the current contents of the cache memory. This value is added to the amount of the output medium which has previously been used. The sum can then be compared to the maximum storage parameter.

If the sum exceeds the maximum storage parameter the contents of the cache memory are immediately flushed or written to the output device. If the sum is less than the maximum storage parameter then the size of the cache memory can be increased by the size of the block or file of data which is to be next written into the cache memory. The comparison process is then repeated. In accordance with the present apparatus and method, the size of the cache memory can be increased to a maximum corresponding to the maximum output storage parameter before the contents of the cache memory are flushed or written to the output device. Numerous other advantages and features of the present invention will become readily apparent from the following detailed description of the invention and the embodiments thereof, from the claims and from the accompanying drawings in which the details of the invention are fully and completely disclosed as a part of this specification.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a block diagram representation of a portion of prior art computer system;

Figure 2 is a block diagram representation of a portion of another prior art computer system;

Figure 3 is a schematic diagram illustrating portions of a magnetic tape; Figure 4 is a block diagram representation of a computer system in accordance with the present invention;

Figure 5 is a block diagram representation of a computer system in accordance with the present invention with a second embodiment of the present invention;

Figure 6 is a block diagram representation of a computer system in accordance with the present invention with a third embodiment of the present invention; and

Figure 7 is an overall block diagram illustrating steps of a method in accordance with the present invention.

DESCRIPTION OF THE PREFERRED EMBODIMENT

It is typical in computer systems to use special operating system routines to perform the actual communication between the computer and peripheral devices. The special routines are commonly called device drivers. These routines are device dependent and do the actual I/O operations. These routines are typically called as macros and they are called in response to an I/O request.

Referring to Fig. 1, when an application program 2 requires that data be stored on a tape device, it establishes the data to be written in a data buffer 4. It then initiates an I/O request by one or more special instructions 6A. Among other things the I/O request contains information concerning the type of operation being requested and a vector to the data buffer. In a prior art system without a tape I/O request caching system installed this I/O request is essentially passed on to a low level routine, the device driver 8, associated with the target tape device 10. The device driver communicates first-hand with the tape device to accomplish to write operation during which the device driver controls the flow of data between the data buffer 4 set up by the application program and the tape device. Referring to Fig. 2, a tape I/O request a prior art caching system with a fixed length data cache is illustrated. In this system, the I/O request 6B initiated by the application program 2 is intercepted before it can pass to the device driver 8 and is placed in an I/O request queue 12. The data associated with the I/O request 6B is also moved to a data cache 14. From this point the I/O requests are sent to the device driver in an optimum regulated flow as determined by an algorithm. However, as previously explained with respect to other known prior art systems, the size of the cache is limited to the amount of data that can be written past the end-of- media indicator. In a virtual memory system, a fixed length cache can be implemented using "pages" of memory assigned by the system program. The present invention provides a variable size cache memory based on the amount of available space remaining on the storage unit, such as the tape drive 10. With respect to a tape storage unit, the steps of the present invention include: (a) monitoring by means of a first algorithm the physical position, within a tolerance, of the tape device's writing head relative to an end of the tape device's storage media; (b) monitoring by means of a second algorithm the length of the tape device's storage media that will be consumed when all of €he currently cached data is written to that media; and (c) whenever the length determined by the second algorithm plus the position of the tape device' writing head relative to a starting end of the media is greater than or equal to the total available writable length of the media, then stopping the caching of tape device I/O operations and flushing or writing the current contents of the cache memory to the tape device. Referring to Fig. 3, a tape or tape-like storage medium 20 as illustrated has data being recorded upon it serially in a direction from left to right, i.e., from the end designated "START OF TAPE" to the opposite end which is designated "END OF TAPE," by means of a writing head. The portion of the medium designated by "A" has already been consumed, i.e, data has actually been recorded upon it. This already- consumed portion extends from the start of the medium to the current physical position of the medium's writing head which is designated "HEAD." The portion of the medium designated by "B" is a portion that has logically been consumed, i.e., it is a portion that will become actually-consumed when all of the data currently in the cache has been flushed from the cache to be written on the medium, i.e, recorded upon the medium. The portion designated by "C" is that portion of the medium not yet logically or actually consumed. The location of the border between the "B" and "C" portions is the logical position of the writing head as opposed to the physical position, the physical position being located at the border between the "A" and "B" portions.

The physical and logical positions of the writing head can be determined by considering the following fixed and known factors: (a) the writable length of the medium, (b) the density at -which the data is written onto the medium, and (c) the size of any inter-record gaps. Although the units of measurement and data can vary without departing from the scope of this invention, for illustrative purposes only it will be assumed hereinafter that the units of length are inches and the units of data are bytes.

The first algorithm determines at selected times the current physical position of the writing head in relation to the front of the tape by: (a) keeping a current total of all data units (bytes) already written to the medium, (b) dividing that total by the data density (bytes/inch) , (c) adding to the result of the division the total length (in inches) of all inter-record gaps associated with the already written data. The result is the physical position, within an acceptable tolerance, of the writing instrument from the front of the medium in terms of length units (inches) . The second algorithm determines at selected times the current logical position of the writing head in relation to the front of the tape by: (a) keeping a current total of all cached write data units (bytes) , i.e., data slated to be written to the medium, (b) dividing that total by the data density (bytes/inch) , (c) adding to the result of the division the total length (in inches) of all inter-record gaps which will be associated with said cached write data, and further adding to that total the physical position of the writing head. The result is the logical position, within an acceptable tolerance, of the writing instrument from the front of the medium in terms of length units (inches) .

The length of the past-end-of-tape media determines the range of acceptable tolerances. It has been found that industry standards or such media provide more than enough excess writable media beyond the end-of-tape reflector to compensate for any differences in the-calculated from the real or actual positions. It has been found that these algorithms mathematically predict at any time both the physical and logical positions within 2% of actual for conventional tape devices.

Whenever the caching system recognizes that the currently queued I/O requests will bring the tape device's writing instrument to within a selected proximity of the end of the storage media, the caching system discontinues the queuing of I/O requests until this condition is no longer true. Preferably the selected proximity is approximately 150 feet for tapes having a total length of 2400 feet. Thus according to the method of this invention, when the logical position of the writing head is within 150 of the end of the tape, caching of I/O requests and data is stopped and the cache is flushed in its normal course. Thereafter the I/O requests are directed to the system device driver. The net effect is that for 95% of the tape the I/O operations are accelerated. But for the 5% immediately preceding the end of tape, the system functions as in the prior art shown in Fig. 1. Hence, the size of the data cache can automatically increase to expand and encompass the entire available space on the medium to which information or data can be written. For example, if the medium is empty, the cache size could expand to almost the entire capacity of the medium.

Another way of expressing the steps of the present method is as follows. The following factors are utilized in determining when to write the current cache contents to tape;

1) number of inches of tape on the reel (total inches) ;

2) number of inches of tape already consumed (already consumed) ; 3) density that data will be written at

(bytes per inch density) ;

4) inter-record-gap size (irg inches) . The apparatus calculates the number of inches of tape that would be consumed if the data in the variable length cache memory is written to the tape drive by: consumed = [data size in bytes]/[bytes per inch density] + [irg size] if [already consumed] + consumed > = [total inches] then begin flush the contents of the current cache out to tape stop data caching end [already consumed] = [already consumed] + consumed

Using this method, the size of cache is dynamically alterable and is no longer limited to the amount of data that can be written past the end-of-tape reflector. Instead, the cache is limited only by: 1) amount of data that can still be written to the tape and

2) memory caching resources

It will be understood that the present apparatus and method could be implemented in a variety of different ways without departing from the spirit and scope of the present invention. For example, the method could be implemented using a stored program computer. In this instance, a dynamic cache manager can be provided in the form of a program. Alternately, the apparatus could be implemented as a hardwired system instead of a programmed system. It could be implemented with a dedicated cache management microprocessor with the management program stored in ROM.

It will also be understood that the apparatus and method of the present invention are usable with a wide variety of different types of computer systems. For example, the dynamically adjustable cache memory of the present invention can be used in conjunction with a relatively standard memory management system. However, beyond that, the apparatus and method of the present invention could be used very advantageously in a virtual memory management system.

When the apparatus and method of the present invention are used in conjunction with a virtual memory management system, that portion of the virtual memory management syste which deals with input/output operations could communicate with a dynamic cache manager program. The cache management program could take in as inputs the length or total number of inches of tape on a reel, the density in bytes per inch at which data will be recorded as well as the inter-reσord- gap size.

The dynamic cache management program in a virtually memory system could use, to implement the variable size cache memory, one or more "pages" of memory taken off of th virtual memory system free page list. The number of memory pages taken from the free page list in this instance is dynamically controlled.

Hence, the cache memory itself is implemented as a plurality of one or more pages from the virtual memory system. As is well known, such pages could be implemente on disk drives. In such an instance, as is standard for virtual memory systems, a page of information recorded on disc drive is brought into computer memory as needed. Hence, the cache memory itself could be partly resident i main computer memory and partly resident on one or more d drives depending on the assignments made by the virtual memory management system. Thus, as a practical matter, t cache memory can be made as large as necessary within the limits of the virtual memory management system.

Figure 4 illustrates an overall block diagram of a virtual memory type system 50 which incorporates a dynamically variable cache memory in accordance with the present invention. The system 50 includes a computer 52 which is being controlled in part by a control program wh includes a virtual memory management system kernel 54. T kernel 54 receives input/output requests from application programs 56 which are being executed.

The kernel 54 communicates with a dynamic cache management program 58 which could be resident for executi purposes in the main memory of the computer 52. The dyna cache management program receives as parameters, informat pertaining to the length of the media magnetic tape, whic is available, the tape write density as well as the inter record gaps.

Utilizing the above described method steps, the dyna cache management program 58 adjusts the size of a variabl length cache memory 60. In accordance with the present method, when the amount of information currently in the cache 60, plus the amount of information previously writt to the tape drive 10 exceeds the total available space th the current contents of the cache 60 are written to the t drive 10. Where, as in the system 50 of Figure 4, the cach 60 is implemented utilizing one or more pages provided by the virtual memory management system, indicated generally a 62, then the size of the cache 60 can become arbitrarily large and far exceed the size of the main memory of the computer 52.

If the prior sum does not exceed the available space o the tape on the drive 10, then the s.ize of the cache 60 can be increased when the next file or block of information is received to be written to the tape drive. The size of the cache 60 can thus be continuously increased so long as ther is space on the physical reel of tape.

Figure 5 illustrates an alternate system 70 which embodies the present invention. In the system 70, the device driver 8a does not communicate directly with the tap unit 10a. Instead, the tape unit 10a includes a dynamic cache management program 72 in combination with a memory 74 of the fixed maximum size.

It will be understood that the dynamic cache managemen program 72 could be executed by a microprocessor contained within the system 10a. The variable cache management program 72 can be used to store information into the cache memory 74 so long as the amount of information stored therein is less than or equal to the available space on the physical tape drive using the methods described previously. When the contents of the cache memory 74 correspond to the available space on the magnetic tape on the drive 10b the contents of the memory 74 are flushed or written to the tap drive 10b. Input concerning the write density as well as the tota capacity or length of the tape available on the drive 10b can be provided via manually setable switches 76.

Yet another embodiment of the present invention is illustrated by the system 80 of Figure 6. In the system 80 a variable capacity cache memory unit 82 is provided which is coupled between the computer 52 and the tape drive 10. In the embodiment of Figure 6, the cache memory unit can implemented as a unit which is separate from, although coupled to the tape drive 10. Figure 7 illustrates an overall block diagram of the method of the present invention. After each block of information, corresponding to some number of bytes, is written into the cache memory, a calculation is made in a step 90, as previously discussed, to determine the needed extent of the write memory based on the contents of the cache memory and the number of bytes previously written t the tape. The total available extent of write memory is retrieved in a step 92.

The needed extent of write memory is compared to the total available extent of write memory in a step 94. If needed extent of write memory exceeds the total available extent of write memory, in a step 96, the contents of the cache memory are written to the physical tape. In a step 98, the cache memory is released if need be, such as if t cache memory were implemented using pages in a virtual memory system. Once the physical tape is full, then it would then be necessary to reload the tape unit with a fr reel of tape to continue writing to that particular drive. Again with reference to the step 94, if the needed extent of write memory does not exceed the available exte of write memory on the tape unit, the size of the cache memory can be expanded to receive the next vial or block information to be written to the tape drive.

Attached hereto is a computer program module named "TATHROTTLE" which carries out the calculations for dynam cache management as previously described and illustrated Figure 7.

A particular type of virtual memory system which can advantageously use the present invention is represented b the family of VAX-type virtual memory computer systems marketed by the digital equipment corporation.

From the foregoing, it will be observed that numerous variations and modifications may be effected without departing from the spirit and scope of the novel concept of the invention. It is to be understood that no limitation with respect to the specific apparatus illustrated herein i intended or should be inferred. It is, of course, intended to cover by the appended claims all such modifications as fall within the scope of the claims.

KBLF/REQUIRE 'PRETTY'>

%TITLE 'TATHROTTLE - VAX/VMS Tape Accelerator Output Throttling Mechanism

MODULE tathrottle (IDENT « 'V05-001'

)

BEGIN

!++

! FACILITY: VAX/VMS Executive, Device Drivers

!

I ABSTRACT:

This module implements the routines used control tape output caching.

SUBSTITUTE SHEET ISBTT 'Declarations'

LIBRARY 'SYS$LIBRΛRY:LIB' ; ! Systββ Syβbols

LIBRARY 'DRIVER'; I Driver linkages and psacti

LIBRARY 'LIBEXT*; I Extensions to LIB.L32^'

LIBRARY 'TADEF'; ! TADRIVER definitions

SWITCHES ADDRESSING_HODE (EXTERNAL - GENERAL, HONEXTERNAL « WORD_RELATIVE) ;

_driver_linkagβs; 1 Standard driver linkages

LINKAGE throttle btoh_linkage - JSB (REGISTER - 0, REGISTER « 2; REGISTER - 0), throttle-coβpTβte_linfcage » JS8 (REGISTER - 2, REGISTER ■ 3), throttle-consume linkage a JSB (REGISTER « 2, REGISTER • 3), throttla~eo. linkage « JSB (REGISTER » 2, REGISTER - 3), throttle~iniT_linkage « JSB (REGISTER - 2). throttle~write_linkage - JSB (REGISTER ■ 2, REGISTER « 3);

LINKAGE (REGISTER « 3, REGISTER - 5) : NOPRESERVE (0, 1), » JSB (REGISTER - 1; REGISTER - 1, REGISTER - 2) : NOPRESERVE (2 βxβSdeanonpegβd^linkagβ « JSB (REGISTER - 0) : NOPRESERVE (1, 2, 3);

FORWARD ROUTINE throttle btoh : ! Convert bytes to hundreths of inches

NOVALUE throttle_btoh_linkagβ, throttle_coaplete : 1 Process user write completion

NOVALUE throttle_coaplete_linkage , throttle_consu>e : ~ ~ ! Process physical write completion

NOVALUE throttle_con*u»β_linkagβ . throttle_βof : ^— t End of file processing

NOVALUE throttle_βof_linkege, throttle init : ^{— ""} ! Initialization routine

NOVALUE throttle_init_linkagβ, throttle_write : ~ ~ ! Write Throttling Mechanise throTtle_write linkage;

!+

! MACROS:

MACRO

_alloc_block (length, allocated_lengt , block_type, bloek_addr) _

IF (status « exe$alononρaς*d (length; allocated_length, block addr)) THEN ~

BEGIN

CHSFILL (0, .allocatβd_length, .block_addr) ; block_addr [irρ$w_size] « .allocated_length; block addr [irp$b_type] » block type;

END; ~

*, dec (val) « ~ val « .val - 1 \ , _inc (val) » val • .val + 1 *, _init_queue (head, tail) « head » head; tail s .head; \;

SUBSTITUTE HEET driver ptacts; l Standard driver PSECTS

!+

! EQUATED SYMBOLS:

!-

LITERAL k_βof_bcnt » 14, 1 Bytas par End Of rile k~fal7β ■ 0, ! Logical false k_int»rval « 65535, t Test interval k~irg ■ 60, ! Scaled interrecord gap length k~scalβ_ actor - 100, 1 Scaling factor (hundreths of i

Atrue ^" _ ; 1 Logical true

1+

1 OWN (R/W) STORAGE;

!-

OWN throttlβ_tβst_intβrval : WORD INITIAL (k_interval) ;

!+

1 BUILTIN DECLARATIONS: !-

BUILTIN

INSQUE, REMQUE;

!+

2 EXTERNAL ROUTINES:

EXTERNAL ROUTINE exe$alononpaged : ! Allocate nonpaged dynamic menor exe$ lononpaged_linkag , exe$deanonpaged : NOVALUE ! Deallocaye nonpaged dynamic βea

•xβ$dβanonpagβd_linkagβ; t+

! EXTERNAL REFERENCES:

!-

EXTERNAL bug$_inconstate, ! Inconsistent state bugcheck

•xβ$gl_abstia; ! Age of current incarnation in s

SUBSTITUTE SHEET %SBTTL 'THROTTLE BTOH - Convert Byte Count To Hundreths of Inches'

ROUTINE throttle btoh (bent, tacb ; hcnt) : NOVALUE throttlβ_btoh_linkagβ ■

++

FUNCTIONAL DESCRIPTION:

This routine is called to convert a byte count to the equivalent in hundreths based on tape density.

FORMAL PARAMETERS: bent — Byte count tacb -> Address of Tape Cache Control Block hcnt - Equivalent in hundreths

SIDE EFFECTS:

I/O request is completed.

BEGIN

MAP tacb : REF BLOCK [, BYTE}; ! Disk Cache Control Block

IF .tacb [tacb 1 density] NEQ 0 THEN hcnt - ((.bcnt*k_scalβ_factor*10 + 9)/.tacb [tacb l_dβnsity] J/10 ELSE ~ hcnt ■ 0;

END; ! End of GLOBAL ROUTINE THROTTLE B

ISBTTL 'THROTTLE_COHPLETE - User Output Completion Processing'

GLOBAL XOUTINE throttlβ_complβ β (tacb, irp) : NOVALUE throttle_complete_linkage

FUNCTIONAL DESCRIPTION:

This routine is called for the purpose of accounting for the amount of tape that the user process has written.

FORM PARAMETERS : tacb -> Address of Tape Cache Control Block irp -> Addreaa of I/O Request Packet

IPL ■ IPL$_SYNCH COMPLETION STATUS:

SS$_NORMAL Normal successful completion SIDE EFFECTS:

I/O request is completed

BEGIN MAP tacb : REF BLOCK [, BYTE], J Disk Cache Control Block irp : REF BLOCK [, BYTE); ! I/O Request Packet

LOCAL hcnt;

!+

! Tine stamp the first completion followiung an initialisation..

!-

IF .tacb [tacb_l_timβ_stamp] EQL 0 THEN tacb I taeb_l_timβ_stam ] ■ .βxβ$gl_abstia; throttle btoh (.irp [ irp$l_ben ] , .tacb; hcnt); hcnt _ .hcnt + k_irg; ~ tacb ( acb_l_taρe^"_complβtβd} m .tacb I tacb_l_tapβ_coκplβte J + .hcnt;

END; ~ 7 End~of GLOBAL ROUTINE THROTTLE COMP

1/01525

22

%SBTTt 'THROTTLE_CONSUME - Account For Tape Consumed'

GLOBAL ROUTINE throttle consume (tacb, irp) : NOVALUE throttle_consu»e_linkage ■

++

FUNCTIONAL DESCRIPTION:

This routine is called for the purpose of accounting for tape actually writ

ΓORMAL PARAMETERS: tacb -> Address of Tape Cache Control Block irp -> Address of I/O Request Packet

IPL « IPL$ SYNCH

BEGIN

MAP tacb : REF BLOCK [, BYTE], ! Disk Cache Control Block irp : REF BLOCK [, BYTE]; 1 I/O Request Packet

LOCAL hcnt; throttle btoh (.irp [irp$l_bcnt] , .tacb; hcnt); hcnt -_* .hcnt + k_irg; ^"~ tacb (tacb_l_tape_consumβd] _ .tacb [tacb l_tapβ_consumβ J + .hcnt; tacb [tacb-1~tape-remaining) »= .tacb [tacb X tapβ_rβmaining] - .hcnt;

END; ~ ~ 7 End of GLOBAL ROUTINE THROTTLE CO

%SBTTL »THROTTLE_EOF - End Of File Throttling Mechanism'

GLOBAL ROUTINE throttlβ_βof (tacb, irp) : NOVALUE throttlβ_eof_linkagβ ■

++

FUNCTIONAL DESCRIPTION:

This routine is called for the purpose of throttling end of files,

FORMAL PARAMETERS: tacb -> Address of Tape Cache Control Block irp -> Address of I/O Request Packet

IPL - IPL$ SYNCH

BEGIN

MAP tacb : REF BLOCX (, BYTE], ! Disk Cache Control Block irp : REF BLOCK [ , BYTE] ; 1 I/O Request Packet

LOCAL hcnt; throttle btoh (k_βof_bcnt, .tacb; hcnt); hcnt » .hcnt + k_Trg;~^" tacb [tacb_l_tapβ_remaining] »- .tacb [tacb_l_tapβ_rβmaining] - .hcnt; tacb [tacb~l~tape~consu»ed] β .tacb [tacb_T_tapβ_c^"onsumβd] + .hcnt; tacb [tacb~l~tape~completed] _* .tacb [tacb T tape^" completed] + .hcnt;

END; ~ 7 End~of GLOBAL ROUTINE THROTTLE

SUBSTITUTE SHEET %SBTTL 'THROTTLE_INIT - Throttle Facility Initialization Routine'

GLOBAL ROUTINE throttlβ_init (tacb) : NOVALUE throttlβ_init_linkage •

++

FUNCTIONAL DESCRIPTION:

This routine is called for the purpose of (re)initializing the output throttling facility.

FORMAL PARAMETERS:

IPL _* IPL$_SYNCH

I—

BEGIN

MAP tacb : REF BLOCK [, BYTE];

LOCAL code array : REF VECTOR [0, LONG], statu? : BLOCK [1, LONG] INITIAL (ss$_nor»al) ; tacb [tacb_l_tapβ_rβmaining] _* .tacb [tacb_l_initial length]; tacb [tacb~l~tapβ~consumed] « 0; ~ tacb [tacb_l_tape~cojβplβted] ■ 0; tacb [tacb~l time stamp] » 0; tacb [tacb_l~abcnitl ■ 0; tacb [tacb_l_log_abcnt] « 0; tacb [tacb_l~«riύ_eount] « 0; tacb [tacb_l_requeϊt_count] * 0; tacb [tacb~w_βr cntJ-■ 0; fcode_arra ^" -» tacb [tacb l_fcode array];

INCR i FROH 0 TO 63 DO fcode array [.i] -> 0;

RETURN .status; END; End of GLOBAL ROUTINE THROTTLE I

tJSTTL "THHOTTLE_W«ITE - Write Throttline Mechanise*.

GLOBAL ROUTINE threttle_wri e (tacb, irp) : threttle_write_linkefe ■

rUNCTtONλL DESCRIPTION:

This routine is called loc the purpose of throttling writes. FORMAL PARAMETERS: tacb -> Address ef Tape Cache Control Sleek irp -> Address ef I/O Request Packet

IPL - Ires SYNCH COMPLETION STATUS: k true »> If a status of true is returned, the caller should cache the request. k~ta.se -> If a status ef false is returned, the caller should not cache the request.

■CCIN HAP tacb : _tr BLOCK (, BYTE), I Disk cache Control Block irp : xεr BLOCK I. BYTE); I I/O Request Packet

LOCAL hcnt_cache, hcnt~irp, status : BLOCK (1, LONG) INITIAL (k_true); throttle btoh (.irp {irp.l_bcn ] , .tacb; hcnt_irp); ir {.tacb Itacb_l_tape_co^»apleted) ♦ ,hcnt_irp) OEQU .tacb Itacb_l_step_cachele 1 THEN status «

.status

END; I End of CLOBAL ROUTINE THROTTLE WRITE

END

ELUDOM

I End of -odule TATHROTTLE

SUBSTITUTE SHEET

Claims

I claim:

1. An apparatus for dynamically changing the storage capacity of a cache memory used in the transfer of data to an output device having a predetermined maximum storage capacity, the apparatus comprising: means for determining an available storage capacity parameter of the output device; means for comparing the current capacity of the cache memory to said parameter; and means for transferring the contents of the cache memor to the output device if the current size-thereof exceeds said parameter.

2. An apparatus as in claim 1 including means for increasing said cache memory storage capacity if said parameter exceeds said current storage capacity.

3. An apparatus as in claim 2 including virtual memo storage means for increasing said storage capacity.

4. An apparatus as in claim 2 including capacity increasing storage means linkable to said cache memory. 5. An apparatus as in claim 1 including a cache memo formed of a plurality of sequentially addressable memory locations.

6. A method of dynamically changing the storage capacity of a cache memory used in the transfer of data to an output device having a predetermined maximum storage capacity comprising: determining an available storage capacity parameter o the output device; comparing the current capacity of the cache memory to the parameter; and transferring the contents of the cache memory to the output device if the current size thereof exceeds the parameter.

7. A method as in claim 6 including: increasing the storage capacity of the cache memory i the parameter exceeds the current capacity. 8. A method as in claim 7 including: providing virtual memory storage for increasing the cache memory storage capacity. 9. A method as in claim 6 including determining the quantity of information that had previously been transfer to the output device; and adding the current capacity of the cache memory to t determined quantity. 10. A method as in claim 9 including increasing the storage capacity of the cache memory if the sum of the current capacity plus the determined quantity is a predetermined amount less than the maximum storage capaci

11. A cache memory system with a dynamically adjustable capacity and usable to transfer information to output device having an effective predetermined maximum storage capacity, the cache memory system comprising: means for determining the amount of information previously transferred to the output device; means for forming a sum of said previous amount and said present cache capacity; means for comparing said maximum storage capacity to said sum and for generating an indicium in response to sa maximum storage capacity exceeding said sum; and means responsive to said indicium, for increasing sa capacity.

12. A cache memory system as in claim 11 with said increasing means including: means for providing one or more groups of storage locations linkable to said cache memory for increasing th storage capacity thereof.

13. A cache memory system as in claim 12 with said providing means including at least a portion of a virtual memory management system. 14. A cache memory system as in claim 12 with said providing means including additional storage capacity linkable to said cache memory.

15. A cache memory as in claim 12 with said providing means including solid state memory means contiguously addressable with said cache memory.

16. A cache memory system as in claim 11 with said output device including a linearly moveable storage medium.

17. A cache memory as in claim 16 with said storage medium including a magnetic tape. 18. A cache memory system as in claim 11 including virtual memory storage means with said increasing means including means for allocating additional storage units fro said virtual storage means to increase said cache memory storage capacity. 19. A cache memory system as in claim 11 including means, responsive to an absence of said indicium for transferring contents of said cache memory system to the output device.

20. A cache memory system as in claim 19 including means, responsive to said transferring means, for reducing said capacity of said cache memory.

21. A method of operating a cache memory system with dynamically adjustable capacity and usable to transfer information to an output device having a predetermined nominal maximum storage capacity, which is a selected amou less than the actual maximum storage capacity, the method comprising: providing a cache memory; determining the amount of information previously transferred to the output device; forming a sum of the previous amount of information a the present cache capacity; comparing the nominal maximum storage capacity to the sum and generating an indicium in response to the nominal maximum storage capacity exceeding said sum; and increasing said capacity in response to the presence the indicium.

22. A method as in claim 21 including: providing one or more groups of storage locations linkable to the cache memory for increasing the storage capacity thereof.

23. A method as in claim 22 including providing a virtual memory management system. 24. A method as in claim 22 including linking additional storage capacity to the cache-memory.

25. A method as in claim 21 including providing virtual memory storage means and allocating additional storage units from the virtual storage means to increase cache memory storage capacity.

26. A method as in claim 21 including transferring contents of the cache memory system to the output device response to an absence of said indicium.

27. A method as in claim 26 including reducing the capacity of the cache memory in response to the transferr step.

28. A method as in claim 27 including reducing the current capacity of the cache memory to zero after the transferring step. 29. In a computer system having a computer writing data through a data cache to a data storage medium of fin size, the data cache comprising: means for determining a remaining, unwritten, capaci of a finite size data storage medium; and means for limiting the capacity of the data cache to amount that is capable of being stored in the determined remaining capacity of the finite size data storage medium

30. The computer system according to claim 29 where the data storage medium is written serially from a beginn to an end, and wherein the means for determining comprise means for calculating the capacity of the serially written data storage medium that remains before its end.

31. The computer system according to claim 30 wherein the serially written data storage medium is magnetic tape having an end-of-tape marker before its end, and wherein th means for calculating comprises: means for computing the cumulative number of inches of magnetic tape required to write all data both previously written from the cache, and currently held within the cache to the magnetic tape; and wherein the means for limiting comprises: means for comparing the computed number of inches of magnetic tape to the total number of inches of magnetic tap remaining before the end-of-tape marker. 32. An improved method of writing data through a data cache to a data storage medium of finite size, the improved data caching method comprising: determining a remaining capacity of a finite size data storage medium; and limiting the quantity of data accepted from the computer into the cache to that quantity capable of being stored in the determined remaining capacity of the finite size data storage medium.

33. The data caching method according to claim 32 wherein the writing is serial from a beginning to an end of a data storage medium, and wherein the determining comprises: calculating the capacity of the serially written data storage medium that remains before it send. 34. The data caching method according to claim 33 wherein the serial writing of a data storage medium is of magnetic tape having an end-of-tape marker before its end, and wherein the calculating comprises: computing the cumulative number of inches of magnetic tape required to write all data previously and currently within the cache to the magnetic tape; and wherein the limiting step includes: comparing the computed number of inches of magnetic tape to the total number of inches of magnetic tape remaining before the end-of-tape marker.

35. A data cache for buffering data between a data origination means for originating data and a data storage means, having a finite data storage capacity, for storing the data, the data cache comprising: an addressable memory means for holding the data to stored intermediary in its transfer fro -the origination means to the data storage means; means for varying the addressable space of the addressable memory means in accordance with the data stor capacity of the data storage means.

36. The data cache according to claim 35 wherein th addressable memory means comprises: a buffer memory having a physically addressable spac that is less than the data storage capacity of the data storage means; means for determining a remaining data storage capac of the data storage means; and means for controlling the addressable space of the buffer memory to be the greater of its physically addressable space and the determined remaining data stora capacity.

37. A method of buffering data from a data originat means for originating data and a data storage means, havi a finite data storage capacity, for storing the data, the data buffering method comprising: holding the data to be stored in an addressable memo intermediary in its transfer from the origination means t the data storage means; and varying the addressable space of the addressable mem in accordance with the data storage capacity of the data storage means.

38. The data caching method according to claim 37 wherein the holding is in a buffer memory having a physically addressable space that is less than the data storage capacity of the data storage means; and wherein the varying comprises: determining a remaining data storage capacity of the data storage means; and controlling the addressable space of the buffer memory to be the greater of its physically addressable space and the predetermined remaining data storage-capacity.

39. A computer that caches its data output to a data storage device that in turn writes the output data on a dat storage medium of finite capacity, the data-caching compute comprising: an addressable cache memory; and means for dynamically varying the addressable space of the addressable cache memory to be less than or equal to a remaining, unstored, data capacity of the.data storage medium.

40. A data storage device that caches a data output that it receives from a computer prior to writing the computer output data on a data storage medium of finite capacity, the data-caching data storage device comprising: an addressable cache memory; and means for dynamically varying the addressable space of the addressable cache memory to be less than or equal to a remaining, unstored, data capacity of the data storage medium.