1
DISK DRIVE INITIATED DATA TRANSFERS
RESPONSIVE TO SEQUENTIAL OR NEAR
SEQUENTIAL READ OR WRITE REQUESTS
FIELD OF THE INVENTION 5
This invention relates to direct access storage devices (DASDs) having either cyclic, multitracked, or cyclic, spiral-tracked storage media utilizing a discrete, double, or circular-buffered data transfer path, and more particularly to methods and means for effectuating an increased data 10 throughput in said transfer path.
DESCRIPTION OF RELATED ART
It is well appreciated in the prior art that physically a DASD comprises a spindled motor; at least one cyclic, multitracked disk coated with remanent magnetic material affixed to the spindle; a plurality of transducers for reading or recording data on selected ones of the tracks; a servo mechanism for radially positioning a honeycomb of arms 2Q tipped with counterpart transducers positionable over a selected one or more tracks; a buffered data path for transferring data between the transducers and a device interface; and several sequencers and/or microprocessor electrically coupling the interface, buffer, servo, and other means such ^ as ECC. At least one disk surface has positioning signals recorded thereon for use by the servo as a location reference for positioning transducers over selected ones of the tracks.
Logically, a DASD comprises either a cyclic, multitracked, or spiral storage medium for data and means 30 responsive to external read and write requests for establishing a buffered path to and copying data from at least one logical block address (LBA) location on the tracked medium identified by each read request and for establishing a buffered path to and for recording data located on at least one 35 LBAon the tracked medium identified by each write request. The path is directionally responsive to the counterpart read or write request and includes a circular buffer.
Architecturally, DASDs operate either as a demand/ responsive or a queued access storage subsystem externally 40 attached to one or more sources of read and write requests. Parenthetically, the term "demand/response" connotes that a new request will not be accepted from a higher echelon until the last request is satisfied by a lower echelon and a positive indication is made by the lower to the higher echelon. The 45 term "queued access" connotes a system interface such as the ANSI standard small computer system interface (SCSI) where several commands may be batched together and transferred to a subsystem or device at one time (or as one message), and the subsystem or device executes them asyn- 50 chronously. This does avoid the overhead of a tag and response for each and every command or request. Also, DASDs can operate individually, clustered, or orchestrated as in various forms of RAID arrays. Where they are clustered as part of a larger subsystem, then storage management 55 control and functionality can be enriched. This usually requires a cluster or array control unit or the like. Such higher levels of storage subsystem integration are exemplified by the IBM 3990/3390 DASD and are described in Luiz et al, U.S. Pat. No. 4,207,609, "Method and Means for Path 60 Independent Device Reservation and Reconnection in a Multi-CPU and Shared Device Access System", issued Jun. 10,1980; and more recently in Beardsley et al., U.S. Pat. No. 4,916,605, "Fast Write Operations", issued Apr. 10, 1990.
Over the years, storage functions have migrated to/from 65 CPU operating systems to attached subsystems or devices. This process has been accelerated because of availability of
2
inexpensive microprocessors and their simplified cousins called "sequencers". Thus, contemporary disk drives will include an interrupt-driven, task-switched microprocessor and satellite processors or sequencers. The microprocessor conducts overall device management including error detection and data recovery, while the sequencers handle specialized functions such as device interface attachment, disk reading and writing, and buffer management.
Historically, buffers were expensive and were sparingly used at any level in a storage subsystem. However, the high availability of inexpensive random access memory (RAM) permits use of double or circular buffering in the movement or staging of data between the device interface and the disk storage medium. Additionally, such RAM can also be implemented as a writable control store (WCS) for sequencers and as a local memory for the microprocessor.
The availability of processing and local memory within the disk storage device permits the device to work smarter and harder. Thus, a control program executing on a processor or sequencer within the DASD can implement operating system features. These features include monitor and interrupt-driven, task-switching and managing subordinate asynchronous functions as resource managers or objects. Typically, a DASD will include an interface manager having code resident in a WCS local to a dedicated sequencer and the same for a disk manager. The interface manager will interpret external read or write requests and invoke the control processor via an interrupt. The latter will perform address translation from the logical address specified in the request and cause the servo to position the DASD transducer over the translated track address. Each read request will cause the disk manager to stage the data on a track to the buffer, and the interface manager will move data from the buffer to the device interface. Likewise, each write request will cause the interface manager to move externallysupplied data into the buffer, while the disk manager will cause data from the buffer to be written on the referenced track.
The elapsed time between presentation of a read or write request to a storage device and the responsive staging up or destaging of data is a significant performance measure. It is evident that such time will vary due to the latency, patterns of referencing (read or write and random or sequential), locality of the data, track switching, error susceptibility and data recovery procedures, and path enablement and disablement. Significant attention has been paid to improving DASD performance. This includes enhancing the ability to detect and correct errors on the fly as to where data is staged responsive to read requests, and byte and/or burst correction can be made to digital data whether caused by noise, thermal asperity, track misregistration, or the like. However, one aspect needing improvement is that of data rate under various patterns of referencing.
In this specification, the storage model, as viewed from an application executing on a host CPU, comprises at least one disk storage device having a bounded number of consecutively numbered logical block addresses (LBAs) for storing data. Any mapping or address translation between an LBA and the real DASD location (cylinder, track, and head) is performed within the device.
SUMMARY OF THE INVENTION
It is accordingly an object of this invention to devise a method and means for increasing the data throughput in the buffered transfer path of a DASD under read or write referencing.
