Electrical computers and digital processing systems: memory – Storage accessing and control – Specific memory composition
Reexamination Certificate
2000-08-04
2002-10-22
Kim, Matthew (Department: 2186)
Electrical computers and digital processing systems: memory
Storage accessing and control
Specific memory composition
C711S113000, C360S048000
Reexamination Certificate
active
06470421
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to hierarchical, demand/response, disk storage subsystems, and more particularly to a method and means for reducing contention among one or more direct access storage devices (DASDs) in the presence of concurrent accessing of data formatted according to one addressing convention, but formatted and stored across one or more DASDs according to a second convention.
DESCRIPTION OF RELATED ART
In this specification, the acronym DASD signifies a cyclic, multitrack, direct access storage device of large disk diameter and low recording density along any track. Also, HDD is the acronym for high-density disk drives having a relatively small disk diameter with high recording density along any track and a high radial number of tracks. Lastly, the terms “subsystem”, “storage control unit”, and the IBM 3990 SCU are used interchangeably.
Data Storage Models and Format Conversion at DASD Level
One early storage model of data was denominated CKD. CKD is an acronym for count, key, and data. This is a variable-length record formatting convention used by IBM for DASDs. This convention required a count field defininig the length in bytes of the data recorded in a variable-length data field and a key field avaible for use as a record identifier. In practice, the count field is frequently also used to provide record identification. Each of the fields as recorded was spaced apart by a gap along the DASD track. The gap was designed as a pause interval on the continuously rotating DASD, permitting the system to adjust itself to process the next field. The gaps were occasionally dissimilar in length and also served as a place for inserting metadata That is, the gap between the C and K fields differed from the gap between the K and D fields.
Each CKD-formatted record consisted of at least the fixed-length count field and a variable-length data field. The use of the key field was optional and relegated primarily to sort intensive applications. The records were stored or mapped onto a cylinder (track), head (disk), (sector) addressable group of synchronous and constant speed rotating magnetic disks.
Major operating systems such as the IBM MVS, access methods such as VSAM, and significant amounts of application programming became heavily invested with the CKD data model and the simple cylindrical, physical storage addressing of large diameter disk drives. While some records would be less than a track extent, theoretically other CKD records could span several tracks. However, the advent of virtual memory, demand paging, and page replacement operations between mainframe CPUs, such as the IBM S/370 with MVS OS, and large disk-based storage subsystems, such as the IBM 3390, tended to conform CKD records to approximate a 4-kilobyte page. Relatedly, the typical 3390 recording track could accommodate up to twelve pages or 48 Kbytes+5 Kbytes worth of gaps between the fields within a record and between records.
With the passage of time, the recording densities of disk drives substantially improved and it was economically desirable to map data recorded in one format (CKD) onto a disk drive programmed to record data in another format (fixed-block architecture or FBA). Relatedly, FBA is an acronym for fixed-block architecture. That is, a string of extrinsically formatted information is blocked into a succession of equal-length blocks. One way of ensuring recording synchronism between the formats is to have the initial count field of each new CKD record start on an FBA block boundary. In such a scheme, the last FBA block should be padded out to its block boundary.
Reference should be made to Menon, U.S. Pat. No. 5,301,304, “Emulating Records in One Record Format in Another Record Format”, issued Apr. 5, 1994. Menon exemplifies the state of the art in format conversion disclosing an emulation method for rapidly accessing CKD records in which the CKD records are stored on a disk drive in FBA format.
Menon maps CKD to FBA blocks by embedding one or two indicators in the mapped information. The term “mapped information” is consonant with the FBA image of the CKD track. In this regard, an “indicator” is coded information of location displacement or a data attribute with respect to a CKD record being accessed on an FBA-fonnatted device. The indicators permit a general orientation and then a precise location of the head with reference to a record of interest on a given FBA DASD track measured from the index or other benchmark. Thus, when CKD records were written out to the FBA-formatted device, the indicators were placed in the stream. Consequently, when the records had to be accessed and staged for both reading and write updating, the access time or disk latency is perceptibly shortened using the indicators.
Overview of Hierarchical Demand/Response DASD Storage Subsystems
In the period spanning 1970 through 1985, IBM developed large-scale multiprogramming, multitasking computers, S/360 and S/370 running under the MVS operating system. A description of the architecture and that of the attached storage subsystem may be found in Luiz et al., U.S. Pat. No. 4,207,609, “Method and Means for Path Independent Device Reservation and Reconnection in a Multi-CPU and Shared Device Access System”, issued Jun. 10, 1980. Such systems were of the hierarchical and demand/responsive type. That is, an application running on the CPU would initiate read and write calls to the operating system. These calls, in turn, were passed to an input/output processor or its virtual equivalent (called a channel) within the CPU. The read or write requests and related accessing information would be passed to an external storage subsystem. The subsystem would responsively give only status (availability, completion, and fault) and pass the requested data to or from the CPU.
The architecture of hierarchical demand/response storage subsystems, such as the IBM 3990/3390 Model 6 and the EMC Symmetrix 5500, is organized around a large cache with a DASD-based backing store. This means that read requests are satisfied from the cache. Where the data or records are not in the subsystem cache, the data satisfing those requests are staged up from the DASDs to the subsystem cache. Write updates result in data being sent from the CPU to the cache or to a separate nonvolatile store (NVS), or both. This is the case with the IBM 3990 Model 6. The cache-stored data is then destaged or written out to the DASDs on a batched basis asynchronous to processing the write requests. Records stored in NVS are destaged only if the modified tracks are not available in cache. In these subsystems, 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.
In order to minimize reprogramming costs, applications executing on a CPU (S/390) and the attendant operating system (MVS) would communicate with invariant external storage architecture even though some components may change. Relatedly, the invariant view of storage associated with an MVS operating system required that data be variable-length formatted (CKD) and stored in that CKD format on an external subsystem of attached disk drives (IBM 3390) at addresses identified by their disk drive cylinder, head, and sector location (CCHHSS). Significantly, requested CKD-formatted data is staged and destaged between the CPU and the storage subsystem as so many IBM 3390 disk drive tracks worth of information. One address modification is to use CCHHR, where R is the record number with CC and HH refers to the cylinder and head numbers, respectively.
It is well appreciated that an improved disk storage facility can be attached to a subsystem if the new facility is emulation compatible with the unit it has replaced. Thus, a RAID 5 storage array of small disk drives can be substituted for a large disk drive provided there is electrical and logical interface compatibility. Illustratively, the IBM 3990 Model 6 storage control unit can attach an
Bui Thao B.
Chen James C.
Ng Chan Y.
Bataille Pierre-Michel
International Business Machines - Corporation
Kim Matthew
Klein Esther E.
Millett Douglas R.
LandOfFree
Method and means for reducing device contention by random... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and means for reducing device contention by random..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and means for reducing device contention by random... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2998977