Browse Prior Art Database

System and Method for Efficient Redundancy on a Single DASD

IP.com Disclosure Number: IPCOM000013705D
Original Publication Date: 2000-Nov-01
Included in the Prior Art Database: 2003-Jun-18
Document File: 3 page(s) / 57K

Publishing Venue

IBM

Abstract

Disclosed is a system and method for efficient redundancy on a single direct access storage device (DASD). The invention uses multiple copies of data to: reduce access time reduce error recovery time reduce the number of required error correcting code (ECC) bytes increase reliability increase sustained data rate Throughout the mid-80s and 90s, data capacity on DASDs has grown at a much faster rate than the rate at which data can be retrieved. To combat this, many system designers implemented redundant array of independent disks (RAID) solutions to stripe data across multiple drives, thereby speeding the access to the data by allowing multiple drives to retrieve a portion of the information. The use of RAID solutions has worked well in the high-end server market but is not a reasonable approach on lower cost workstation and desktop systems. In fact, it appears that hard drives will soon outstrip the ability of RAID systems to deal with the drive's dramatically increased storage capacities. The next logical step seems to be in the direction of placing RAID-type functionality in the drives themselves. This invention is a step in that direction.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 49% of the total text.

Page 1 of 3

System and Method for Efficient Redundancy on a Single DASD

    Disclosed is a system and method for efficient redundancy on a single direct access storage device (DASD).

The invention uses multiple copies of data to: reduce access time reduce error recovery time reduce the number of required error correcting code (ECC) bytes increase reliability
increase sustained data rate

Throughout the mid-80s and 90s, data capacity on DASDs has grown at a much faster rate than the rate at which data can be retrieved. To combat this, many system designers implemented redundant array of independent disks (RAID) solutions to stripe data across multiple drives, thereby speeding the access to the data by allowing multiple drives to retrieve a portion of the information. The use of RAID solutions has worked well in the high-end server market but is not a reasonable approach on lower cost workstation and desktop systems. In fact, it appears that hard drives will soon outstrip the ability of RAID systems to deal with the drive's dramatically increased storage capacities. The next logical step seems to be in the direction of placing RAID-type functionality in the drives themselves. This invention is a step in that direction.

Given one or more rigid disks, each of which has two writeable surfaces, a system is presented where each track of data is duplicated on the accompanying surface of the same disk in a fashion whereby each block of data is offset 180 degrees from its duplicate. A system similar to this was proposed in 1973, where n copies of the data were present on the drive, though not in the configuration mentioned above. The reason that this system did not become popular at the time has three sources; capacity was still king, lack of adequately sized cache memories, and the ability to increase the drive's RPM. The invention at the time required that the second copy of the data be written immediately after the first; resulting in a substantial access penalty. The present invention overcomes this by using a portion of the cache to store the write information until it can be written at a more advantageous time.

Preferred Embodiment:

Format - As mentioned earlier, each disk surface is duplicated on its opposing surface. The figure shows a track of data and its duplicate, where the start of each track is indicated by sid 0.

1

Page 2 of 3

Queueing - Each command in the queue will be associated with two physical locations on the disk. When access times are calculated, both locations will be used in scheduling. The only exception is when only one copy of data is valid on the disk, in the case when only one copy of the data has been rewritten. At that time the other write is still pending in the DupWrite cache, which will be explained later.

Read Accesses - Read commands are chosen from the queue by finding the command with the shortest access time.

Write Accesses - Write commands are chosen exactly like reads when considering the writing of the first copy....