Browse Prior Art Database

Staged Read Error Recovery for a Magnetic Tape Subsystem

IP.com Disclosure Number: IPCOM000109399D
Original Publication Date: 1992-Aug-01
Included in the Prior Art Database: 2005-Mar-24
Document File: 3 page(s) / 155K

Publishing Venue

IBM

Related People

Bentley, SR: AUTHOR [+4]

Abstract

This article focuses on a read error recovery strategy which combines a parameterized hardware dataflow and microcode controls based on a staged matrix. It is implemented on the 3490E magnetic tape subsystem which is capable of interpreting more than one recorded format on tape (18-track format AXP ECC encoding vs. 36-track format Reed Solomon ECC encoding). Read Error Recovery configurations are defined by a staged matrix designed to address the standard problems of tape defects, debris, and tension as well as the special problems of read interchange and track misregistration. The key advantages are: 1. A generalized, parameter matrix is used to define staged recovery. Recovery matrix parameter options are implemented as data areas within microcode, so changes are confined to data areas, not lines of code. 2.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 49% of the total text.

Staged Read Error Recovery for a Magnetic Tape Subsystem

       This article focuses on a read error recovery strategy
which combines a parameterized hardware dataflow and microcode
controls based on a staged matrix.  It is implemented on the 3490E
magnetic tape subsystem which is capable of interpreting more than
one recorded format on tape (18-track format AXP ECC encoding vs.
36-track format Reed Solomon ECC encoding).  Read Error Recovery
configurations are defined by a staged matrix designed to address the
standard problems of tape defects, debris, and tension as well as the
special problems of read interchange and track misregistration.  The
key advantages are:
1.   A generalized, parameter matrix is used to define staged
recovery.  Recovery matrix parameter options are implemented as data
areas within microcode, so changes are confined to data areas, not
lines of code.
2.   A 36-track format strategy capitalizes on a parameterized Reed/
Solomon ECC dataflow to vary the base operating point.
3.   An 18-track format strategy is incorporated to improve the
probability of error recovery in cases of track misregistration.
Electronic switching of the dual head elements is used to attempt
readback verification from either set of recording head elements.
4.   Read recovery summary data is condensed from the final state of
the staged Read Recovery Matrix for the purpose of algorithm
optimization and volume analysis. This data is stored internally and
is also sent to the host as a part of channel sense data.  The data
area implementation makes for flexible feedback of optimization
information into the design in a straightforward manner.

      When reading 18-track format tapes with a 36-track head, the
read recovery matrix directs unique interchange recovery.  This
scheme utilizes electronic switching to attempt readback verification
from both sets (banks) of recording head elements.  When the read
track width is reduced by a factor of 2, it increases the sensitivity
to defects and also reduces the available SNR margin which translates
it to overall lower reliability.  With only half the track width,
track misregistration problems can be introduced on tape interchange.
The smaller track width also has a higher exposure to read errors due
to debris and tension errors.

      When reading 36-track format tapes with Reed/Solomon ECC, the
Read Recovery Matrix takes advantage of a parameterized hardware
dataflow that allows a wide variation in its base operating point.
In the latter stages of read error recovery, microcode makes
extensive use of this hardware flexibility in its attempt to read
problem blocks on tape.

      In general, the Read Error Recovery Matrix directs a series of
read and repositioning sequences around the block in error.  Attempts
to read the block are tried in both directions, under both nominal
and parameterized dataflow configurations.  There are a maximum of 7
stages in the matrix, with...