Amazon Technologies, Inc.

Mark Robert Sieklucki

James Christopher Sorenson

Paul David Franklin

Rajesh Shanker Patel

Dave Rahmany

A magnetic tape drive includes a head assembly with a write head and a read head. The read head is positioned ahead of the write head, so that as a magnetic tape is moved across the head, the read head is able to read data that is about to be overwritten by the write head. When a client computer system commands the tape drive to write data to the magnetic tape, the tape drive writes the data to the tape but preserves any data that is overwritten by reading the data before it is overwritten. The preserved data is returned to the client computer system. In various examples, the returned data can be written back to the tape, discarded, or stored elsewhere under the control of the client computer system.

Concurrent read/write apparatus

A tape drive allows for an increased track density by partially overlapping adjacent tracks as each successive track is written. Each successive track overwrites a portion of the width of the previous track, thereby reducing the width of the previous track. This process can be applied repeatedly thereby producing an arrangement of shingled data tracks across the width of the tape. In an embodiment, single-pass verification of data written in this manner is accomplished using a head assembly with two read heads per track. A first read head positioned behind the write head verifies that the data written is correct, and a second read head positioned over the previous track verifies that data on the previous track remains intact after being partially overwritten.

Data recording method and apparatus

A storage manager may be used to perform a refinement operation on a coding matrix as part of a coding operation (e.g., an encoding operation or a decoding operation) in a storage system, such as an object-redundant storage system. The refinement operation may include identifying a plurality of computations to be performed as part of the coding operation, where the plurality of computations have common terms and operations. The refinement operation may include refining a coding matrix associated with the coding operation based on the identified computations and precomputing the identified computations. The precomputed computations may be used instead of performing the computations when performing the coding operation.

Computation refinement in a data storage system

A storage manager may be used to perform a coding operation (e.g., an encoding operation or a decoding operation) on a data object using a refined version of a coding matrix in a storage system, such as an object-redundant storage system. The coding operation may include identifying a coding matrix for the data object and retrieving a refined version of the identified coding matrix from a data store of pre-refined coding matrices. The refined version of the coding matrix may identify two or more computations to be performed as part of the coding operation, where the two or more computations have common terms and operations. The coding operation may include determining one or more precomputed results of the two or more identified computations. The precomputed results may be used instead of performing the computations when performing the coding operation.

Growing community of inventors

Seattle, WA, United States of America

Mark Robert Sieklucki