Hitachi Global Storage Technologies Netherlands B.v.

International Business Machines Corporation

James R Shipman

David Robison Hall

Adam Michael Espeseth

George Anthony Dunn

Richard New

Richard Rolls

Embodiments of the present invention provide a data storage apparatus with new features to more easily enable atomic transactions. Rather than having the host system issue the multiple logging commands to the data storage apparatus, the data storage apparatus can be modified so that it can perform the logging function itself. In one embodiment, a data storage controller of a data storage apparatus for implementing an atomic transaction comprises a receiving module configured to receive from a host one or more commands to be executed as an atomic transaction; a log recording module, configured to record in a nonvolatile storage a log containing the one or more commands of the atomic transaction, the log to be administered by the data storage controller and not by the host; and an execution module configured to perform the one or more commands of the atomic transaction.

Hard disk drive with support for atomic transactions

Expected access times (EATs) of write request to a disk drive are essentially a measure of the predicted service time for the write request. Write Requests generated by a caching storage controller to a disk drive are essentially maintenance functions used to clear the cache. The disk drive modifies the EATs of write commands with a penalty such that read requests requiring disk access are preferentially satisfied. The penalty may be constant or may be established based on one or more factors, and may even be negative if necessary to clear a cache full of writes requiring destaging to disk.

Read priority caching system and method

A command queue is maintained in a disk drive. The command queue includes a plurality of access commands that are awaiting execution. The command queue is sorted to provisionally select one of the plurality of access commands as the next command to be executed. A seek start deadline is calculated for the provisionally selected access command based on an estimated extra latency of the provisionally selected access command. A seek operation with respect to the provisionally selected access command is deferred based on the calculated seek start deadline, to await arrival at the disk drive of a new access command. If the new access command arrives at the disk drive prior to the seek start deadline, the respective estimated access times for the provisionally selected access command and the newly arrived access command are compared to select one of the two access commands for execution.

Method and apparatus for improving efficiency of operation of a hard disk drive by deferring command execution

A method and means for performing logical-combining functions such as XOR operations in a RAID 3 or 5 array or a combinational configuration of disk drives. The logical combining occurs below the device attachment interface and is used in generating new parity for write update operations and data rebuilding. Each disk drive is modified to use a segmented and address synchronized buffer in the write and read paths, and a logical circuit that combines the data in the write path before buffering with the data in the read buffer as it is read out from the buffer. The logically combined result is presented at the device read path interface. The combined data from a first disk drive interface is introduced into the write path of a second path through the second device interface and the logical combining repeated ad seriatim, the device interfaces being connectable in a daisy-chain manner and configurable by array controller commands.

Method and means for managing disk drive level logic and buffer modified

A method and apparatus for performing preemptable calibration and housekeeping functions during idle time between commands thereby maximizing data throughput. The idle time activities are scheduled on an infrequent time scale. A single seek is used to get to a measurement location. Once the read/write head is positioned at the measurement location, it stays there for a fairly long time in order to make many measurements. If a new host command arrives, the test is aborted and the host command is serviced. The new command is begun without performing any clean-up activity and may or may not record the point at which the idle time activity was abandoned and idle time activity is reinitiated after the new command is completed. Results from the idle time activity is committed to the disk or RAM only after completion of an idle time activity, and thereafter another idle time activity is begun. Thus, constant data delivery is provided by the preempting of the idle time activity which thereby maximizes data throughput.

Growing community of inventors

San Jose, CA, United States of America

James R Shipman