International Business Machines Corporation

Philip Matthew Dudas

Kenneth Wayne Boyd

John Jay Wolfgang

Kenneth Fairclough Day, Iii

Jason Lee Peipelman

William Frank Micka

Todd Charles Sorenson

Paul Anthony Jennas, Ii

Larry Juarez

David Montgomery

Joshua M Rhoades

Kenneth Fairclough Day Iii

A read/write ratio for each of a plurality of data segments classified in a hot category as hot data segments is determined. Each of the plurality of hot data segments is ordered by the read/write ratio in a descending order. Each of a plurality of available SSD devices is ordered by a remaining life expectancy in an ascending order. Those of the plurality of hot data segments are matched with those of the plurality of hot data segments with those of the plurality of available SSD devices such that a hot data segment having a higher read/write ratio is provided to an SSD device having a smaller remaining life expectancy than another hot data segment having a lower read/write ratio.

Sub-lun input/output profiling for SSD devices

Sub-LUN input/output profiling for SSD devices

Disclosed are a system, a method, and a computer program product to provide for the optimization of the output transfer load balance between the peer computers transferring data to one or more storage devices. The peer computers receive, organize and transfer the data to storage devices. The data set is composed of a plurality of data transfers. After an initial division of the data transfers between the two peers, each peer will have assigned responsibility for a number of data transfers. If the one of the peer computers completes offloading transactions earlier than the other peer, then the peer that is still transferring data will employ the other peer to execute a portion of the remaining data transfers. The operation of the system is symmetrical in that either peer may assist the other peer depending upon which peer has idle time. In addition the operation is autonomous and self-adjusting resulting in the peer nodes optimizing the size of the portion of data transfers that are reassigned during the operation of the invention resulting in the minimization of idle time for either peer. The self-adjusting feature allows the system to react to changing conditions that affect data transfer rates to the storage devices.

Autonomic learning method to load balance output transfers of two peer nodes

Disclosed are a system, a method, and a computer program product to optimize the overall latency of transferring data from peer computers to storage devices. The latency optimization occurs after a group of data is received and organized by the peer computers. The average amount of time to transfer data to each particular storage device from the peer computers is used to determine the assignment of data transfers between the peer computers. Each peer computer maintains continuously updated measurements of the average time to transfer data to each storage device. The updated transfer time measurements are based upon a moving average with adjustable weighting of past and present measurements enabling the system to adapt to changing conditions.

Autonomic predictive load balancing of output transfers for two peer computers for data storage applications

Disclosed is a system, method, and article of manufacture for maintaining consistent copies of data between a primary site and a secondary site. One or more backup components receive backup messages from the primary site describing transactions. Transactions received from the primary site at the backup components are referred to as 'inbound transactions.' The backup components associate these inbound transactions with an identifier such that all transactions associated with a certain identifier comprise all of the activity on the consistency group for a certain period of time. The backup captain periodically notifies the backup components to pause (i.e., stop receiving inbound transactions temporarily), and once all backup components indicate that they have paused, the backup components are notified to resume receiving transactions and associate the newly received transactions with a new identifier. At the same time, the backup components are sending transactions to the secondary storage devices. Transactions sent from caches at the backup components to the secondary site are referred to as 'outbound transactions.' The backup components send all transactions from the current outbound consistent transactions set, temporarily stop sending outbound transactions, and notify the backup captain that they have completed sending the current outbound consistent transactions set. When all backup components have completed sending the current outbound consistent transactions set, the backup captain causes consistent copies to be made at secondary storage and notifies the backup components to send the next outbound consistent transactions set.

Method, system, and article of manufacture for creating a consistent copy

Disclosed are a system, a method, and a computer program product to efficiently create consistent transaction sets to maintain one or more copies of data at different data storage sites. All transactions sent to a primary backup appliance during a consistent transaction set creation interval are formed into a consistent transaction set by efficiently adding new transactions as they are received and removing unnecessary transfers as newer versions arrive. When the creation interval has expired, the complete consistent transaction set is transferred to a secondary backup appliance to be used to update a consistent backup copy of the primary site data. For each consistent transaction set, there will be a tree data structure (a search tree) created that contains the addressing information for all of the blocks of data in the consistent transaction set. The tree data structure used is a modified splay tree, which is a specialization of a binary search tree such that accessed nodes are 'percolated' to the top of the tree for faster subsequent access. Secondary data consistency is maintained because the consistent transaction sets are applied whole at the secondary site, and after application, the secondary volumes are exact copies of the primary volumes at the time the consistent transaction set was completed.

Autonomic link optimization through elimination of unnecessary transfers

Disclosed is a system, method, and article of manufacture for maintaining data integrity. A frequency for creating consistent transactions sets is negotiated. A consistent transactions set creation duration is negotiated. At an interval determined by the frequency, during the consistent transactions set creation duration, processing of a first consistent transactions set is completed, while processing of a second consistent transactions set is delayed until the end of the consistent transactions set creation duration. After the first consistent transactions set have been offloaded to secondary storage, consistent copies of data may be created.

Growing community of inventors

Tucson, AZ, United States of America

Philip Matthew Dudas