International Business Machines Corporation

Oliver Benke

Peter Bendel

Roland Seiffert

Tobias Ulrich Bergmann

Namik Hrle

Daniel Martin

Maryela Evelin Weihrauch

Ruiping Li

Knut Stolze

Stefan Letz

Thomas Lumpp

Hartmut Penner

Boas Betzler

Oliver Draese

Jan Kunigk

Klaus Theurich

Eberhard Pasch

Joerg-Stephan Vogt

Johannes S Kern

Felix O Beier

Utz Bacher

Terence Patrick Purcell

Thomas Gentner

Isabell Schwertle

Dennis Butterstein

Markus Mueller

Jan-Bernd Themann

Martin Reitz

Wolfgang Schaeberle

Bernd Jostmeyer

Walter Schueppen

Muhammad Usman Karim Khan

Dominic Röhm

A method for maintaining a memory replica of a primary computer system to be used by an alternate computer system is provided. Provided are detecting changes performed on a memory device of the primary computer system by software executed on the primary computer system. Also provided is monitoring at least one memory controller of the primary computer system; triggering, based on the monitoring, a snapshot creation procedure, the snapshot creation procedure comprising creating a memory snapshot that comprises the detected changes.

Maintaining a memory replica of a primary computer system

A computer-implemented method for seamlessly performing a maintenance operation on a stateful system includes mapping a network address of the stateful system to a primary server that uses a primary database to respond to incoming data requests. In response to receiving a maintenance request, the primary database is replicated to a secondary database of a secondary server. The secondary server is updated according to the maintenance request. The method further includes caching, in a replay buffer of the primary server, incoming data requests during the replicating. After the replicating, the data requests from the replay buffer are executed by the secondary server. Write operations to the primary server are disabled during the replicating, and the network address of the stateful system is mapped to the secondary server. Subsequently, the primary server is updated and reinstated by mapping the network address, and enabling the write operations.

Updating stateful system in server cluster

A method for distributed training in a parameter dataset; the method for at least one coordinating node; a network of at least one distributed node; receiving a parameter dataset from a first distributed node; decrypting the received dataset; training the parameter dataset based on a training data record to obtain a trained parameter dataset; and forwarding the trained parameter dataset to a second distributed node.

Distributed training in a parameter dataset

Methods and systems for generating database statistics. Table statistics in a metadata catalog of a source database system are observed, statistics generation costs utilizing a target database system are estimated, and source statistics generation costs utilizing a source database system are estimated. The statistics generation costs are compared and statistics generation queries by the target database system are triggered in response to the statistics generation costs utilizing the target database system having a predefined relationship with the source statistics generation costs utilizing the source database system. The statistics generation queries are performed by the target database system in response to the triggering by the source database system. The generated statistics are sent from the target database system to the source database system, the table statistics in a metadata catalog are updated based on the generated statistics, and the updated table statistics are used to optimize a query plan.

Offloading statistics collection

A processor(s) initiates a database transaction, in a computing environment that includes a database that includes one or more memory devices. The processor(s) forks a first address space that represents a current state of the database, to create a second address space. The processor(s) writes an entry indicating timing of the initiating to a log file and generates a file that is mapped to the one or more memory devices. The file includes differences in state between the current state of the database and a state subsequent to executing and committing the database transaction, and a timestamp indicating timing for committing the database transaction. The processor(s) write the database transaction to the second address space.

Database recovery using persistent address spaces

A method for merging transactions for execution by a database management system. The method includes a computer processor receiving a database query that is comprised of a plurality of transactions. The method further includes a computer processor placing each of the plurality of transactions, of the database query, in respective chronological order, in a scheduling queue. The method further includes a computer processor determining that at least a first transaction and a second transaction within the plurality of transactions each contains one or more transaction delineation descriptions. The method further includes computer processor responding to the determination that the first transaction and the second transaction, within the plurality of transactions, each contains one or more transaction delineation descriptions, and merging the first transaction and the second transaction to create a merged transaction combination and merging also includes combining two or more table-scan operations.

Merging database operations for serializable transaction execution

A method, system and computer program product for providing consolidated access to data of a plurality of source databases. Tables of each of the source databases are replicated to a shared accelerator. The source DBMSs are configured to dispatch queries to the accelerator for accelerating query execution. The accelerator is configured such that the replicated tables can only be accessed by the source DBMS having provided said tables for executing a dispatched query. A user can select one of the source DBMSs to act as a consolidated DBMS&#x2014;C-DBMS. The C-DBMS provides the consolidated access. The user is enabled to select tables managed by another one of the DBMSs. In response to receiving the selection of the tables, the accelerator is re-configuring such that the C-DBMS is granted access also to the copies of the selected tables in the accelerator.

Accelerator based data integration

Secure operations are performed on encrypted code. A processor in a first operating mode obtains encrypted code. The processor switches from the first operating mode to a second operating mode, and decrypts the encrypted code to obtain decrypted code. The decrypted code is executed, based on the processor being in the second operating mode, to provide a result. The result is encrypted, and the encrypted result is sent to a user, based on the processor switching back to the first operating mode.

Secure operations on encrypted data

Embodiments of the invention are directed to methods for improving performance of a multi-core processor. A non-limiting method includes increasing a first operating frequency to a first elevated operating frequency of a first core of a gang of cores, the gang of cores comprising a plurality of cores of the multi-core processor. The method further includes upon a determination that an operating temperature of the first core is above a threshold temperature, switching processing of a thread from the first core to a second core in the gang of cores. The method further includes reducing the first operating frequency of the first core. The method further includes increasing the operating frequency of the second core to a second elevated operating frequency.

Thread migration and shared cache fencing based on processor core temperature

A method for a continuous mutual extended processor self-test is provided. The method is implemented by a system including a plurality of cores. The system sets an operating condition for the continuous mutual extended processor self-test. An assist processor of the plurality of cores executes a test program that implements the continuous mutual extended processor self-test on a core under test of the plurality of cores. The system determines a pattern and a response during the test program execution and repeats the test program until the test program has finished or failed.

Continuous mutual extended processor self-test

A multiprocessor computer system facility is provided for selectively reallocating a logical processing unit. The logical processing unit is one logical processing unit of multiple logical processing units backed by a socket of a plurality of sockets of a multiprocessor computer system. The selectively reallocating includes: monitoring, during execution of program code, a performance metric of a hardware resource of the socket, the hardware resource being shared by the multiple logical processing units; and reassigning the logical processing unit, based on the monitoring, to a different socket of the plurality of sockets of the multiprocessor computer system to, in part, improve the performance metric of the hardware resource of the socket. The reassigning includes reassigning the logical processing unit based on the monitoring determining that a differential between a resource utilization rate of the socket and a resource utilization rate of the different socket exceeds a set threshold.

Performance-based reallocating of logical processing units to sockets of a computer system

A distributed database system (or federated database system) where a catalogue is used to store metadata descriptive of data containers for the constituent database system(s) and/or accelerator system(s) of the federated database. In response to the generation of the trigger signal corresponding to a database definition language (DDL) statement, the accelerator system updates metadata in the catalogue.

Growing community of inventors

Stuttgart, Germany

Oliver Benke