International Business Machines Corporation

Joseph L Temple, Iii

Michael Ruffin

Timothy I Morrison

Allen H Preston

Ann C Merenda

Eva L Yan

Carlos A Ordonez

Brian William Curran

Robert Dov Herzl

Anthony S Coscarella

Kristin R Jayaram

William John Starke

Jeffrey Mark Nick

Jeffrey Paul Kubala

Peter Bergersen Yocom

Michael E Baskey

David Andrew Schroter

Kenneth Anthony Lauricella

Richard Frank Rizzolo

Lishing Liu

Charles B Winn

Frank J Degilio

Robert W St John

David R Pruett

Michael Gerard Nealon

Jih-Kwon Peir

Allan Rowe Steel

Dennis Felix Clocher

Joseph John Lisowski

Linda L Quinn

John J Ellison

Robert E Vik, Ii

Jack A Silver

William D Reeder

Joseph M D'Onofrio

Christian F Rohrbach

Christopher D Pratt

Richard N Fuqua

Louis J Milich

James S O'Higgins

Paul M Moore

Peter A Nagel

Anderson H Hunt

Joseph O Celtruda

Donald Jue

Kein A Hua

Rachmil Jacobovits

Luis I Hernandez

A method includes determining an anticipated usage pattern for an information technology system and a desired service level for the information technology system. The method includes determining a utilization saturation point for the information technology system and selecting a number of servers in the information technology system. The method also includes calculating one or more performance characteristics for the information technology system based on the anticipated usage pattern, the desired service level, the utilization saturation point, and the number of servers and comparing the one or more performance characteristics for the information technology system with one or more desired performance characteristics for the information technology system.

Facilitating the design of information technology solutions

According to an exemplary embodiment of this disclosure, a computer-implemented method may include selecting a first machine to be sized. A first throughput value and a first capacity value may be obtained for the first machine. A base machine may be selected. A base throughput value and a base capacity value may be obtained for the base machine. The first throughput value may be normalized, with a computer processor, to the base throughput value to product a normalized throughput value. The first capacity value may be normalized to the base capacity value to produce a normalized capacity value. A workload weight may be determined for the first machine, based at least in part on comparing the first machine to the base machine. A workload metric may be calculated as a weighted average of the normalized throughput value and the normalized capacity value.

Hardware sizing

Utilization of at least one resource is managed in a computer system by monitoring usage of capacity of the computer resource, including monitoring usage of base capacity of the computer resource and marginal capacity of the computer resource exceeding the base capacity and determining whether usage of the computer resource exceeds a predetermined threshold. The determination may be based on an average usage of multiple computer resources over the predetermined time period. If usage of the computer resource exceeds the predetermined threshold, a charge for the usage of the computer resources is adjusted by an amount based on the usage of the computer resources exceeding the threshold.

Method, computer program product and system for managing usage of marginal capacity of computer resources

A data caching approach is provided for a distributed computing environment employing coherent data caching. The data caching approach includes dynamically deciding whether to associate a priority tag with requested data for a processing unit, wherein the priority tag is to be employed in deciding whether to hold the requested data in a local cache associated with the processing unit. The dynamically deciding includes determining whether latency of arrival of requested data at the processing unit meets a set threshold, and if so, associating the priority tag with the requested data. The priority tag indicates that at least one of the requested data is to be captured in the local cache or the requested data is to be maintained in the local cache over other data without an associated priority tag. In one implementation, the threshold is a programmable threshold, and the local cache is a lookaside cache.

Method, system and computer program product for data caching in a distributed coherent cache system

A partitioned processing system is disclosed wherein a workload calculation is implemented for at least one partition. CPU utilization and I/O throughput information is used in calculating resource control parameters which is indicative of resource usage for the partition. The system collects and analyzes the computer system capacity data in a partition which determines a sizing metric. The system includes a manager in the computer system which issues a command to obtain throughput information of a computer system first partition, and issues a command to obtain resource utilization information of the computer system first partition. The manager then calculates a resource control parameter using the information obtained. The system includes a monitor which indicates resource allocation responsive to the resource control.

Inter-partition message passing method, system and program product for throughput measurement in a partitioned processing environment

A programmable prefetch mechanism is presented for prefetching instructions for a processor executing a program, and in particular a non-procedural program such as object-oriented code. The prefetch mechanism includes prefetching instructions from memory which are sequential instructions from where the processor is currently executing in a sequence of instructions of the program, and when the prefetching encounters a new update prefetch stream (UPS) instruction, the prefetching includes executing the UPS instruction and subsequent thereto, branching to a new memory address for prefetching of at least one non-sequential instruction from memory for execution by the processor. The UPS instruction can be inserted into the program at compile time and when executed causes the loading of a prefetch buffer in the prefetch mechanism which in one embodiment includes a set associative array of x,y address pairs. When an incremented prefetch address is matched to an x address of the array, the prefetching branches to the new memory address y paired with the matching x address in the prefetch buffer. In this manner, cache misses associated with unconditional branches to non-sequential instructions are avoided.

Programmable prefetching of instructions for a processor executing a non-procedural program

A partitioned processing system capable of supporting diverse operating system partitions is disclosed wherein throughput information is passed from a partition to a partition resource manager. The throughput information is used to create resource balancing directives for the partitioned resource. The processing system includes at least a first partition and a second partition. A partition resource manager is provided for receiving information about throughput from the second partition and determining resource balancing directives. A communicator communicates the resource balancing directives from the partition manager to a kernel in the second partition which allocates resources to the second partition according to the resource balancing directives received from the partition manager.

Inter-partition message passing method, system and program product for managing workload in a partitioned processing environment

A method, system and program product for determining a cost differential resulting from a migration of computational workloads between computer platforms is disclosed herein. Information regarding current and planned hardware and software use on the computer platforms and the costs associated therewith is provided and supplemented with model data including historical data and industry averages. A cost and capacity measurement associated with the current and planned use and cost is derived. A required processing capacity migration between platforms is determined for a proposed computational workload migration therebetween and a resultant cost and capacity measurement is calculated for each platform to determine the resultant cost differential.

Method, system and program product for evaluating a computational processing capacity migration between computer platforms

Instruction and data registers of processors of a multiprocessing computing system are joined and forked to allow processing in multiple modes of operation. When joined, the registers of the processors each contain a same piece of information, hence generating single instruction and data streams. In contrast, when forked, the registers of the processors contain different pieces of information, thereby generating multiple instruction and data streams. Additionally, information may be stored into partitions of memory and fetched and broadcast by processors local to the particular memory sections thereby resulting in a faster cycle time.

Composite uniprocessor

A method, system and program product for matching the an information technology enhancement project with the resource and priorities of an enterprise is presented herein. A set of objectives for the information technology resources is defined and prioritized as to their importance to the enterprise. The information technology resources for the enterprise are partitioned into segments along a predetermined common property such as geography or skill-base. Each segment is compared to a set of desire information technology characteristics and a score for each characteristic is assigned to each segment and summed providing a total characteristic score for each segment. The total characteristic score is weighted in light of the prioritized objects by virtue of a correlation created between the objectives and the characteristics. The weighted scores are ranked and then classified in accordance with a set of potential information technology enhancement projects to facilitate the selection of one or more of the projects.

Information technology project assessment method, system and program product

Method, system and program storage device are provided for implementing a register pipe between processing engines of a multiprocessor computing system. A register pipe includes at least one first register of a first processing engine and at least one second register of a second processing engine. Data is transferred between the first processing engine and the second processing engine through the register pipe without passing through memory. In one embodiment, general purpose registers within the first processing engine and within the second processing engine are employed to implement the register pipe. A control mechanism is provided within each processing engine to dynamically enable or disable the register pipe coupling any two processing engines of the multiprocessor computer system. A technique for broadcasting to multiple register pipes and for implementing barrier synchronization using a register pipe addressed to a processing engine itself are also provided.

Register pipe for multi-processing engine environment

Synchronization &ldquo;states&rdquo; of programmable length imposed on processing nodes are disclosed, the last of which contains an event requiring action by fully synchronized nodes. The nodes are shifted between the states using a shared timing source to ensure synchronization for the event. The first state is a &ldquo;non-impending event&rdquo; state where the nodes continue their routine processing; the second state is an &ldquo;impending event state&rdquo; closer to the event, and within which nodes &ldquo;wait&rdquo; for the event by suspending their routine processing. In the third state, immediately preceding the final, event state, any non-waiting nodes are interrupted for the impending event. The present invention therefore allows some node autonomy during the first state within which the nodes poll at their leisure, and during the second, &ldquo;wait&rdquo; state, but during the third, interrupt state, any nodes not yet arriving by polling are interrupted.

Multi-node synchronization using global timing source and interrupts following anticipatory wait state

A method, apparatus and program product for determining a required processing capacity to accommodate a computational workload migration from a first computer system to a second computer system is described herein. A processing benchmark for the workload on the first computer system is determined along with a scaling factor for converting the processing benchmark to the second computer system. An adjusted processing benchmark is then determined for the workload on the second computer system and is multiplied by the number of computers in the second system on which the workload is to be run to provide a total processing benchmark for workloads on the second computer system. From the total processing benchmark a mean and maximum utilization factor may be derived to determine the average and peak use of the workload during a predetermined time period. A skew factor may further be derived to represent the excess capacity required on the first system due to the imbalance of the workload across multiple computers. The invention may be repeated for multiple workloads to determine the required capacity for all workloads on the second computer system.

Computational workload-based hardware sizer method, system and program product

The present invention provides a method for optimizing an IT environment, the method incorporating relational databases in conjunction with unique data acquisition and querying techniques providing a flexible query feedback of IT software, hardware and network parameters.

Method for managing a heterogeneous IT computer complex

A method, system and program product for determining the value to an enterprise of a proposed technology modification is presented herein. The information technology resources of an enterprise are partitioned into segments along any number of various lines such as business process, geography, etc. The partitioning creates one or more sets of partitioned segments. Within a given segment the resources are profiled in accordance with the information technology priorities of the enterprise and mapped against the complexity of the proposed modification to derive an opportunity score. The opportunity scores for the profiled segments are examined to determine if the partitioning has been effective and if not, the process is repeated. Once an effective partitioning has been effected the opportunity scores of the segments indicate a measure of the respective value of the proposed modification within each segment.

Method, system and program product for determining the value of a proposed technology modification

Growing community of inventors

Hurley, NY, United States of America

Joseph L Temple, Iii