International Business Machines Corporation

Dieter Wellerdiek

Elpida Tzortzatos

Harris M Morgenstern

Robert Miller, Jr

James H Mulder

Qais Noorshams

Tobias Orth

Bernard R Pierce

Karla Kay Arndt

Robert M Abrams

Nicholas C Matsakis

Norman Christopher Böwing

Marco Selig

Horst Sinram

Karin Genther

Tobias Huschle

Anastasiia Didkovska

Toni Pohl

Norman C Böwing

Friedrich Matthias Gubitz

A computer-implemented method includes storing usage of a resource of said computer system as time-stamped resource usage values, comparing said time-stamped resource usage values with predetermined time-stamped performance goal values, assigning a time-stamped priority value to an application that is running based on at least one of said performance goal values, identifying a future workload demand value by applying a time-series analysis algorithm to at least some of said time-stamped resource usage values and a corresponding at least some of said time-stamped performance goal values for said application resulting in workload demand time frames and related amplitudes of said workload demand time frames, and adjusting a dispatch priority value for said application by setting a minimum dispatch priority for said application based on said future workload demand value.

Computer system workload manager

Aspects of the invention include determining, by a machine learning model, a predicted workload for a system and a current system state of the system, determining an action to be enacted for the system based at least in part on the predicted workload and the current system state, enacting the action for the system, evaluating a state of the system after the action has been enacted, determining a reward for the machine learning model based at least in part on the state of the system after the action has been enacted, and updating the machine learning model based on the reward.

Workload management using reinforcement learning

A method, system, and computer program product are provided for performance anomaly detection. Velocity data is periodically received from a workload manager for one or more address spaces. An expected velocity value is created for each of the one or more address spaces. A factor of the expected velocity value is compared to a current velocity value from the velocity data. Based on the current velocity value being lower than the factor, a remedial action is generated indicating an anomaly.

Programmatic performance anomaly detection

A method for enhancing a workload manager for a computer system includes sampling and storing usage of a resource of the computer system as resource usage values, comparing said resource usage values with predetermined performance goal values, assigning a time-stamped priority value to an application that is running based on at least one of the performance goal values by the workload manager, retrieving a portion of the resource usage values and a related portion of the performance goal values for the application, identifying a future workload demand value by applying a time-series analysis algorithm to the resource usage values and the performance goal values for the application resulting in workload demand time frames and related amplitudes of the workload demand time frames, and adjusting a dispatch priority value for the application by setting a minimum dispatch priority for the application based on the future workload demand value.

Memory management is provided which includes a page replacement process managed by a storage manager and a workload manager. The page replacement process swaps out the content associated with a frame of physical memory to an auxiliary storage in order to provide a free frame. The memory management process includes: determining that the physical memory runs out of free frames; providing priority information from the workload manager to the storage manager, the priority information indicating the priority or business relevance of a certain process; selecting one or more pages to be swapped to the auxiliary storage based on the priority information; and swapping out the contents of the one or more selected pages to the auxiliary storage.

Importance based page replacement

In one embodiment, in accordance with the present invention, a method, computer program product, and system for accessing non-functional host capabilities and services of a host operating system, by a guest operating system, are provided. The method includes a hypervisor, of a host OS, intercepting a hardware instruction from a guest operating system, wherein the hardware instruction is associated with data located in a register and in memory that is accessible by the hypervisor. The method further includes analyzing the hardware instruction and data to determine if the hardware instruction is a non-functional service request. The method further includes, upon determining that the hardware instruction is a non-functional service request, executing the related non-functional service request and writing a response to the associated memory. The method further includes notifying the guest operating system that the execution of the non-functional service request is complete and the response is available.

Accessing host services for virtual guest operating systems

A method for accessing non-functional host capabilities and services of a host operating system, by a guest operating system, is provided. The method includes receiving a function code and additional data from an application of a guest operating system, wherein the function code corresponds to a non-functional service request on a host operating system. The method further includes creating a parameter control block using the function code and the additional data in memory of the guest operating system, wherein the parameter control block is accessible by a hypervisor of the host operating system. The method further includes issuing a hardware instruction by the guest operating system, wherein the hypervisor of the host operating system receives control. The method further includes receiving a notification from the hypervisor that the execution of the non-functional service request is complete.

Technical solutions for reducing page invalidation broadcasts in virtual storage management are described. One general aspect includes a method including allocating, by a storage manager, a virtual memory page to a memory buffer that is used by an application being executed by a multiprocessor system, the virtual memory page being allocated from an address space of the application. The method also includes recording, by a memory management unit, a mapping between the virtual memory page and a physical location in a memory. The method also includes in response to a request, from the application, to deallocate the memory buffer, delaying invalidation of the mapping between the virtual memory page and the physical location in a memory, based on a count of free frames in the address space of the application.

Reducing page invalidation broadcasts in virtual storage management

The invention relates to a method for storage management in a data processing system having a plurality of storage devices with different performance attributes and a workload. The workload is being associated with respective sets of data blocks to be stored in said plurality of storage devices. The method comprises the steps of dynamically determining performance requirements of the workload and dynamically determining performance attributes of the storage devices. The method further comprises the step of allocating data blocks to the storage devices depending on the performance requirements of the associated workload and the performance attributes of the storage devices.

Storage management in a data processing system

A method for automatically determining performance problems in a computer system due to a metric indicating a current memory peak load in the computer system is disclosed. In order to decrease CPU consumption in large memories it is proposed to perform the steps of:

Growing community of inventors

Ammerbuch, Germany

Dieter Wellerdiek