System and method for autonomic system management through modulation of network controls

• US 8,473,602 B2
• Filed: 06/11/2005
• Issued: 06/25/2013
• Est. Priority Date: 06/11/2005
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Liu, Zhen, Pendarakis, Dimitrios, Silber, Jeremy I., Wynter, Laura
• Primary Examiner(s): Nawaz, Asad
• Assistant Examiner(s): Waqas, Saad A
• Application Number:
  US11/150,351
• Publication Number:
  US 20060280161A1
• Time in Patent Office: 2,936 Days
• Field of Search: 709/224, 709/234, 709/223, 709/233, 709/229, 455/450, 455/13.1, 455/427, 370/389, 370/351, 370/232, 370/236, 370/400, 370/235, 370/395.21, 370/229, 370/392, 370/252, 706/14, 710/200
• US Class Current: 709/224
• CPC Class Codes: H04L 41/0816 the condition being an adaptation, e.g. in response to network events
• CPC Class Codes: H04L 41/083 for increasing network speed
• CPC Class Codes: H04L 43/0876 Network utilisation, e.g. volume of load or congestion level

Abstract

Systems and methods are provided to determine an allocation of network resources in a distributed on-demand information technology (IT) systems using existing control mechanisms for other operating system resources in order to achieve a desired operating point within the IT system. This desired operating point is obtained by optimizing a goal-based objective function while taking into account system constraints. The relationship between utilization of all system resources, i.e. network resources and processing resources, and attainment of performance objectives is autonomously obtained for a plurality of actions that could be required by a range of system applications. This relationship is used to allocate network resources to applications while maintaining desired performance objectives. The allocation is enforced using existing control mechanisms.

Claims

1. A method for processor and network resource management in a computing system, the method comprising:

identifying a plurality of performance objectives for a plurality of applications that utilize network and processing resources and are running on a given computer in a computing system containing a plurality of computers each running a plurality of applications, each identified performance objective dependent upon utilization of the network and processing resources by the application and comprising a prescribed performance target for execution of the application and a prescribed allocation of central processing unit processing power among the plurality of applications running on the given computer;

recording observations of utilization levels of the network and processing resources and associated realization levels for each performance objective over time during the execution of the applications, each observation comprising a sample point in a multi-dimensional space that comprises a separate dimension for each controllable processing resource and the utilization of that controllable processing resource, each allocatable network resource and the utilization of that allocatable network resource and each performance objective and the realization levels of that performance objective such that the utilization of controllable processing resources central processing unit power, disk space and main memory contained in the given computer and utilized by the applications and allocatable network resources communication bandwidth, communication buffer space, link bandwidth, network switching bandwidth, network interface buffer capacity, operating system buffer capacity, application buffer capacity, network switching buffer capacity and queueing order within the computing system and utilized by the plurality of applications and each one of the plurality of performance objectives and the realization of that performance objective are covered;

using the recorded observations to determine a multi-dimensional function expressing a dependency within the multi-dimensional space between realization of the identified performance objectives and utilization of the network and processing resources by the applications;

using the determined multi-dimensional function to identify a desired set of actions with regard to allocation of the network and processing resources to the applications that optimizes the identified performance objectives, the desired set of actions comprising increasing and decreasing the allocation of the network and processing resources among the applications;

maintaining functions expressing utilization of processing resources by each application as a function of network resource allocation to that application;

using the maintained functions to analyze network resource allocations proposed by the identified desired set of actions for potential effects on processing resource allocations and performance objectives;

scaling the proposed network resource allocations based upon the potential effects; and

implementing the desired set of actions with regard to the allocation of the network and processing resources and the scaling of the proposed network resource allocations.

2. The method of claim 1, wherein the step of using the recorded observations to determine the function further comprises using learning algorithms to determine the function.

3. The method of claim 1, wherein the step of using the recorded observations to determine the function further comprises using a stochastic control algorithm to determine the function.

4. The method of claim 1, further comprising updating the determined function and the calculated allocation of the network and processing resources continuously in real time.

5. The method of claim 1, wherein the plurality of computers communicate across a network.

6. The method of claim 1, further comprising identifying at least one constraint on the allocation of network resources.

7. The method of claim 1, wherein the step of implementing the allocation of network and processing resources comprises using network filters or network resource reservation mechanisms to implement the allocation of network resources.

8. A non-transitory computer readable storage medium containing a computer executable code that when read by a computer causes the computer to perform a method for processor and network resource management in a computing system, the method comprising:

identifying a plurality of performance objectives for a plurality of applications that utilize network and processing resources and are running on a given computer in a computing system containing a plurality of computers each running a plurality of applications, each identified performance objective dependent upon utilization of the network and processing resources by the application and comprising a prescribed performance target for execution of the application and a prescribed allocation of central processing unit processing power among the plurality of applications running on the given computer;

recording observations of utilization levels of the network and processing resources and associated realization levels for each performance objective over time during the execution of the applications, each observation comprising a sample point in a multi-dimensional space that comprises a separate dimension for each controllable processing resource and the utilization of that controllable processing resource, each allocatable network resource and the utilization of that allocatable network resource and each performance objective and the realization levels of that performance objective such that the utilization of controllable processing resources central processing unit power, disk space and main memory contained in the given computer and utilized by the applications and allocatable network resources communication bandwidth, communication buffer space, link bandwidth, network switching bandwidth, network interface buffer capacity, operating system buffer capacity, application buffer capacity, network switching buffer capacity and queueing order within the computing system and utilized by the plurality of applications and each one of the plurality of performance objectives and the realization of that performance objective are covered;

using the recorded observations to determine a multi-dimensional function expressing a dependency within the multi-dimensional space between realization of the identified performance objectives and utilization of the network and processing resources by the applications;

using the determined multi-dimensional function to identify a desired set of actions with regard to allocation of the network and processing resources to the applications that optimizes the identified performance objectives, the desired set of actions comprising increasing and decreasing the allocation of the network and processing resources among the applications;

maintaining functions expressing utilization of processing resources by each application as a function of network resource allocation to that application;

using the maintained functions to analyze network resource allocations proposed by the identified desired set of actions for potential effects on processing resource allocations and performance objectives;

scaling the proposed network resource allocations based upon the potential effects; and

implementing the desired set of actions with regard to the allocation of the network and processing resources and the scaling of the proposed network resource allocations.

9. The non-transitory computer readable storage medium of claim 8, wherein the step of using the recorded observations to determine the function further comprises using learning algorithms to determine the function.

10. The non-transitory computer readable storage medium of claim 8, wherein the step of using the recorded observations to determine the function further comprises using a stochastic control algorithm to determine the function.

11. The non-transitory computer readable storage medium of claim 8, wherein the method further comprises updating the determined function and the calculated allocation of the network and processing resources continuously in real time.

12. The non-transitory computer readable storage medium of claim 8, wherein the plurality of computers communicate across a network.

13. The non-transitory computer readable storage medium of claim 8, wherein the method further comprises identifying at least one constraint on the allocation of network resources.

14. The non-transitory computer readable storage medium of claim 8, wherein the step of implementing the allocation of network and processing resources comprises using network filters or network resource reservation mechanisms to implement the allocation of network resources.

15. A method for processor and network resource management in a computing system, the method comprising:

identifying a performance objective for an application that utilizes network and processing resources and is running on a computer in a computing system containing a plurality of computers each running a plurality of applications, the identified performance objective dependent upon utilization of the network and processing resources by the application and comprising a prescribed performance target for execution of the application and utilization of the network and processing resources;

recording observations of utilization levels of the network and processing resources and associated realization levels for each performance objective over time during the execution of the applications, each observation comprising a sample point in a multi-dimensional space that comprises a separate dimension for each controllable processing resource and the utilization of that controllable processing resource, each allocatable network resource and the utilization of that allocatable network resource and realization levels of the performance objective such that the utilization of controllable processing resources central processing unit power, disk space and main memory contained in the given computer and utilized by the applications and allocatable network resources communication bandwidth, communication buffer space, link bandwidth, network switching bandwidth, network interface buffer capacity, operating system buffer capacity, application buffer capacity, network switching buffer capacity and queueing order within the computing system and utilized by the plurality of applications and the realization of the identified performance objective;

using the recorded observations to determine a multi-dimensional function expressing a dependency within the multi-dimensional space between realization of the identified performance objective and utilization of the network and processing resources by the applications;

using the determined multi-dimensional function to identify a desired set of actions with regard to allocation of the network and processing resources to the application that optimizes the identified performance objective, the desired set of actions comprising increasing and decreasing the allocation of the network and processing resources among the applications;

maintaining functions expressing utilization of processing resources by each application as a function of network resource allocation to that application;

using the maintained functions to analyze network resource allocations proposed by the identified desired set of actions for potential effects on processing resource allocations and performance objective;

scaling the proposed network resource allocations based upon the potential effects; and

implementing the desired set of actions with regard to the allocation of the network and processing resources and the scaling of the proposed network resource allocations.

16. The method of claim 15, further comprising identifying a constraint comprising a minimum allocation of central processing unit processing power to any one of the plurality of applications running on the given computer at any time;

and

using the determined function and the constraint to calculate an allocation of the network and processing resources to the applications that optimizes the identified performance objective.