Browse Prior Art Database

A Methodology for Automated Process CPU and Network Transaction Throughput Management

IP.com Disclosure Number: IPCOM000149787D
Original Publication Date: 2007-Apr-06
Included in the Prior Art Database: 2007-Apr-06
Document File: 15 page(s) / 94K

Publishing Venue

IBM

Abstract

In today’s network transaction monitoring environment, both network applications and monitoring tools face tremendous high transactions volume in tens of thousands per minute. All programs stress to their maximum in terms of design logics, programming perfection, and resources utilization efficiency, and competing for the same scarce CPU and storage resources especially under high workload scenarios. They are all requiring more CPU and storage to handle high transaction volume at the same time. The network and system monitoring tools play an important role in gathering and detecting useful customer operation and service matrixes such as transaction response time, volume, resource consumption, and exceptions. Nevertheless, the tools are not as important when compared to customer’s production WEB server, database management system, and application programs as they are the vital components of customer on-line business. Consequently, the monitoring tools cannot require significant portions of system resources and degrade customer production program operation and end-user transaction response time regardless of workload volume. The monitoring tools must be able to: a) Accurately determine its system resource usage. b) Automatically adjust and adapt its own process controls and operate within customer defined CPU and storage limit. c) Delivering maximum possible useful output within resource limit under stress. No automated self resource control solution found. The customers commonly assign unessential jobs/tasks at lower dispatching priority and virtual storage size so that they have lesser opportunity to run and gain system resources while higher priority jobs/tasks are busy. However, there is no assurance that the lesser important jobs, once started, will not monopolize system resources and impact system operation. Further, most programs, business programs or tools, shutdown, hang, or crash when unable to obtain or exhausted processing resources. Operation scenarios must be prevented and avoided.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 22% of the total text.

Page 1 of 15

A Methodology for Automated Process CPU and Network Transaction Throughput Management

The Automated Process CPU and Network Transaction Throughput Management is a closed continuous feedback implementation that periodically examines its process CPU usage, determines current operation state based on current and past observations, selects appropriate action, if it deems necessary, perform programmed action, and monitoring action results. The algorithm and methodology can be modified, fine-tuned, and further enhanced based on application objectives, system configuration, workload characteristics, and specific customer requirements.

Interval Timer

  Calculate current interval CPU Usage

Timer Expired

       Set Interval Timer

   Determine Interval CPU State (Normal, Target, or critical)

   Perform Action Procedures

    Determine Current Operation State per this interval state and historical interval states

Determine New action based on process active state and action in effect

   Determine Process Active State based on this interval state and current states

The ITCAM RTT WRM/WSA product Analyzer agent component captures network packets from physical network, processing network protocol, such as TCP/IP, application protocol, such as HTTP, and formats on-line transaction response time data. Under very high network load, it may require high CPU and storage. By leveraging this self control methodology, it successfully maintains CPU usage at or below customer defined target limit and delivers maximum possible transaction response time reports proving its monitoring usefulness without impacting customer's

1

Page 2 of 15

internet operation under extremely high network volume. For example, The Analyzer's CPU limit is set at 45%. In the chart below, the Analyzer successfully tracks its CPU usage at below preset target zone for a given workload volume (60,000+ per minute) of this particular machine. The transaction arrival rate oscillated around 38,000 per minute per Analyzer workload adjustment.

Analyzer CPU Management on AIX Machine ELAIX03

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93

80

70

60

50

40

30

20

CPU Percent CPU Target Trans Per Minute

3. Description: Describe how your invention works , and how it could be implemented , using text, diagrams and flow charts as appropriate .

The following three parameters are used in explaining operation of this methodology. CPU_TARGET_THRESHOLD

     This parameter defines maximum program CPU usage limit in percentage. CPU_MANAGE_PERIOD

     This parameter defines CPU time calculation period, such as 60 seconds. CPU_ACTION_INTERVAL
This parameter defines the number of consecutive CPU usage management periods needed in order for program to initiate adjustment actions for maintaining its CPU consumption within the desired target zone, such as 2.

STEP 1 - CPU Time Calculations


The first step in CPU management is accurately determine own process's CPU usage in a given time in...