System, method and program product for forecasting the demand on computer resources

• US 7,415,453 B2
• Filed: 07/05/2005
• Issued: 08/19/2008
• Est. Priority Date: 07/08/2004
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Suzuki, Yasuhiro, Yamamoto, Naoya, Yamamoto, Katsuyoshi
• Primary Examiner(s): Vital; Pierre
• Assistant Examiner(s): Vo; Truong V
• Application Number:
  US11/174,981
• Publication Number:
  US 20060010101A1
• Time in Patent Office: 1,141 Days
• Field of Search: 707/2
• US Class Current: 1/1
• CPC Class Codes: G06F 11/3409 for performance assessment
• CPC Class Codes: G06F 11/3452 Performance evaluation by statistical analysis
• CPC Class Codes: G06F 11/3476 Data logging G06F11/14, G06F11/2205 take precedence
• CPC Class Codes: G06F 2201/87 Monitoring of transactions
• CPC Class Codes: G06F 2201/875 Monitoring of systems including the internet

Abstract

An automated system and method for predicting a future resource utilization tendency by analyzing a change in resource usage on the basis of a real operation condition in a computer system. A CPU utilization estimating unit acquires a resource log and a transaction log, and estimates utilization for each transaction, and information on the number of executions for each transaction, a processing amount predicting unit performs a statistical time-series analysis on the basis of time-series data of the number of executions for each transaction included in the transaction log and predicts an amount of processing for each type of a transaction, and a resource utilization predicting unit predicts the computer resource utilization or demand on from the estimated resource utilization for each transaction and the future predicted amount of processing for each transaction.

Claims

1. A computer implemented method for monitoring a server, comprising the steps of:

acquiring a first log containing resource utilization data for a plurality of times and a second log containing transaction data from the server at corresponding times, and storing the first log and the second log in a storage device of the computer;

estimating resource utilization for each type of transaction on the basis of resource utilization data at the plurality of times included in the first log and information on a number of executions processed of each transaction type included in the second log by calculating an average value of resource utilization data for a fixed time period, calculating an amount of processing for each transaction type for the fixed time period from the number of executions processed of each transaction type, and estimating the resource utilization for each type of transaction by a multiple linear regression analysis of the average value of resource utilization data and the amount of processing for transaction type;

performing a statistical time-series analysis on time-series data of the number of executions for each transaction type included in the second log by identifying a plurality of time series models for each transaction type using the time-series data of the number of executions for each transaction type, comparing each time-series model of the plurality of time-series models for each transaction type to the time-series data of the number of executions for each transaction type to determine an error for each time-series model, selecting a time-series model for each transaction type from the plurality of time-series models for each transaction type which has the smallest error and predicting a future amount of executions for each transaction type using the selected time-series model for each transaction type; and

predicting server resource demand on the basis of the resource utilization for each type of transaction estimated by the estimating step and the future amount of executions for each transaction type predicted by the predicting step by summing products of the resource utilization for each type of transaction and the future amount of executions for each transaction type.

2. The computer implemented method according to claim 1, wherein the performing step comprises:

excluding a part of the time-series data and identifying the plurality of time-series models for each transaction type on the basis of remaining data;

predicting the number of a future amount of executions for each transaction type corresponding to the excluded part of the time-series data by each of the plurality of time-series models; and

comparing a result of the prediction with actual data of the excluded part to determine an error for each time-series model.

3. The computer implemented method according to claim 1, wherein said resource comprises one of a CPU (central processing unit) utilization, a memory usage, a storage capacity of a storage device, and a network usage.