HP 3000 Manuals HP 3000 Manuals
Managing Data Effectively Forecasting is a technique that predicts the future based on past trends. The more information you have about the past, the better your chance of accurately predicting the future. HP RXForecast relies on the HP LaserRX/MPE SCOPE collector to provide the data on which to forecast. By collecting system information for a long time, you can more accurately recognize trends and predict system resource requirements. You can use the HP LaserRX/MPE EXTRACT program to build logfiles that are optimal for HP RXForecast use. EXTRACT lets you extract data from either raw logfiles (specifying the LOGGLOB file) or from previously extracted logfiles. For HP RXForecast, you would use both of these features to create ideal logfiles. For information about how to use EXTRACT, see the HP LaserRX/MPE User's Manual: Collection Software. For example, suppose you size all SCOPE logfiles to hold 40 days of data. This provides a normal month of data and 10 days of additional space if needed. Monthly, you use EXTRACT to create an extracted logfile with all data, including global summaries and detail, application summaries and detail, and process detail. You could name this file using a naming convention of systemyymm. Thus, if the system name were MOE and the data pertains to January 1989, the filename would be MOE8901.
NOTE If you upgrade your system, you should not extract logfiles to the same system name. If you do, you will get misleading information about the system.
Then, from the extracted file you have created, you would extract only global and application summaries and append this data to the yearly summary file (called, in this example, MOE1989) for the same system. You would archive each monthly file to tape when the new monthly extract is completed, but keep the yearly file on line for forecasting and long-term analysis. Even after you create yearly files, you can use EXTRACT to append them together to build a file that contains several years of data. As needed, you can restore the monthly files from tape and extract other sets of data, or just use the detail information for in-depth analysis. It is imperative that you extract all data (summaries and detail) to the monthly logfiles. You cannot get data from a logfile on a subsequent extraction that is not included in the original extraction. If you extract only summaries from a logfile, you cannot get detail data on it later. You cannot recreate detail data from summaries. And remember, you do not have to stop the SCOPE collector to run EXTRACT on a logfile, even a logfile that is currently being used by the SCOPE collector.
Tip We strongly recommend that you forecast from extracted logfiles (hourly summaries), rather than from raw logfiles (5-minute summaries). Forecasts from raw logfiles are needlessly time-consuming. Also, there are unusual instances in which the raw logfiles will give you a different forecast than one based on an extracted logfile.
Maintaining Forecast Logfiles It is recommended that you continually maintain both global and application logfiles for forecasting. Ideally, you will keep up to 3 years of data from which to forecast. That much data could become prohibitively large unless you maintain separate global and application logfiles. You can use the EXTRACT program to extract global and application data every month and append these monthly logfiles to your ongoing global and application forecast logfiles. Furthermore, you can use this "extract and append" technique to eliminate unusual or irrelevant data from your forecast logfiles. Such data would be any data from a time period that you consider atypical for your system, such as holidays. For instance, you may want to forecast from data spanning October through January. You know, however, that this period was marked by several holidays during which your company experienced little activity. Any forecast you make from all of the data is not likely to be valid. But you don't want to ignore all of the data just because some of it is unusual. The solution is to extract and append. That is, you create extracted logfiles that exclude unusual activity and append them together for forecasting. In our example, you would extract a logfile up to the Thanksgiving holiday (November 22 and 23, 1990). You would append this logfile to your ongoing forecast logfile. Then you would extract and append data from the beginning of December up to the Christmas and New Year holidays. The forecast will be based only on those data points included in the logfile, so the graph will contain blank spaces for the periods of activity you eliminated. Note that zeros are NOT substituted for the missing data, as this would skew the forecast. Instead, in creating the forecast, HP RXForecast just disregards the dates you've excluded. By maintaining global and application logfiles of data typical of your system, you can use HP RXForecast effectively to forecast system resource utilization.