 |
Features
- This is OraPub Collector main page. All updates and related details will be posted on this web-page.
- This tool allows one to easily collect multiple workload data providing very business centric and robust forecasts. It collects Oracle data using dbms_monitor (which does not require any special license) and from v$sysstat, v$osstat, and v$system_event. The data collected can easily be used to determine Oracle CPU consumption, Oracle IO consumption (r/w, IOPs, MB/s), IO subsystem response time (single and multi-block reads and also writes) for the entire system and also for specific workload classes.
- The collector download contains both the single collector shell script and a How To step-by-step Word document. The document introduces the topic, explains how it works, shows you how to make it work for you, and how to load the data into Excel. It's a very rough document but it does the job for now.
- The first two introductory sections of the document are shown below. Enjoy and please send feedback to craig @ orapub.com. We look forward to hearing from you!
- The Need
- While forecasting based upon a single and general workload (for example, user calls) can be surprisingly useful and precise, many forecasts require a more business-centric workload characterization. For example, if the question to be answered is something like, "If we double the OE users in Hong Kong, can the system handle the load?" two workload categories are required; Order Entry and the remaining workload.
- The Challenge
- In the past, an expensive and potentially disruptive data collection facility had to be used to gather this type of information.
- Starting with any Oracle 10g license, the dbms_monitor package is supplied and can be used to effectively gather very specific workload category statistics with no noticeable impact on a production system. While dbms_monitor is usually thought of as a tracing facility, it also allows Oracle statistics (think v$sesstat and v$sysstat) to be gathered at targeted business and technical workloads. Any session associated with a specific v$session.client_identifier, service, module, or action can have workload statistics gathered and subsequently characterized. And in addition to this, with the aid of a simple logon trigger, a session can be identified by any combination of service, module, action, username, terminal, program, ip address, OS username, sid, server process pid, etc. If the workload category can be identified by some combination of these parameters, by using dbms_monitor you can easily capture its activity.
- Change History
- 3c. Add the capability to gather IO consumption (reads and writes by server process and background process) and IO response times (single and multi-block reads and writes). The actual interval duration is also included in the output file.
- The script was written based upon Linux, but the only change you may need to make is how the CPU utilization information is captured from sar and the date call. For HP-UX, look for lines similar the below and replace them with the below changes:
- Change 1: seq=`date +%S`
- Change 2: echo "$seq $line $full_date" | awk '{print $1 "," $10 " " $11 ",sar,cpu busy," ($3+$4+$5)/100 }' thequote="'" >> $wlc_file
|
|
|
 |
|
 |
|
