I have been successfully collecting the accounting cards for a long time. I have them from the very first IPL of this z890. I am not doing chargeback. I would not do charge back from the accounting cards, I would do it from the Monitor data stream because the per user utilization and the rest of the system utilization is in a single place for analysis.
Needless to say, I would love to buy all the great PRODUCTS that people sell, but I have no budget. Anyway, that would be going forward in time. My current need is to process some data from BACK IN TIME. I have to work with data from FY2007 (Oct 1 2006 to Sep 30 2007) and I don't have the detailed Monitor data to process with ESALPS or even with PerfTK that I do have. I have accounting cards and rexx and pipelines and wanted to see if there were any use of NEW tools in LINUX for processing this data. Apparently no one has thought of using new tools (new to VMers at least) for processing old data. I will go back to using rexx or if I am up late at night sometime I might resurrect the FORTRAN G compiler I got from the Waterloo Mods tape and do it all in my first programming language. /Tom kern -----------Original Message----------------- From: Rob van der Heij <[EMAIL PROTECTED]> To set up something on CMS to collecting the account records is very easy. Either set up 2 users to manage that, or a simple pipeline to do it. The hardest part is to ensure your disk does not fill up. My experience is that when you have accounting data and charge-back based on that, customers will often challenge the reported usage and related charges. When they normally use 1 hour of CPU per day and suddenly one day 10 hours, they often claim it must be bad accounting since they don't remember doing more work. I have been at an installation where on average 40% of the CPU capacity was used by looping servers and other operation problems. When you have detailed monitor data available, you can still tell when those 10 hours were consumed. And when you have process data available as well, as you suggest, it is even possible to point at the process consuming those CPU hours. This does require correct CPU usage inside Linux, and sufficient high capture ratio to be relevant. Unfortunately the new Linux instrumentation runs short on both aspects. Rob PS Needless to say that ESALPS keeps performance history for such analysis and collects CPU usage with typically > 99.5% capture ratio, so good enough for accounting purposes as well. -- Rob van der Heij Velocity Software GmbH http://velocitysoftware.com/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
