This page gives an overview of the steps to install the various software components needed to enable Job Performance data in XDMoD. Instructions for upgrading the software from are available on the upgrade page
The Job Performance (SUPReMM) XDMoD module should be installed on an existing XDMoD instance. The XDMoD instance must have job accounting data shredded and ingested and present in the UI. Do not begin the install procedure until the accounting data is loaded into XDMoD. See the main XDMoD documentation for instructions on installing and configuring XDMoD.
The steps to install and configure the software are as follows:
It is important that the various software components are installed and configured in the order listed. For example, if the XDMoD module is not installed and configured then the summarization software will not be able to access the tables in XDMoD datawarehouse that are needed.
The Job Performance (SUPReMM) XDMoD module version number must be identical to the version of the base XDMoD install. The versions of the summarization software and PCP software for a given XDMoD version are listed below.
|Package||Recommended Version||Compatible Version(s)|
|PCP||3.12.2||3.11.x - 3.12.x|
The SUPReMM software has been tested with MongoDB version 3.4.15. We expect that the software is compatible with any supported release version of MongoDB.
The summarization software should be compatible with the 4.x releases of PCP. However the XDMoD team have not tested this configuration and will not be able to provide support for it.
The detailed list of software packages for each subsystem is listed in the respective system requirements pages.
The XDMoD Job Performance (SUPReMM) module must be installed on an existing, functional XDMoD instance.
The PCP data collection software must be installed on the existing HPC compute nodes.
The Job summarization software and Mongo database may be installed on the same server as the XDMoD instance. However, for best performance, it is recommended that a dedicated server is used to host the Mongo database. This server can also be used to run the Job summarization software.
We do not recommend installing databases or webservers on any HPC compute nodes.
The amount of data storage depends on the amount of historical data that is desired to be stored.
The following estimates are based on using the default configuration:
These estimates were obtained from the average storage usage for the SUPReMM system installed on the HPC resource at CCR Buffalo. The actual quantity of data storage required depends on many factors including the rate of job creation and which metrics are collected.