3.1.1.2.1 Virtualized environment modelling (that is, Hypervisor / VMware hosted systems)
The sizing provided by IBM ICCP assumes that the host operating system has exclusive access to the hardware resources available to the server.
In a virtualized environment, this level of availability cannot be assumed.
There is overhead for running a virtualized instance which is above the loading calculated for a conventional hardware and operating system environment, that will result in increased CPU utilization.
The extent of the overhead will vary depending on multiple factors, including how many unique virtualized instances are running. Due to this IBM ICCP cannot be used to assess the loss in performance due to virtualization. This overhead is difficult to quantify and therefore, to predict to what extent there will be a loss in performance.
3.1.1.2.2 IBM FileNet Process Engine Server Farming & Sizing
Empirical studies have shown that the capacity of an N server PE farm should be about N times the capacity of a single non-farmed PE server, as long as the system is not limited by the database server or some other external resource. So, to size for a PE Server farm, simply set the hardware configuration to have the number of PE servers.
IBM ICCP spreads the workload equally across all of the configured PE servers in the Server Farm.
3.1.1.2.3 IBM ICCP Interface Overview and Layout
The IBM ICCP Interface consists of the following graphical elements:
• Pull-down menus: File, Edit, Project, Walkthrough Wizard, View, Settings, and Help.
• A Toolbar, providing quick access to many of the most used options selected from the pull-down menus.
• A Treeview control, which features a list of nodes. Treeview is in the left pane of the IBM ICCP main window.
3.2 Inputs required from the client
The IBM CCP utilization is analyzed using Windows performance reports and manual calculation, while the utilization of the server is done using ICCP tool by feeding the load data for users and the system data on the target server.
The recommended threshold of 40% or 0.4 shown on the charts, is applicable to physical servers.
Virtual servers have a recommended threshold of 80% CPU utilization. The graphs which can be produced show the utilization demands per server. The existing demands should be collected from the target servers, then this data can be imported in equivalent servers in the model.
If the transactions are routed through a load-balancer then the utilization cannot be allocated precisely to one server or the other, so the model should be set up with a cluster of servers to allow for this. The working hours need to be recorded for any activity patterns to be simulated. These should be recorded for access hours from Monday to Friday and for weekend working.
The System Dashboard Monitor, described earlier, for use to collect the input data for ICCP, can be downloaded with the product numbers shown in Figure 3.1:
Note: The link to the IBM Software Access Catalog, https://www-356.ibm.com/partnerworld/partnertools/eorderweb/ordersw.do, was used for the System Dashboard Monitor software.
Figure 3.1: The System Monitor is downloaded (Part Nos. M07PTML and M07PWML)