Server virtualization planning
Another very important aspect of MAP is the Server Virtualization Planning Wizard. Organisations are increasingly virtualising their server infrastructure for reducing costs and increasing ROI. MAP toolkit allows you to analyse your current server setup and find out how many virtual server host computers you would need to virtualise one or more physical machines.
For this to work, MAP will require you to create a text file containing the names of the servers you want to be virtualised. The Performance Metrics Wizard will then need to be run for at least an hour to gather performance data. You will then create a model of your virtual machine host by specifying how many CPUs it has, what disk configuration is present and how much memory is available. Once the steps have been completed, you can run the Server Virtualization Planning Wizard and specify what Microsoft virtualisation technology is to be used. The wizard will guide you through the steps and when completed, a detailed analysis will start.
We will not go into the details of creating a hardware configuration library and planning for virtualisation here, but will talk about the Performance Metrics Wizard.
Measuring performance metrics with MAP
Apart from server consolidation, there could be various reasons for running the Performance Metrics Wizard: base-lining being one example. You may want to capture a snapshot of the memory, disk, CPU and network utilisation for your servers over a period of time.
MAP allows you to run performance Monitor (PerfMon) counters on one or more servers simultaneously. That way, you do not have to go to each individual server to create the counter logs. But you also lose some flexibility in such scenarios: MAP only exposes a predefined set of counters. This means unlike creating your own custom performance counter log, you are restricted to what is available. Fear not, the counters available from MAP are useful enough and sufficient to get you started.
When you click on the Server Consolidation node of the MAP toolkit, the detail pane lists a number of graphical options as shown below:
Clicking on the Performance Metrics Wizard link will result in the following dialogue box appearing:
As you can see, the performance metric wizard does not allow you to get computer names from AD. Instead, it expects you to have a plain ASCII text file with the name of each server listed in a new line.
In our case, I am interested to capture the performance workload of two of my servers: an application server and a database server. I have created a file and I chose that.
Once you select the text file, a pop up message shows how many servers will be inventoried for performance data. The next screen lets you specify the user account that will be used to connect to each machine via WMI
The next screen allows you to specify the performance counter end time. By default, it is 1 hour from your present time. I have set it to run for 2 hours:
You can also see that MAP does not allow you to choose how often the performance data will be collected. MAP collects this data from each target machine every five minutes.
The final screen lets you finish the wizard and the counter log process will start:
Once the performance data has been collected, you can click on the Performance Metrics Results item under the Server Consolidation node. This time the details pane would show the results in a tabbed format. To get a better look, click on the Generate Report link from the actions pane.
Once the report generation has completed, you can click on the Saved Reports and Proposals option on the View menu. This will open the same explorer window where the discovery reports are saved (by default, it is in a folder under your user profile’s My Documents directory). The Excel spreadsheet you would look for is called PerfMetricResults with a date time stamp.
The Excel spreadsheet has several tabs, each one showing a different list of metrics. These are:
- Processor Utlization (%Processor Time plus other values)
- Memory Utilization (page faults per second, available bytes etc.)
- Network Utilization (Bytes received and sent per second etc).
- Physical and Logical Disk Utilization (Average Disk Queue Length and other counters)
The values contained in the spreadsheet do not show the maximums, minimums and averages over the period of time. All it gives you is a single line of result for each machine. For CPU and memory, the values of the counters are averages. For network and disks, the values reflect maximums.
Like the discovery and readiness reports, the performance metric report also shows entries for machines that could not be reached for performance analysis.
One thing that can worry system administrators is the amount of data travelling across the network as MAP collects performance metrics from remote servers and saves it locally. The performance overhead is about 1 MB of data per machine every five minutes. If you want to monitor a large number of servers, Microsoft suggests you group them in batches of up to 150 machines in each and run the performance metrics on each batch separately.
Now if you run the Performance Metrics Wizard again, a message like the following will appear:
If you click No, the performance data for the new machine(s) will be appended to the existing performance data of other machines. However, if your list contains the same server, the previously collected data for the machine will be overwritten.