Extracting Performance Data from Your .NET Applications

Extracting Performance Data from Your .NET Applications

ou always want the software you write to have great performance. The reason isn’t shocking?users look to software to quickly and efficiently handle their workload. Often times, meeting this performance requirement (whether it is explicit or implied), can be a difficult, even daunting task. Tuning an application to perform at its peak level involves a thorough understanding of the architecture and environment into which you will deploy your application. However, you can’t truly begin to optimize an application’s performance if you don’t understand how to empirically measure that performance. From this perspective, your application must emit enough data to enable real time performance monitoring.

Imagine for a moment that a new client has commissioned you to develop a .NET component. Your client gives you fairly simple specifications for the component. Your application must process comma-separated data contained within flat-files, and insert this content into a SQL Server database. The application itself does not even need to expose a graphical user interface?a simple console application will suffice. Beyond these very basic instructions, the client also imparts this golden nugget of information: Your component must process over 1,000 transactions per second.

How would you go about proactively publishing enough empirical data from your component so that, at the end of the day, you can accurately address the question of application performance? The answer?use custom performance counters.

Figure 1: The Windows Performance Monitor application allows you to select and view the current data associated with any registered performance counters.

What Are Performance Counters?
Performance counters are discrete sets of data published by an executing application, service, or operating system process. Windows 2000, Windows XP, and Windows Server 2003 intrinsically support counters. They exist for the express purpose of diagnosing and analyzing throughput, bottlenecks, and so on.

Performance data is organized according to objects, counters, and instances. Objects represent the application or process that publishes the performance data. A counter is the actual measurement point. You use an instance to further delineate processes. Instances are optional.

Data is exposed for inspection via the Windows Performance Monitor console (see Figure 1). This interface allows you to select specific counter instances to view graphically or log to a file.

You can see that all of the plumbing is in place to monitor and view performance counters. Now let’s discuss how to publish performance data from within your application.

Figure 2: The FileWatcher class waits for data files to be dropped into a file system directory. It then passes control to the Loader class to insert the records from the file into the Northwind database.

Creating Performance Counters
Your solution could take many different forms depending on the actual attributes of the system you want to develop. To set the stage a bit, consider a console application tasked with inserting rows into the Products table in the Northwind database. The data will arrive in the form of a CSV file that gets dropped into a pre-determined directory. The console application merely has to read each line in the file and copy its contents over verbatim to the Products table?as fast as it possibly can.

Figure 2 and Figure 3 show one possible design for this application. The FileWatcher class watches a specific file system directory. When a file arrives there, FileWatcher passes the filename to the Loader class, which will establish the database connection, read the file in line by line, and insert the data into the Northwind database.

To add performance data to this application, you must first identify the data that will enable useful performance analysis. Potential statistics useful for performance analysis might include:

  1. The number of records inserted per second. (This will provide a good indicator of raw processing speed.)
  2. The number of errors per second. (This will provide a slightly different view on the processing speed dimension.)
  3. A count of the total records inserted. (This will provide a view of the workload gradient over time.)
Figure 3: The basic design for the Northwinds DataLoader component involves a three-step execution. (1) Files containing the product data are dropped into a directory. (2) The DataLoader component waits for files to arrive in this directory. (3) New files are processed line by line, and their contents are inserted into the Northwind database.

Other possibilities include: percentage of errors to successful inserts, total execution time per file, average records per file, and so on. The three statistics listed above provides a good enough basis to use for exploring performance counters.

With the different counters identified, let’s create them. You can add performance counters to the current inventory in one of three ways. You can add them programmatically at run-time, you can add them via an install action as part of a setup package, or you can add them via the Visual Studio .NET Server Explorer. In this article, we’ll concentrate on creating counters using Server Explorer, and we’ll also take a look at what it takes to create counters at run-time. Note that the Standard Edition of Visual Studio .NET does not ship with Server Explorer.

Creating Performance Counters with Server Explorer
To create your set of counters using Server Explorer, fire up a version of Visual Studio .NET that ships with Server Explorer (Enterprise Architect, Enterprise Developer, or Professional). Make Server Explorer visible. (Press Ctrl-Alt-S to make it active if it isn’t currently being displayed.) Under the Servers node, locate and expand the tree for the computer that will host the performance counters. Expand the Performance Counters node. Figure 4 shows the performance counters configured on a machine.

Figure 4: The Visual Studio .NET Server Explorer will display the list of all performance counters registered on a particular machine. Server Explorer makes it easy to add custom performance counters to the current list of counters.

To add your own performance counter, you will first create a new category. A category is typically the name of the application that will publish the performance data, although you may wish to use multiple categories for larger systems. Right-click on the Performance Counters node and select Create New Category. This will launch the Performance Counter Builder dialog box shown in Figure 5.

To follow along with our Northwind data loader sample, type Northwinds Data Loader as the Category name then click New to add a new counter. Working through the list of previously identified counters, enter the name, type, and description for each one, filling in the information that you see in Table 1. Note that there is a slight disconnect in the nomenclature here. A category is the same thing as an “object.” (Refer above where we discussed the organization of performance counters.)

Table 1: For the Northwinds DataLoader class, we are interested in tracking statistics across three different counters.



Load Errors/sec


Records Loaded/sec


Total Records Loaded


Click OK?and you’re done. You’ve created the performance counters and registered them on your machine.

Figure 5: The Performance Counter Builder will allow you to create new performance counters or edit existing ones.

Creating Performance Counters Programmatically
To create your performance counters in code, you will use a series of classes from the System.Diagnostics namespace. Let’s start off with the CounterCreationDataCollection class. This collection class holds CounterCreationData instances; these physically represent the custom counters. Therefore, after instantiating a CounterCreateDataCollection object, you will then create your counters by building CounterCreateData objects and adding them to the CounterCreationDataCollection instance. The following snippet shows the code necessary to create the records loaded/second counter.

   // C#   // First create the CounterCreationDataCollection    // object   counters = new CounterCreationDataCollection();      // Create the counter instances to add to the   // collection...   CounterCreationData ctr =       new CounterCreationData();      // Set the counter's properties   ctr.CounterName = "Records Loaded/sec";   ctr.CounterType =      PerformanceCounterType.RateOfCountsPerSecond32;      // Add the counter to the collection   counters.Add(ctr);      

After you have built up your collection of counters, you must create them in a particular category using the PerformanceCounterCategory class. This class exposes a static/shared member called Create. After passing the category name, category help topic, and the CounterCollectionData instance in to the Create method, the counter(s) will be registered on the local computer. In addition to creating the counter, the Create method will also create the category if it doesn’t currently exist on the target machine.

   // Create the counter   PerformanceCounterCategory.Create      ("Northwind Data Loader",      "", counters);

Publishing Performance Data
The counters you created can now accept data. Let’s add the publishing code to the data loader. Figure 3 shows the basic architecture of the Northwinds data loader component. We will place all of the important activity, from a performance perspective, inside the Loader class. Through its ProcessFile method, the Loader class opens the data file and sends each line through to a SqlCommand object (via the private InsertRow routine) that performs the insert into the Products table in the Northwind database. In the following pseudo-code for the record insertion process, it becomes clearer exactly where you should publish the performance counter data.

   public void ProcessFile()   {   // Open the file   // Loop: read in each line    //    Parse the file line   //    Try: Insert the values   //         Update perf counters   //         Continue loop   //    Catch:   //         Update per counters   //         Continue loop   // End loop   // Close the file   }

There are really two important points within the source: when a record is successfully inserted, and when an exception is encountered during a record insert.

With the code sites located, you can now focus on publishing the performance data. Writing to performance counters involves using the PerformanceCounter class from the System.Diagnostics namespace. The PerformanceCounter class represents a registered instance of a performance counter. Performance counters are uniquely identified by name; the x class allows you to identify the counter instance you want to work with through its CounterName property. You will also identify the counter through its category name (via the CategoryName property) and through the machine name that is hosting the counter (via the MachineName property). This code will create a ‘connection’ to our previously created records loaded/sec counter.

   // Configure the loadsPerSec counter   loadsPerSec = new PerformanceCounter();      // Set the category name   loadsPerSec.CategoryName = "Northwind Loader";      // Set the counter name   loadsPerSec.CounterName = "Records Loaded/sec";      // Set the machine name   loadsPerSec.MachineName = ".";

You will also set the ReadOnly property (to indicate that you want to write into the counter), and the RawValue property. The code uses the RawValue property to directly assign a value to the counter. In the code snippet below, I use RawValue to initialize the counter to 0.

   loadsPerSec.ReadOnly = false;   loadsPerSec.RawValue = 0;

You can create handles to the other counters in a similar fashion. Once you have the handle to a counter, to record a newly loaded record you just need to call the PerformanceCounter.Increment() method. This method will increment the counter’s raw value by 1. Although the records loaded/sec counter represents an average value over time, you don’t need to write any other code. The counter handles averaging and other calculations internally.

Figure 6: This graph shows the DataLoader application in action. At its peak, it processed just short of 1,100 records per second.

Listing 1 shows the completed Loader class. To simplify the code a bit, I created a CounterHelper class (see Listing 2). I use CounterHelper to create the PerformanceCounter objects and expose three public methods that the Loader class uses to increment those counters: IncLoadsPerSec, IncErrorsPerSec, and IncTotalLoaded.

To see the performance counters in action, run the Windows Performance Monitor and then run the data loader application. You begin the load process by dropping a test file into the loader’s data directory. Figure 6 shows the performance counters during the load of a file with approximately 31,000 records. Some records contained errors (as shown by the errors/sec counter).

By leveraging performance counters in your applications, you can generate an accurate view of your application’s performance levels, whether during the construction phase or after deployment into production. Accompanying source code for this article, in both Visual Basic and C#, is available at http://www.brilliantstorm.com/.

Share the Post:
XDR solutions

The Benefits of Using XDR Solutions

Cybercriminals constantly adapt their strategies, developing newer, more powerful, and intelligent ways to attack your network. Since security professionals must innovate as well, more conventional endpoint detection solutions have evolved

AI is revolutionizing fraud detection

How AI is Revolutionizing Fraud Detection

Artificial intelligence – commonly known as AI – means a form of technology with multiple uses. As a result, it has become extremely valuable to a number of businesses across

AI innovation

Companies Leading AI Innovation in 2023

Artificial intelligence (AI) has been transforming industries and revolutionizing business operations. AI’s potential to enhance efficiency and productivity has become crucial to many businesses. As we move into 2023, several

data fivetran pricing

Fivetran Pricing Explained

One of the biggest trends of the 21st century is the massive surge in analytics. Analytics is the process of utilizing data to drive future decision-making. With so much of

kubernetes logging

Kubernetes Logging: What You Need to Know

Kubernetes from Google is one of the most popular open-source and free container management solutions made to make managing and deploying applications easier. It has a solid architecture that makes

ransomware cyber attack

Why Is Ransomware Such a Major Threat?

One of the most significant cyber threats faced by modern organizations is a ransomware attack. Ransomware attacks have grown in both sophistication and frequency over the past few years, forcing

data dictionary

Tools You Need to Make a Data Dictionary

Data dictionaries are crucial for organizations of all sizes that deal with large amounts of data. they are centralized repositories of all the data in organizations, including metadata such as