ETWAnalyzer 2.3.3-Preview

Programatic Access to Extracted Data

The IETWExtract interface gives you access to all deserialized data of an extracted Json file in a cohesive object oriented object model. Since ETWAnalyzer was built initially for mass data analysis of automated tests you need to start with a TestRunData instance which gets as input a directory or a Json file. From there you can then query the TestRuns or all contained files. Each TestDataFile has if it adheres to the naming convention the

• TestCase Name
• Measured Duration
• Time when it was taken
• Machine Name
• Client/Server
• Test Status

This is all encoded into the file name which is very fast to query without the need to open any json file. When you want to examine the contents of a Json file you can access of a TestDataFile its Extract property which will deserialize the Json file.

using ETWAnalyzer.Extract;
using ETWAnalyzer.ProcessTools;
using System.Collections.Generic;
using System.Linq;

namespace AutoQuery
{
    internal class Program
    {
        static void Main(string[] args)
        {
            TestRunData data = new TestRunData(args[0]); // read file or a directory with tests
            foreach(TestDataFile test in data.AllFiles)  
            {
                IETWExtract extract = test.Extract; // read Json data
                foreach(KeyValuePair<ProcessKey,uint> topn in extract.CPU.PerProcessCPUConsumptionInMs.Take(10))
                {
                    ColorConsole.WriteEmbeddedColorLine($"[green]{topn.Value,-6} ms[/green] [yellow]{topn.Key,-50}[/yellow]");
                }

            }
        }
    }
}

This very simple application gives you already the top 10 CPU consumers of a file or a directory of files:

Below is the class Diagram of IETWExtract

Scalable Json

The main goal of ETWAnalyzer is to make profiling data queryable in a quick manner. On the other hand many additional extractors can be added in the future. If the queried data is small it can be directly added to the main Json extract file. But if the data can be large, such as FILEIO data which records every touched file operation it can become pretty large. To prevent that performance hit a simple way out was chosen. During serialization for potentially large data extracts a different file is used which contains the suffix _Derived_xxxx.json

These derived files are accessed only when you touch e.g. the FileIO data property. That way we can evolve over time many more extractors with different goals without sacrificing query speed for common things.