Hello, I am Laszlo

Measuring Lock Contention

06/25/2026

Scaling multi-threaded applications is often limited by the proportion of the workload that cannot be parallelized. This concept is captured by Amdahl's law. Code sections guarded by locks are a common example — they restrict scalability by allowing only one thread (or n threads in the case of semaphores) to execute, forcing other threads to wait due to lock contention.

Context

In an HTTP/2 server, requests and responses on a connection are multiplexed. The server needs to write concurrently computed responses to a network stream. Such multiplexing can be achieved in multiple ways:

• Using a Channel in a producer-consumer pattern
• Guarding writes with locks
• Implementing custom scheduling solutions with BlockingCollection or Queues

Find out more »

HTTP/3 with ASP.NET Core and browsers

06/06/2026

Configure ASP.NET Core WebApi

In this post I explore how to test locally an ASP.NET Core WebApi with HTTP/3 in a local environment using a browser. In ASP.NET Core with .NET 10, HTTP/3 is already supported, but it is an opt-in feature.

To enable HTTP/3, add the following configuration to the application, either in the appsettings.json file

"Kestrel": {
  "EndpointDefaults": {
    "Protocols": "Http1AndHttp2AndHttp3"
  }
}

Find out more »

Know Your Data

05/24/2026

There are many social posts, blog posts, and code suggestions online for .NET code snippets titled "Use X instead of Y" or "Why X is better than Y". These posts usually (but not always) include a basic performance comparison using BenchmarkDotNet to validate their catchy titles.

At first glance, these suggestions appear justified by the measurements. However, upon closer inspection, many important details are typically excluded:

• What version of .NET (SDK and Runtime) was used?
• What version of OS was used?
• What is the underlying hardware architecture (ARM or x64)?
• Does the hardware support vectorized operations?
• What are the sizes of the vector registers?
• What is the memory read latency?
• What are the CPU cache sizes?
• What branch prediction algorithm does the CPU use?
• What are the input data types (structs, classes, primitives)?
• What is the input data structure?
• What is the size of the input data (bytes/array length)?
• What is the data access pattern?
• And many more factors

Since explaining all these factors is beyond this post's scope, I will focus on data access patterns and data sizes. For those interested in a deeper analysis, I recommend reading Pro .NET Benchmarking by Andrey Akinshin, which explains many common pitfalls.

Find out more »

Request Cancellation in ASP.NET Core with HTTP/2

05/01/2026

ASP.NET Core allows to create endpoints that receive a cancellation token, such as below:

app.MapGet("/path", async (CancellationToken token) =>
{
});

In this post, I will explore how a server (such as Kestrel) may fire the token to be canceled.

Cooperative Cancellation

Find out more »

Distinct and HashSet

04/21/2026

A common pattern in business-as-usual (BAU) applications is that a class in the business logic layer requests a set of entities by their integer IDs from a repository class. Here is one possible implementation of this pattern:

1. A method in the business layer receives a list of integers as an argument (List<int>).
2. It applies filtering and selection on these integers by creating a new enumerable.
3. It extracts the unique values from the filtered and projected numbers.
4. It passes the filtered, selected unique numbers to a query method in the repository class.
5. This repository method processes the passed in integer enumerable:
   • If there aren't any numbers in the enumeration, it returns null (or empty results).
   • Otherwise, it copies the numbers into an SQL query and returns the query's result.

In this blog post, I will focus on two possible implementations to produce the unique set of integers:

1. Using the Distinct LINQ extension method.
2. Using the ToHashSet<T> method to materialize the result before passing it to the repository class.

Find out more »