Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
263 lines (155 sloc) 17.9 KB
title author description ms.author ms.custom ms.date uid
Web server implementations in ASP.NET Core
guardrex
Discover the web servers Kestrel and HTTP.sys for ASP.NET Core. Learn how to choose a server and when to use a reverse proxy server.
tdykstra
mvc
01/11/2019
fundamentals/servers/index

Web server implementations in ASP.NET Core

By Tom Dykstra, Steve Smith, Stephen Halter, and Chris Ross

An ASP.NET Core app runs with an in-process HTTP server implementation. The server implementation listens for HTTP requests and surfaces them to the app as a set of request features composed into an xref:Microsoft.AspNetCore.Http.HttpContext.

::: moniker range=">= aspnetcore-2.2"

Windows

ASP.NET Core ships with the following:

When using IIS or IIS Express, the app either runs:

The ASP.NET Core Module is a native IIS module that handles native IIS requests between IIS and the in-process IIS HTTP Server or Kestrel. For more information, see xref:host-and-deploy/aspnet-core-module.

Hosting models

In-process hosting model

Using in-process hosting, an ASP.NET Core app runs in the same process as its IIS worker process. This removes the out-of-process performance penalty of proxying requests over the loopback adapter, a network interface that returns outgoing network traffic back to the same machine. IIS handles process management with the Windows Process Activation Service (WAS).

The ASP.NET Core Module:

  • Performs app initialization.
    • Loads the CoreCLR.
    • Calls Program.Main.
  • Handles the lifetime of the IIS native request.

The in-process hosting model isn't supported for ASP.NET Core apps that target the .NET Framework.

The following diagram illustrates the relationship between IIS, the ASP.NET Core Module, and an app hosted in-process:

ASP.NET Core Module

A request arrives from the web to the kernel-mode HTTP.sys driver. The driver routes the native request to IIS on the website's configured port, usually 80 (HTTP) or 443 (HTTPS). The module receives the native request and passes it to IIS HTTP Server (IISHttpServer). IIS HTTP Server is an in-process server implementation for IIS that converts the request from native to managed.

After the IIS HTTP Server processes the request, the request is pushed into the ASP.NET Core middleware pipeline. The middleware pipeline handles the request and passes it on as an HttpContext instance to the app's logic. The app's response is passed back to IIS, which pushes it back out to the client that initiated the request.

In-process hosting is opt-in for existing apps, but dotnet new templates default to the in-process hosting model for all IIS and IIS Express scenarios.

Out-of-process hosting model

Because ASP.NET Core apps run in a process separate from the IIS worker process, the module handles process management. The module starts the process for the ASP.NET Core app when the first request arrives and restarts the app if it shuts down or crashes. This is essentially the same behavior as seen with apps that run in-process that are managed by the Windows Process Activation Service (WAS).

The following diagram illustrates the relationship between IIS, the ASP.NET Core Module, and an app hosted out-of-process:

ASP.NET Core Module

Requests arrive from the web to the kernel-mode HTTP.sys driver. The driver routes the requests to IIS on the website's configured port, usually 80 (HTTP) or 443 (HTTPS). The module forwards the requests to Kestrel on a random port for the app, which isn't port 80 or 443.

The module specifies the port via an environment variable at startup, and the IIS Integration Middleware configures the server to listen on http://localhost:{PORT}. Additional checks are performed, and requests that don't originate from the module are rejected. The module doesn't support HTTPS forwarding, so requests are forwarded over HTTP even if received by IIS over HTTPS.

After Kestrel picks up the request from the module, the request is pushed into the ASP.NET Core middleware pipeline. The middleware pipeline handles the request and passes it on as an HttpContext instance to the app's logic. Middleware added by IIS Integration updates the scheme, remote IP, and pathbase to account for forwarding the request to Kestrel. The app's response is passed back to IIS, which pushes it back out to the HTTP client that initiated the request.

For IIS and ASP.NET Core Module configuration guidance, see the following topics:

  • xref:host-and-deploy/iis/index
  • xref:host-and-deploy/aspnet-core-module

macOS

ASP.NET Core ships with Kestrel server, which is the default, cross-platform HTTP server.

Linux

ASP.NET Core ships with Kestrel server, which is the default, cross-platform HTTP server.


::: moniker-end

::: moniker range="< aspnetcore-2.2"

Windows

ASP.NET Core ships with the following:

When using IIS or IIS Express, the app runs in a process separate from the IIS worker process (out-of-process) with the Kestrel server.

Because ASP.NET Core apps run in a process separate from the IIS worker process, the module handles process management. The module starts the process for the ASP.NET Core app when the first request arrives and restarts the app if it shuts down or crashes. This is essentially the same behavior as seen with apps that run in-process that are managed by the Windows Process Activation Service (WAS).

The following diagram illustrates the relationship between IIS, the ASP.NET Core Module, and an app hosted out-of-process:

ASP.NET Core Module

Requests arrive from the web to the kernel-mode HTTP.sys driver. The driver routes the requests to IIS on the website's configured port, usually 80 (HTTP) or 443 (HTTPS). The module forwards the requests to Kestrel on a random port for the app, which isn't port 80 or 443.

The module specifies the port via an environment variable at startup, and the IIS Integration Middleware configures the server to listen on http://localhost:{port}. Additional checks are performed, and requests that don't originate from the module are rejected. The module doesn't support HTTPS forwarding, so requests are forwarded over HTTP even if received by IIS over HTTPS.

After Kestrel picks up the request from the module, the request is pushed into the ASP.NET Core middleware pipeline. The middleware pipeline handles the request and passes it on as an HttpContext instance to the app's logic. Middleware added by IIS Integration updates the scheme, remote IP, and pathbase to account for forwarding the request to Kestrel. The app's response is passed back to IIS, which pushes it back out to the HTTP client that initiated the request.

For IIS and ASP.NET Core Module configuration guidance, see the following topics:

  • xref:host-and-deploy/iis/index
  • xref:host-and-deploy/aspnet-core-module

macOS

ASP.NET Core ships with Kestrel server, which is the default, cross-platform HTTP server.

Linux

ASP.NET Core ships with Kestrel server, which is the default, cross-platform HTTP server.


::: moniker-end

Kestrel

Kestrel is the default web server included in ASP.NET Core project templates.

::: moniker range=">= aspnetcore-2.0"

Kestrel can be used:

  • By itself as an edge server processing requests directly from a network, including the Internet.

    Kestrel communicates directly with the Internet without a reverse proxy server

  • With a reverse proxy server, such as Internet Information Services (IIS), Nginx, or Apache. A reverse proxy server receives HTTP requests from the Internet and forwards them to Kestrel.

    Kestrel communicates indirectly with the Internet through a reverse proxy server, such as IIS, Nginx, or Apache

Either hosting configuration—with or without a reverse proxy server—is supported for ASP.NET Core 2.1 or later apps.

::: moniker-end

::: moniker range="< aspnetcore-2.0"

If the app only accepts requests from an internal network, Kestrel can be used by itself.

Kestrel communicates directly with the internal network

If the app is exposed to the Internet, Kestrel must use a reverse proxy server, such as Internet Information Services (IIS), Nginx, or Apache. A reverse proxy server receives HTTP requests from the Internet and forwards them to Kestrel.

Kestrel communicates indirectly with the Internet through a reverse proxy server, such as IIS, Nginx, or Apache

The most important reason for using a reverse proxy for public-facing edge server deployments that are exposed directly the Internet is security. The 1.x versions of Kestrel don't include important security features to defend against attacks from the Internet. This includes, but isn't limited to, appropriate timeouts, request size limits, and concurrent connection limits.

::: moniker-end

For Kestrel configuration guidance and information on when to use Kestrel in a reverse proxy configuration, see xref:fundamentals/servers/kestrel.

Nginx with Kestrel

For information on how to use Nginx on Linux as a reverse proxy server for Kestrel, see xref:host-and-deploy/linux-nginx.

Apache with Kestrel

For information on how to use Apache on Linux as a reverse proxy server for Kestrel, see xref:host-and-deploy/linux-apache.

::: moniker range=">= aspnetcore-2.2"

IIS HTTP Server

IIS HTTP Server is an in-process server for IIS and required for in-process deployments. The ASP.NET Core Module handles native IIS requests between IIS and IIS HTTP Server. For more information, see xref:host-and-deploy/aspnet-core-module.

::: moniker-end

HTTP.sys

If ASP.NET Core apps are run on Windows, HTTP.sys is an alternative to Kestrel. Kestrel is generally recommended for best performance. HTTP.sys can be used in scenarios where the app is exposed to the Internet and required capabilities are supported by HTTP.sys but not Kestrel. For more information, see xref:fundamentals/servers/httpsys.

HTTP.sys communicates directly with the Internet

HTTP.sys can also be used for apps that are only exposed to an internal network.

HTTP.sys communicates directly with the internal network

For HTTP.sys configuration guidance, see xref:fundamentals/servers/httpsys.

ASP.NET Core server infrastructure

The IApplicationBuilder available in the Startup.Configure method exposes the ServerFeatures property of type IFeatureCollection. Kestrel and HTTP.sys only expose a single feature each, IServerAddressesFeature, but different server implementations may expose additional functionality.

IServerAddressesFeature can be used to find out which port the server implementation has bound at runtime.

Custom servers

If the built-in servers don't meet the app's requirements, a custom server implementation can be created. The Open Web Interface for .NET (OWIN) guide demonstrates how to write a Nowin-based IServer implementation. Only the feature interfaces that the app uses require implementation, though at a minimum IHttpRequestFeature and IHttpResponseFeature must be supported.

Server startup

The server is launched when the Integrated Development Environment (IDE) or editor starts the app:

When launching the app from a command prompt in the project's folder, dotnet run launches the app and server (Kestrel and HTTP.sys only). The configuration is specified by the -c|--configuration option, which is set to either Debug (default) or Release. If launch profiles are present in a launchSettings.json file, use the --launch-profile <NAME> option to set the launch profile (for example, Development or Production). For more information, see dotnet run and .NET Core distribution packaging.

HTTP/2 support

HTTP/2 is supported with ASP.NET Core in the following deployment scenarios:

::: moniker range=">= aspnetcore-2.2"

  • Kestrel
    • Operating system
      • Windows Server 2016/Windows 10 or later†
      • Linux with OpenSSL 1.0.2 or later (for example, Ubuntu 16.04 or later)
      • HTTP/2 will be supported on macOS in a future release.
    • Target framework: .NET Core 2.2 or later
  • HTTP.sys
    • Windows Server 2016/Windows 10 or later
    • Target framework: Not applicable to HTTP.sys deployments.
  • IIS (in-process)
    • Windows Server 2016/Windows 10 or later; IIS 10 or later
    • Target framework: .NET Core 2.2 or later
  • IIS (out-of-process)
    • Windows Server 2016/Windows 10 or later; IIS 10 or later
    • Public-facing edge server connections use HTTP/2, but the reverse proxy connection to Kestrel uses HTTP/1.1.
    • Target framework: Not applicable to IIS out-of-process deployments.

†Kestrel has limited support for HTTP/2 on Windows Server 2012 R2 and Windows 8.1. Support is limited because the list of supported TLS cipher suites available on these operating systems is limited. A certificate generated using an Elliptic Curve Digital Signature Algorithm (ECDSA) may be required to secure TLS connections.

::: moniker-end

::: moniker range="< aspnetcore-2.2"

  • HTTP.sys
    • Windows Server 2016/Windows 10 or later
    • Target framework: Not applicable to HTTP.sys deployments.
  • IIS (out-of-process)
    • Windows Server 2016/Windows 10 or later; IIS 10 or later
    • Public-facing edge server connections use HTTP/2, but the reverse proxy connection to Kestrel uses HTTP/1.1.
    • Target framework: Not applicable to IIS out-of-process deployments.

::: moniker-end

An HTTP/2 connection must use Application-Layer Protocol Negotiation (ALPN) and TLS 1.2 or later. For more information, see the topics that pertain to your server deployment scenarios.

Additional resources

  • xref:fundamentals/servers/kestrel
  • xref:host-and-deploy/aspnet-core-module
  • xref:host-and-deploy/iis/index
  • xref:host-and-deploy/azure-apps/index
  • xref:host-and-deploy/linux-nginx
  • xref:host-and-deploy/linux-apache
  • xref:fundamentals/servers/httpsys