The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


Starman - High-performance preforking PSGI/Plack web server


  # Run app.psgi with the default settings
  > starman

  # run with Server::Starter
  > start_server --port -- starman --workers 32 myapp.psgi

  # UNIX domain sockets
  > starman --listen /tmp/starman.sock

Read more options and configurations by running `perldoc starman` (lower-case s).


Starman is a PSGI perl web server that has unique features such as:

High Performance

Uses the fast XS/C HTTP header parser


Spawns workers preforked like most high performance UNIX servers do. Starman also reaps dead children and automatically restarts the worker pool.


Supports HUP for graceful worker restarts, and TTIN/TTOU to dynamically increase or decrease the number of worker processes, as well as QUIT to gracefully shutdown the worker processes.

Superdaemon aware

Supports Server::Starter for hot deploy and graceful restarts.

Multiple interfaces and UNIX Domain Socket support

Able to listen on multiple interfaces including UNIX sockets.

Small memory footprint

Preloading the applications with --preload-app command line option enables copy-on-write friendly memory management. Also, the minimum memory usage Starman requires for the master process is 7MB and children (workers) is less than 3.0MB.

PSGI compatible

Can run any PSGI applications and frameworks

HTTP/1.1 support

Supports chunked requests and responses, keep-alive and pipeline requests.

UNIX only

This server does not support Win32.


Here's a simple benchmark using Hello.psgi.

  -- server: Starman (workers=10)
  Requests per second:    6849.16 [#/sec] (mean)
  -- server: Twiggy
  Requests per second:    3911.78 [#/sec] (mean)
  -- server: AnyEvent::HTTPD
  Requests per second:    2738.49 [#/sec] (mean)
  -- server: HTTP::Server::PSGI
  Requests per second:    2218.16 [#/sec] (mean)
  -- server: HTTP::Server::PSGI (workers=10)
  Requests per second:    2792.99 [#/sec] (mean)
  -- server: HTTP::Server::Simple
  Requests per second:    1435.50 [#/sec] (mean)
  -- server: Corona
  Requests per second:    2332.00 [#/sec] (mean)
  -- server: POE
  Requests per second:    503.59 [#/sec] (mean)

This benchmark was processed with ab -c 10 -t 1 -k on MacBook Pro 13" late 2009 model on Mac OS X 10.6.2 with perl 5.10.0. YMMV.


Because Starman runs as a preforking model, it is not recommended to serve the requests directly from the internet, especially when slow requesting clients are taken into consideration. It is suggested to put Starman workers behind the frontend servers such as nginx, and use HTTP proxy with TCP or UNIX sockets.



Starman exposes a callback named psgix.informational that can be used for sending an informational response. The callback accepts two arguments, the first argument being the status code and the second being an arrayref of the headers to be sent. Example below sends an 103 Early Hints response before processing the request to build a final response.

    sub {
        my $env = shift;

        $env->{'psgix.informational'}->( 103, [
            "Link" => "</style.css>; rel=preload"
        ] );

        my $rest = ...


Tatsuhiko Miyagawa <>

Andy Grundman wrote Catalyst::Engine::HTTP::Prefork, which this module is heavily based on.

Kazuho Oku wrote Net::Server::SS::PreFork that makes it easy to add Server::Starter support to this software.

The psgix.informational callback comes from Starlet by Kazuho Oku.


Tatsuhiko Miyagawa, 2010-


This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.


Plack Catalyst::Engine::HTTP::Prefork Net::Server::PreFork