Starman - High-performance preforking PSGI/Plack web server
# Run app.psgi with the default settings
> starman
# run with Server::Starter
> start_server --port 127.0.0.1:80 -- 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
Preforking
Spawns workers preforked like most high performance UNIX servers do. Starman also reaps dead children and automatically restarts the worker pool.
Signals
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 = ...
$resp;
}
Tatsuhiko Miyagawa miyagawa@bulknews.net
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.