Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Daemon for zero-configuration management of a heterogeneous cluster of backend web servers and the accompanying libraries.
C++ Ruby Shell
Branch: master
Fetching latest commit...
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
CMakeLists.txt Add debug info to the build.


Cloudbridge is a server that sits between your front-end webservers (apache/nginx/etc) and
your application servers (currently rails using mongrel). It has the advantage over traditional
appserver clustering techniques of limiting the amount of configuration required while also improving
worst case behaviour.

In the traditional setup, you have a frontend server doing reverse proxy requests to backend app servers.
This setup is a huge win over directly exposing your backend servers because it allows for load
balancing, optimized resource usage (heavy on memory backends aren't sitting idle waiting for
a client to close their connection), and good scalability (since you can add backend nodes with no 
external configuration changes).

However, this configuration has the disadvantage of causing an increase in maintenance effort
in that when a backend is disabled or goes down unexpectedly, the frontend server must either
be told about this or discover it. If it does not, this usually results in hung connections
or broken requests for the user at worst, or extremely more complex frontend servers at best.

Cloudbridge works to avoid this situation by reversing the connection to the backend. It is
a reverse-reverse proxy. In cloudbridge's system, the backends and frontends both connect
to cloudbridge, and it matches connections based on hostname. When a backend goes down or
comes up, it simply stops or starts sending requests to cloudbridge, and no additional
configuration is needed.

Cloudbridge is not the only solution to employ this technique. At the very least, there is
also swiftiply and probably other more proprietary solutions. Where cloudbridge goes one
step beyond swiftiply is in the fact that it requires no configuration for backend hosts.
By using an HTTP protocol extension (the BRIDGE request documented in doc/protocol.txt),
it allows the backend to specify what host it is listening for. This makes setting up
a hosting cluster extremely easy. 

In order to prevent unauthorized access to a domain, the cloudbridge server may be configured
to require a key to be presented to the server in the BRIDGE request that authenticates its
ownership over that host name.

This allows also for the open-cloudbridge technique, which is demonstrated on You
can use to experiment with the system. Instructions are presented there.
Something went wrong with that request. Please try again.