Anatomy of a Walrus
[ Back to Top ]
This page details the organization of a Walrus setup, terminology, etc. This is the blueprint so that engineering can begin on the actual framework and the rest of the pieces that make up this system.
Bird's Eye View
There should be a good name for the entire Walrus infrastructure as deployed for some user. For now, we are going to call this the "installation." With that defined ... a Walrus installation basically consists of various clouds (logically or physically separated networks) containing systems.
As an example we are going to use the Mozilla architecture as it seems like it will be fairly representative of server environments spread across multiple data centers/geographies/etc.
Mozilla has three data centers at present. San Jose, Amsterdam, and China. These data centers are independent, machines in one cannot necessarily talk to machines in another, but they do have the ability to converse through machines that are located on the edges. Walrus will operate in such an environment through defining each data center as a separate "cloud" and indicating how one cloud can talk to another.
Most actions for this kind of environment will happen within a particular cloud. Monitoring, configuration, etc will all be localized. It is assumed that with such setups, there will be a single cloud that is the "master" cloud and will be responsible for such things as sending notifications, issuing configuration commands, and other tasks that typically fall to the administrator.
That's it for the bird's eye view: the installation contains clouds which contain systems. Systems can talk to systems within their own cloud, communication with other clouds can only be done through defined proxies, one cloud is the master cloud, each cloud has a master system, ...