High Availability will keep your hotspot going and
your users happy. Public IP's authentication and control system consists
of six servers. A master server, three authentication nodes, and
two network database servers running four db nodes. The master
server is used for configuration, updates, and development. The auth
nodes are three synchronized slave servers. Updates that are made
in the master server are pushed/replicated to the three authentication
nodes (aka cerberus). The three nodes combined with the gateway programs
provide high availability (failover) for your hotspot users and load
balancing for the public ip system. The network database cluster
is hosted on two servers and stores user and system data that comes
into the public ip system from the gateways. User records, accounting
records, and gateway updates are all stored in a data cluster for
high availability. If one data node goes down the data will still
be available from the other node.
The datacenter housing Public IP's servers is located in McLean,
Virginia, just across the bridge from
Washington DC. The datacenter has it's own generator, and
have backup battery systems capable of running our systems indefinitely The
datacenter network consists of a Juniper M20-based Routing Layer, a Cisco Catalyst
6500/11000 series Distribution Layer and a Cisco Catalyst Customer Switching
layer. All switch ports are managed and monitored 24/7. The network is connected
to several locations including Equinix and Switch & Data where
it has transit connections. Major upstream bandwidth providers include Deutsche
Telekom's T-Systems and PCCW's BTN networks, with a total of over 500 peers.
Network capacity is variable. Fiber connections are always maintained in far
excess of what is needed, and access to plentiful dark fiber in the datacenter
is available as well. Therefore providing connection details is inaccurate, because
we always maintain what we need and more.
|
