• Fully-redundant architecture
• Multi-gigabit backbone
• SONET transit connections for self-healing fault-tolerance
• Multi-layer network design
• Customized routing policies for customers
• BGP best-path routing
• Automatic detection and re-routing around network failures
• Open peering policy allowing for local data delivery
• Carrier fault-tolerance
• Non-oversubscribed bandwidth
• Quality of Service monitoring and analysis
• Multiple redundant power back-up
• Multiple Levels of Security
• Multiple levels of Fire suppression
• Continuous, uninterruptible power supply
• Environmental Monitoring Systems maintain constant temperature and humidity

The use of non-blocking gigabit devices throughout the network ensures regional latency of a few milliseconds or less, suitable for the most demanding delay-sensitive traffic. Use of redundant fiber rings ensures network reliability and availability.

The data center has connections to many different Internet backbones. By connecting to multiple tier 1 backbones, the data can be distributed through many sources. This architectural design also means that the network connections are not dependent upon an single Internet backbone. Thus when probems occur, traffic rerouting is automatic, thereby ensuring the integrity of the network and continued access for our high-speed servers.

This takes the term "multi-homing" to a whole new level. Presently bandwidth utilization is 5% during peak traffic times. Therefore, the network is very flexible. If one of the backbone connections experiences problems, the traffic can simply be rerouted over other paths, thereby ensuring that users receive fast access times to sites hosted on the network.

In addition, the network runs Border Gate Protocol (BGP4). BGP is used at a provider with more than one access point to the Internet. It helps create a truly redundant network. In fact, in an ideal situation, a lease line failure should result in the BGP routing session to close on the bad leased line and the router on a working circuit should then begin to accept the additional traffic. In other words, traffic from a down circuit is redistributed across other circuits, thereby maintaining network integrity. Providers that are multi-homed and correctly setup can actually be more reliable than a single backbone provider because they have multiple paths to multiple providers.

A provider's local area network is not often enough being seen as a point of latency.

The two main sources of latency for a full-time Internet connection are the user's local area network and the Internet provider's local area network. Ether switches and high-end Cisco routers anchor the local network. This top-of-the-line network hardware ensures that data requests get to their destination and back out of the network as fast as possible.

We use ether switches instead of hubs because of their speed and their security capabilities. Whereas only one computer plugged into a hub can talk at one time, all the machines connected to a switch can talk at the same time.

We use high-end HP Procurve 2524 Switches

Servers on the network do not share a single path (T3). Instead, the servers are connected into a high-speed Ethernet switch. This switch is connected to the core router at the data center.

From the core router, data is sent back to the end user across the fastest available path. Whereas statically routing traffic over one path creates a single point of failure, this distributed architecture ensures that users can access data extremely quickly and have multiple paths both into and out of our network.