Our servers are Intel Pentium III class, running at the clock speed of 733 MHz, with 512MB of SDRAM and an 18 GB SCSI hard drive.

The Network Operations Center (NOC) where our servers reside is located in Baltimore, Maryland, and is OnNet with Frontier GlobalCenter (FGC) and Qwest Communications through two separate bandwidth-on-demand connections which enter Downtown Baltimore just a few floors below the NOC.

FGC, a Tier 1 provider whose 13,000-mile fiber optic network and Dense Wave Division Multiplexing (DWDM) technology provide an enormous 460 gigabytes per second (Gbps) of capacity worldwide, has an ATM fiber node located just a few floors below the NOC.

Qwest comes into Baltimore with an OC-48 line. They also have an ATM fiber node 14 floors below the NOC. Qwest comes into Baltimore at the same location with an OC-48 line. Our Qwest connection enables us to offer additional redundancy and better routes to Europe, Latin America, and Asia. With these two carriers, our router will have up to 150,000 possible routes to send each packet of traffic.

Furthermore, because of these unique connections, we no longer need to link to the Internet though an OC3 or T3 Telecom circuit. Instead, independent cables run inside the building directly from the NOC to both the Global Center point of presence and the Qwest point of presence. These lines can handle the bandwidth of a T3 or an OC3, and with DWDM they can handle several times the bandwidth of an OC3.

Network Redundancy
We use intelligent end-user routing software called Border Gateway Protocol (BGP), between Qwest and FGC, who use it as well. BGP can identify which path is the most efficient for each data packet, and then route the packet to its destination on the fastest path. This increases the speed at which web pages sent from our NOC arrive at their destination.

Studies have shown that the second most common reason for downtime is circuit failure on the Tier 1 backbone, the major data highway. To guard against this potential problem, we have two Tier-1 providers. If one has problems, we can route traffic down the other one. Furthermore, because we are OnNet with Frontier GlobalCenter and Qwest, we share their digital distribution architecture, which includes private peering network connections to major Internet carriers such as MCI, Sprint, UUNET, EUNET, AT&T, AOL, Best, Erols, @Home, IBM Advantis and others. These private peering arrangements allow us to exchange packets of data with every major backbone carrier in a one-to-one environment quickly and efficiently.

In addition, FGC has high-speed links to 8 public exchanges including both MAE East and West and several NAPS. Through these many public exchanges, customers have the ability to reach their site wherever they are coming from on the Internet. Thus we have the best of both worlds: a network that is both efficient and wide reaching.

Network Reliability
Industry analysis reveals that 70 percent of downtime over 10 hours with any ISP is caused by telephone circuit failure. Since our NOC is in the same building as Frontier Global Center, circuit failure is virtually eliminated because there is no phone circuit between us and FGC. Instead, there is a direct connection between our Cisco 7200 router and theirs.

The second most common reason for downtime is circuit failure on the Tier 1 backbone. FGC, themselves a backbone, also have peer connections with other major Tier 1 providers, which allows traffic to be switched to other backbones quickly in the event of a crisis.

How reliable is this? Yahoo is another fine company who connects directly using only FGC. If you can reach Yahoo, you can reach our network. FGC's groundbreaking 460 Gbps network runs BGP to 25 other major carriers through private peering arrangements, providing the fastest, most efficient and most reliable network available today.

Raw Performance equals Low Latency/High Throughput
You can't drive an engine as hard as it will go and expect consistent high performance. But often providers operate their networks at three to four times responsible capacity, and as a result their corresponding transfer times reach over 300ms. Our main network daily average is 6.5 percent of its capacity, with midday peak spikes reaching only 15.5 percent capacity.