Our Network

The network connection is another area where performance is relatively easy to achieve and where reliability presents additional challenges. Our facilities are stellar in both respects.

For performance, a single T1 (1.5 Mbps link) to an internet backbone is more than adequate for the vast majority of internet sites, AS LONG AS the T1 is not saturated. A 90% full T1 will perform far better than a full DS3 (45 Mbps link, which has 30 times the raw capacity of the T1).

All that said, at Axion-IT, we have two redundant DS3 (45 Mbps) connections to two different internet backbones. And they are not even close to being saturated. Currently, the combined peak load handled by these DS3's is about 8 Mbps, so we have 82 Mbps of spare capacity on our trunks to the Internet.

To evaluate reliability, we have to take a closer look at both the trunks to the Internet and the local network joining our servers to these trunks.

All of our servers are housed in the collocation facility run by Atlantech Online, Inc., an Internet Service Provider (ISP) based in Silver Spring, MD. The network and facilities we describe here are theirs. We have worked very closely with Atlantech Online since 1994. Our consultants helped to design, build, and maintain Atlantech's network and the collocation facility (on which our hosting services depend). So, we are very familiar with their systems and technical staff.

Everything about this network has been designed to eliminate any "single point of failure." In other words, everything is redundant, so that if any part of the system fails, traffic is automatically routed around the failure and service continues as normal.

Our servers are connected to two redundant routers through a "multi-layer switched network". The multi-layer switched network uses redundant ISL trunking between the switches. The redundant routers use HSRP (Hot Standby Router Protocol) to allow either router to handle traffic in case one of them fails. Each router has a DS3 (45 Mbps link) connecting it to a different Internet backbone provider. One goes to AboveNet and the other goes to Exodus. Full BGP peering is used with both backbone providers, so if our connection to either provider goes down, all traffic is automatically routed through the other provider.

There are two independent backup power supply systems. Half of the network equipment is powered through each backup power supply system. Each backup power system can keep the equipment running for 4 hours in case of a power failure. And, there is a backup generator available in case the power goes out for more than 4 hours. Two different power company substations feed the building in which all of the equipment is located. We first placed our equipment in this building in 1994 and during the subsequent years it has NEVER suffered a power failure.

The only way to get better redundancy in your internet services is put identical servers in two or more different locations and load share between them. If you need that kind of reliability, please contact us to discuss your needs. We have agreements in place to allow us to host servers at both AboveNet and Exodus, so we can provide this service when you need it.