DCA3 Network Infrastructure & Architecture Summary



Our DCA3 advanced Data & Network Operations Centre (NOC) in Springfield, Virginia is an essential part of our core network. The DCA3 NOC is staffed by highly experienced Systems & Network Administrators 24 hours a day, 365 days a year, and features eight (8) parallel (redundant) UPS (uninterrupted power) systems, two parallel (redundant) 1.25MW diesel generators, and multiple (redundant) AC systems.

The front-end local network is arranged in three logical segments:

Core - redundant Cisco 12416 routers connected to the Internet and the rest of our national coast-to-coast network via numerous 10 Gigabit Ethernet lines; running the full BGP4 protocol, thus ensuring route optimization and redundancy.
Distribution - redundant high-end Cisco Catalyst 6513 MSFC3 / Supervisor 720 master routing switches that distribute access to the physical and virtual LANs and route internal traffic.
Access - redundantly 2 Gb/s uplink connected Catalyst 2950G-48-EI switches to which all servers are connected. Each and every server gets a full 100Mb/s full duplex port to the access layer. 1 Gb/s access ports are also available.


The private back-end local network is a fully separate, autonomous network comprised of the following two layers:

Distribution - redundant high-end Cisco Catalyst 6509 MSFC3 / Supervisor 720 master routing switches that securely route the private back-end network physical and virtual private LAN internal traffic.
Access - redundantly 2 Gb/s uplink connected Catalyst 2950G-48-EI switches to which all servers are connected. Each and every server gets a full 100Mb/s full duplex port to the private back-end network. 1 Gb/s ports are also available.

Our network does not use any 10Mb/s or 100 Mb/s switches, or any hubs - at all. It is completely 10 Gigabit Ethernet (multiple, redundant 10,000 Mb/s for all network layers) and Gigabit Ethernet (1,000 Mb/s for the individual server access layer) based. Both internal and external network provide multiple, redundant, paths. Each and every piece of the network is fully redundant: there is no single point of failure. The network also includes a number of multi-terabyte tape and RAID-based backup systems.

Physical Network Infrastructure

Cox Communications, the nation's third largest cable service provider, and a very financially solid carrier, provides our private fibre-optics through multiple inter-connected OC-192 SONET rings (DWDM capable) throughout northern Virginia and DC.

IntelliFiber, a part of Cavalier Telecom, another financially solid carrier, provides additional private fibre-optics out of DCA3 for us, through multiple inter-connected OC-192 SONET rings (DWDM capable) throughout northern Virginia and DC.

The multiple, different carrier, OC-192 SONET rings consist of two redundant OC-192s traversing two separate paths and having two diverse routes to our facilities, thus providing the ultimate in reliability and performance. Even if both of the lines are cut at any point, the OC-192 ring is still operational, with less than 50 ms (0.05 second) self-healing restoration time.

The multiple private SONET rings provide us the ultimate in reliability: no downtime even in the case of a major fibre cut in the area (due to separate paths and multiple fibres).

Other fibre carriers, such as Verizon, are also present on-site.

The DCA3 site has lines run within northern Virginia and the District of Columbia. The directly-connected IAD1 and DCA2 sites further augment that by additional Virginia, DC, New York, Illinois, California and Washington state based connectivity. Due to this fully redundant physical connectivity -- lines traversing from us in all directions -- ensures that we will remain up and running even in case of a major disaster in the area.

Back to Top Navigation

Logical Network Architecture


DCA3 Network Winter, 2016 Spring (map)
DCA3 Network Winter, 2016 Spring (footer)
   

Run a traceroute from our DCA3 network.

Traffic from DCA3 considers all the outgoing routes - be it at the same site or at one of the many other sites - weighted for physical distance, reliability, performance and latency, and picks the most direct, lowest latency route available to the destination. That is, all the packets take the most direct, fastest, way out and have a full BGP selection of all the transit providers and peers at all the sites, not just the transit/peering lines present at that particular site. The network is fully inter-connected and meshed.

For more details on our core network, such as a full backbone connectivity and peering diagram and network maps of our transit Tier 1 backbone connectivity providers, please review details on our core network.

All customers and independent tests find our network to be better performing than any single competitor. Our industry-leading SLA guarantees that.

Back to Top Navigation

Further Advantages of Our Network

The northern Virginia area is an ideal location for a Data Centre. A very large portion of the world's Internet traffic passes through here, and the area is not subject to earthquakes, tornadoes, floods or other natural disasters.

We are located at the true heart of the Internet, and in a very safe and secure area. This central location ensures optimal connectivity, speed, and routing, minimizing the number of hops - and thus the chance of any problems. We are located right beside a great number of private and public peering points, in one of the most densely and best connected areas in the world, including directly peering at the Equinix Exchanges (IBX'es) in Ashburn, Virginia and Chicago, Illinois, at the NYIIX in NYC, New York, at PAIX in Palo Alto, CA and Seattle, WA, at the SIX in Seattle, WA, at Any2 in both Los Angeles, CA and San Jose, CA, and at PacketExchange in Ashburn, VA, Chicago, IL, Palo Alto, CA and Seattle, WA. All providers have multiple, redundant routes into the DC/VA area. This is further built upon by our redundant connectivity to multiple Tier 1 providers via separate circuits, terminating at diverse locations in Virginia and the District of Columbia. We are redundantly set-up Internet connectivity wise (multiple backbone connectivity providers), as well as physically. In terms of fibre optics, we have multiple private OC-192 SONET rings with multiple financially-healthy fibre providers, taking different physical paths.

We have built our facilities in the best possible area for Internet connectivity, and then further reinforced it with redundancy in all possible forms. We operate four sets of two parallel (redundant) PowerWare UPS systems - eight (8) systems total, with enough battery capacity being able to power all the DCA3 servers for over an hour in case of a power outage. In case of a more extensive outage, the two parallel 1.25MW diesel power generators - that can run indefinitely, if need be - would kick into effect. Even if one generator fails, there would be no impact whatsoever and all would remain running, as the generators are set-up in a parallel, fully redundant, mode. Multiple redundant DataAire precision cooling AC Units - over 200 tons all together - are in use, too, maintaining a consistent temperature of 18C and a dust-free environment at our Data Centre. In addition, there is ample room for expansion: for adding additional generators (including large fuel tanks), AC units, outside condensers, and bringing in more utility power, past the current 2000A @ 480 Volts, 3-phase, that it has now; in fact, all is prepared already in advance to double the power and AC capacity of the facility.

To ensure the best routing of our customers' data, we pay for commercial top-priority transit network capacity to our tier 1 backbone providers and have full burst capacity available at all times on all circuits. Thus we receive the highest level of service, including 99.99% or 100% uptime and no packet-loss guarantees, from all our connectivity providers. 90%+ of traffic from us to end users only passes over one or two backbones on its way to reaching the end destination. In addition, we are committed to under 50% network capacity utilization at all times; we ensure that our network always has double the capacity it needs, even at peak hours.

Our network design and operations ensure the very best connectivity, speed, and reliability possible: 100% network uptime, no packet loss, and under 3 ms low latency connection to Tier 1 backbones - guaranteed by our industry-leading SLA.

Our DCA3 network & site in summary:

located in Springfield, VA - right at "the centre of the Internet" and a natural disaster free zone
multiple (parallel, redundant) UPS systems & diesel generators provide for power backup & redundancy
200+ tons of AC cooling capacity, expandable to over 500 tons
fibre redundancy through multiple private OC-192 SONET rings
redundant architecture 10 Gigabit Ethernet-based internal network, using Cisco backbone-grade equipment
redundant, geographically dispersed, connectivity to all of the world's major backbones