Data Centers

Geographically Dispersed

Our customers business are worldwide so we made sure that our physical server deployments would match them. We first launched services from the NYC metro area. Then to provide better service to the Asian demographic we opened up space in San Jose, CA. Finally to better serve Europe we opened up two separate spaces in Amsterdam. Customers can choose any location individually, or through load balancing or our content delivery network (CDN) be dispersed transparently amongst all locations, allowing their users to be automatically directed to the geographically closest location tremendously improving their load times.

Cooling

The two largest problems with building out datacenter space are power and cooling. With cooling we made sure to select facilities that have very stringent power requirements. These strict guidelines ensure that power consumption and the heat it creates is properly dissipated by the on site air conditioning and monitoring units. We also employ datacenters with raised flooring which helps in the best distribution of cooling throughout the large spaces as well as creates an easier cable management environment than overhead ducting.

Power

All points of power delivery to the racks are doubled up to ensure that failure of any part of the system is automatically switched over to the redundant carrier. In addition each facility has onsite power generators which are automatically switched on in the case of an emergency. Each rack comes equipped with metered PDUs that display the AMP usage for the rack.Each one also has two circuit breakers. To further ensure that we can scale and grow we leave each PDU with 33% spare capacity.

Security

All of our datacenters are staffed with security personal 24/7/365. Each requires biometric hand scans and key cards for employees to gain access to all areas. The facilities themselves are monitored by the security staff as well as by a separate network of cameras that cover every inch of the datacenter, ensuring that there are no blind spots. The facilities themselves are located in non-descript buildings which helps to avoid unwanted visitors.

Network

Design

It is imperative that networks have built in fault tolerance and can properly switch over traffic from failed links. Our networks were all designed with this mind ensuring that when a link goes down our customers are not affected in either downtime or increased latency. This extends to how we lay out network equipment as well, using a stand-by fail-over in case of hardware failure for every piece of the network. Additionally depending on the service the customer selects we also have the application layer which provides load balancing, not only between servers, but also between geographical and datacenter locations to further ensure uptime.

Hardware

Everything is dependent on the network as a result we exclusively use only the market leader, Cisco, when purchasing network equipment. This allows us to not only create the network configurations necessary but also ensures that the hardware is up to the task. As the networks grow and more data is pushed through the switches the hardware pushing that traffic needs to be reliable and Cisco has proven themselves time and time again.

In addition to Cisco switches we also deploy firewalls and load balancers when customers request them to increase security and redundancy. We also provide private networks as well for customers that have large application development environments and want to create the fastest and most secure links between their servers.

Carriers

When it comes to bandwidth it's imperative that the providers be reliable and varied. Even when downtimes are mitigated all providers still have routine maintenance that they must perform that can impact service. Because of this we use a wide variety of providers in our network such as Level3, Verio/NTT, MFN Abovenet, Global Crossing, and others.

In addition to transit carriers peering is equally important. These are direct connections to other large networks. This means that packets traveling between our network and the peer have the best possible routes. We have a large peering network both in the United States (which also includes many cable providers) and in Europe that we rely on. Peering and multi-homed carriers allow us to provide the best possible service to our customers.