Choosing Your Data Center

Choosing Your Data Center: What to Expect from Your Hosting Provider :PART II

By: Paul Froutan VP of Data Center Operations ,from ; developersnetwork.com

HEATING, VENTILATION AND AIR CONDITIONING (HVAC)

The HVAC system is another critical part of a reliable infrastructure. As HVAC units operate continuously and require ongoing maintenance, it is critical to have redundant units available in case of failures and during maintenance windows. The provider should also have equivalent HVAC to the power system. For example, if there are 100 kilowatts of power in a data center, simple conservation of energy dictates that the HVAC system should be able to displace 100 kW of heat (equivalent to about 30 tons). Customers should therefore beware of providers who quote more power than HVAC. While the power system may really be able to provide that power, the HVAC system will not be able to cool down the data center properly. Liebert is still the premier supplier of data center HVAC systems. However, competitors such as Data Aire can provide similar units at lower prices.

IP NETWORK

The final component of a reliable system is the IP network. The network consists of two parts; the provider’s equipment, including routers and switches, and the transit links that enter the data center. First, let’s look at the provider’s equipment.

Until two or three years ago, Cisco was the undisputed premier vendor of network equipment. However, with the emergence of several new and nimble players in the field, there are now many other vendors to choose from. One issue to consider in multi-vendor environments, however, is inter-operability. While a switch, for example, may have great performance on its own, it may not be very reliable when working in conjunction with a router from another vendor. Providers should have a lab where their production setups can be duplicated for testing and for planning upgrades and maintenance.

Another aspect is reliability. Providers should have redundant switches and routers in every layer of the network, and these layers should be meshed so they can withstand several failures without any impact on the customer. The network should withstand at least one failure in each layer and continue to function at full performance. Some providers have backup systems consisting of lower capacity equipment. However, if there is a failure at peak times, that setup will lead to poor performance.

WIDE AREA NETWORKS

The second part of the network is the provider's WAN (Wide Area Network) connection. This part again consists of two different pieces; the "transport" and the "transit". The "transport" is the circuit that carries the data, while the "transit" is the actual data. Unless you’re familiar with IP networking, that may be confusing so here’s an analogy. The “transport” is like a water pipe. You can have a large or small pipe. That pipe comes into your house. On the other end, the pipe can be connected to one or several water systems that provide water. The water is the “transit”. You can have a large empty pipe that doesn’t do much good. You need the pipe and water for it to be useful. Many providers state that they have multiple OC-12s or OC-48s. However, they don’t mention how much actual transit they have. Again, empty circuits are useless. It is therefore important to find out how much actual transit they have and to which providers. Just as empty pipes are useless, having all the transit on a single pipe is very unreliable. That circuit can have many problems, from a physically cut fiber in the street, to a misconfiguration at the Central Office. Therefore, a provider should have multiple fiber providers and multiple transit providers, a minimum of three of each. Any two should be able to handle the maximum traffic of the network. Therefore, two T-1s (at 1.5 Mbits) and one T-3 (at 45 Mbits) is not considered redundant as if the T-3 fails, the provider loses over 90% of its capacity.

<<Back>>

Other Related Links