Now that we have reviewed the evolution of servers and data center technologies, we are ready to look more closely at the architecture of a typical data center. Pictured below are the various elements in a representative format, resembling some designs that I have created over the years.
As we have discussed in dome detail, servers lay at the heart of the data center, performing numerous tasks and providing applications to the enterprise they serve. This may be a mix of rack-mounted and blade servers, though trends for greater density in servers would slant toward a blade-based infrastructure.
Mentioned only briefly earlier, large-scale storage arrays provide massive data archiving capabilities, often using completely separate networks using fiber channel protocols. Data transfer in this type of network takes place at a block level, which is intolerant of loss or delay. Using specialized fiber channel switches and unique server adapters called Host Bus Adapters (HBAs), this adds another layer of complexity to the overall infrastructure. This is referred to as a Storage Area Network (SAN).
From a design perspective, the preferred method of network management is to make use of an out-of-band (OOB) network that is logically, or even physically, separate from the network carrying user data (production traffic). Many network devices actually have dedicated Ethernet management ports to facilitate this exact structure. In a physically separate management network, a separate set of switches connects the OOB interfaces, in essence requiring another network for this purpose.
Cisco Data Center Training