Cisco’s UCS: a new architecture for a new data centre

You may wonder why Cisco is competing in the already heavily populated server market. The answer is, they aren’t. The Cisco Unified Computing System (UCS) is not so much a server as a complete integrated data centre: a new architecture that redefines the way we have hitherto used our server systems. 

Cisco has long dominated the large switches and routers market. These large switches have effectively been blade servers, combining CPU functionality and I/O for more than a decade. So the move into mainstream servers is simple and logical. But the company hasn’t stopped there. It has taken the best building blocks of contemporary data centres – the very best of computing, networking, server virtualisation and storage resources – and has integrated them into one unified fabric. No longer should we say ‘the network is the computer’; with UCS now we can say ‘the fabric is the computer’, which needs to be wired only once and can dynamically provision resources.

UCS integrates server virtualisation software from VMware, network storage from EMC and NetApp, applications from any vendor, and its own converged network intelligence (native Ethernet or Fibre Channel over Ethernet) so that combined they provide the foundation that forms the UCS architecture. A new Ethernet standard called the Data Centre Ethernet (DCE) is used to handle all traffic inside the system by dividing the total available bandwidth into eight virtual lanes to reserve bandwidth for a specific type of traffic. This provides a guaranteed bandwidth even in an oversubscribed system. For external connectivity, UCS uses converged network adapters (CNA) to allow server TCP/IP network I/O (iSCSI) and storage I/O (FCoE) to run concurrently on the same adapters.

The basic building blocks of UCS are:

- UCS Manager
- UCS Blade Server chassis
- UCS Blade Servers
- UCS Fabric interconnect
- UCS Network Adapters
- Fabric extenders

The result is a consolidated, unified and virtualised server architecture that combines rack-mounts and blade servers under a single management window with the tools for rapid deployment. It integrates a low-latency, lossless 10 Gigabit Ethernet unified network fabric with enterprise-class x86-architecture servers to be able to connect to the LAN, SAN, or HPC environment.

Key benefits of a UCS data centre include:

- a smaller footprint in the data centre
- reduced energy consumption
- fewer servers with more available memory
- simpler management through the unified management window
- fewer switches, cables and adapters

In future posts we’ll take a closer look at other virtual data centre technologies.

Written by: Zahid Hassan