Have an account?

  •   Personalized content
  •   Your products and support

Need an account?

Create an account

Design Zone for Massively Scalable Data Center

Build a Data Center Framework

Architect a Massively Scalable Data Center for big data and web 2.0 applications

Massively Scalable Data Center

December 21, 2012

A Cisco Massively Scalable Data Center (MSDC) is a collection of design tenets realized in a lab architecture. The overall architecture builds a scalable Layer 3 data center fabric over a three-stage Clos topology. The architecture is deployed to enable horizontally distributed applications, such as those found in warehouse-scale, hyperscale, and Web 3.0 environments. These applications generate workloads that require symmetric bisectional bandwidth between any computing or storage nodes attached to the network.

These applications are deployed on a cluster that is expected to scale up and down from hundreds of nodes to tens of thousands of nodes in order to serve millions of concurrent connections from clients. To support the elasticity in the control plane, the MSDC reference architecture uses Border Gateway Protocol (BGP) and open shortest path first (OSPF) as the control plane protocols.

To support the peak bandwidth requirements of some of these applications, which easily exhaust the bandwidth offered by a 10 Gbps link, the reference architecture uses Cisco Nexus 7000 and Nexus 3000 Series network platforms.

Cisco MSDC framework guides a data center architect in defining an approach to critical issues involved in MSDC architecture such as:

  • Data center fabric topology choices
  • Control plane protocol selection
  • Network provisioning and management
  • Network diagnostics, analytics, and troubleshooting
  • Modeling port buffer utilization in the face of specific traffic patterns
  • Platform selection with an eye on power management
  • Network programmability and open APIs

Additional Resources


For Partners

Log in to access tools and resources specifically designed for Partners.