Data Center Grade Switch

James Liao

Data Center Grade of Switches

How is Data Center Grade of switches different from the typical switches?

The term Data Centers could apply to various entities of computing resources in many organizations. Data Centers could be the mega computing warehouse of cloud providers, where they host hundreds of thousands of servers and peta-bytes of storage. It could also be the enterprise data center of a Fortune-500 company, where hundreds of applications were hosted. It could also mean a cluster of High Performance Computing servers where researchers simulate the nuclear reactions or the biological process of DNA.

These data centers all share one unique requirement - Scalability, which makes Data Center Grade of switches distinctly different from the typical wiring closet or desktop switches.

So, how exactly does scalability make Data Center Grade switches different?

  • Bandwidth - DC (Data Center) switches typically come with higher bandwidth.
  • Density - DC switches have higher density to accommodate the real estate requirements.
  • Packet Buffer - DC switches have bigger packet buffer to handle the jitter of data rate.
  • Traffic control - DC switches have various traffic flow controls, such as Pause, over the traffic.
  • Classification - DC switches support traffic classification to ensure the bandwidth of high priority traffic.
  • Virtual LAN - DC switches support VLAN provisioning for better management.
  • Table size - DC switches should have bigger MAC/IP tables to handle the number of servers in Data Centers.
  • Trunking - DC switches support trunking capability to allow scalable bandwidth to other networking peers.
  • Storm control - DC switches should support storm control to avoid single source from bringing down the network.
  • Routing - DC switches support routing capability so users could partition the network into smaller L2 domains

Besides these scalability requirements, we are also seeing some new requirements coming from data centers.

  • Multi-path - we see more data centers demanding multi-path capability, which increases the fault-tolerance of a network and provides on-demand bandwidth increase.
  • Multi-protocol - specifically on FCoE type of application



Cancel  Add attachments