Cisco Data Center Ethernet

Abstract

Cisco has worked very closely with VMware to create new technologies that enable virtual machine aware policies in data center switches that have migratory properties. This Cisco training white paper provides an overview of some of Cisco's data center platforms and their virtualization-aware features; specifically the Nexus product line of switches (1000V, 2000, 5000, 7000) and the Cisco Unified Computing System (UCS).

Sample

Introduction

Cisco's Catalyst 6500 (Cat6K) series platform is one of the most widely deployed switches used for data center and enterprise network connectivity. The Cat6K platform has been in production for nearly a decade and the longevity of the platform is a testament to Cisco's value proposition of investment protection. Cisco has introduced many chassis, power supply, supervisor and line card modules to assist with ever-changing network capacity requirements and technology evolution.

VMware and other virtualization products are revolutionizing network requirements of the data center as dedicated application servers are migrated to virtual machines (VMs). These larger servers can accommodate the computational requirements of running multiple server-based operating systems on one hardware platform. Traditional network switch platforms can only apply network policies statically on a per port basis. VMware virtualization has created a requirement for network policies to be performed on a per virtual machine basis. VMware technologies like VMotion and Distributed Resource Scheduling (DRS) enable automatic virtual machine migration across physical servers, further increasing the complexity of switch requirements. Network policies need to be able to migrate between switches (or ports on the same switch) to assure proper network policy for migrating virtual machines.

Cisco has worked very closely with VMware to create new technologies that enable virtual machine aware policies in data center switches that have migratory properties. This white paper will provide an overview of some of Cisco's data center platforms and their virtualization aware features; specifically the Nexus product line of switches (1000V, 2000, 5000, 7000) and the Cisco Unified Computing System (UCS).

Data Center Evolution

Data centers began as centralized mainframe-based systems that had the benefit of centralized administration, but the disadvantage of limited flexibility and vendor choice. Most of the applications maintained in mainframe environments were migrated to server-based systems in the late 1990s to early 2000s due to lower costs, open standards, the Internet revolution, and a diverse talent pool. The exponential growth of server-based systems has resulted in a condition referred to as "server sprawl" where large enterprises have hundreds of different servers dedicated to certain applications. Network managers have been tasked to minimize the costs associated with operations, administration, maintenance, and provisioning (OAM&P) of these vast distributed systems, while also maximizing the utilization of computing resources.

Computing Resource Optimization

When put into service, servers and clustered server systems are sized for peak traffic loads to accommodate the busiest hour of the busiest day of the year. For instance, if a retail organization did not have enough resources to accommodate the holiday shopping rush, the cost of lost business could be very detrimental. Servers sized for these maximum utilization periods will typically run at a very low utilization for eleven months of the year, however. Under-utilized servers are very costly to IT budgets. VMware virtualization technology allows multiple operating systems to share the same hardware resources, while also providing fault isolation. One misbehaving operating system (OS) should not be able to affect another OS running on the same hardware server platform unless the two servers have dependencies on each other. Virtualization software allows the OS of individual virtual machines to share the multi-core processors and large random access memory (RAM) pools to maximize the utilization and efficiency of server solutions.

Virtualization

Virtualization technology operating systems are referred to as a "hypervisor". A hypervisor is a thin operating system that recognizes the hardware resources of a server and makes the resources available to the operating systems running on top of the hypervisor in virtual machines. VMware's hypervisor (VMkernel) uses virtual network interface cards (vNICs) that direct traffic to the physical NIC of the server hardware platform. vNICs allow many virtual machines to map network traffic to the same physical server NIC.

Network switches forward Ethernet traffic based on the destination media access control (MAC) address of the network traffic. vNICs use vMAC (virtual MAC addresses) to individually identify a virtual machine using the same physical NIC. Traditional network switches are not able to forward traffic received on a switch port back to the same port (hair pinning), which is a requirement for applications communicating between VMs that are running on the same physical server. Virtual machines located on the same server communicate through VMware software vSwitches without sending traffic through any physical NIC cards. vSwitches enable communication between two VMs on the same server, something that would otherwise be impossible using a physical network switch.