Cisco Nexus 5000 Series Switches

As businesses move to evolve their data center infrastructure to meet increasing business demands, the Cisco Nexus^TM Family provides a means to protect investments in existing infrastructure and incrementally add capabilities and capacity in a cost-effective manner. This document describes the Cisco^® solution.

Businesses are facing a growing and increasingly complex set of challenges in the data center. Traditional measures of operational capability such as system availability and resource utilization are facing increasingly stringent expectations, and new metrics such as energy efficiency have been added to the mix. At the same time, because of IT's increasingly important role in business, data centers are facing a new set of external demands such as global availability, regulatory compliance, and the expectations of the empowered user.

Cisco Data Center 3.0 provides a granular, practical framework for transforming the data center to meet the evolving needs of the business with a data center strategy designed for long-term sustainability. Cisco Data Center 3.0 emphasizes investment protection and the incremental incorporation of new technologies and infrastructure, so investment is more closely aligned with business need. The underlying tenets of Cisco Data Center 3.0, consolidation, virtualization, automation, yield both reduced total cost of ownership (TCO; better asset utilization, reduced power and cooling needs, and improved operational efficiency) and improved responsiveness to the needs of the business.

Cisco Data Center 3.0 envisions the evolution of data center infrastructure through a set of defined, overlapping phases. Although the endpoint is well defined, the rate at which a business moves through these phases will vary based on the business needs of the specific business.

Today, the typical data center is populated with servers that are connected to the network with Gigabit Ethernet network interfaces and separate Fibre Channel host bus adapters (HBAs). These data centers generally have low production virtual machine density, with fewer than 10 percent of production workloads running on virtual machines, and an operating structure built around this siloed environment (Figure 1).

At this point, customers can continue to use their existing investment in infrastructure. As they begin to outgrow their existing network switches, the Cisco Nexus Family offers a number of options, including the Cisco Nexus 7000 Series Switches and the Cisco Nexus 2000 Series Fabric Extenders, to support their Gigabit Ethernet attached servers. This approach allows customers to maintain operational and management consistency with existing Cisco Catalyst^® Family infrastructure and provides forward investment protection with the capability to deploy technologies such as 10 Gigabit Ethernet, unified fabric, and virtual machine-aware networking (Cisco VN-Link) in the future.

Phase 2 is defined by customer efforts to reduce TCO by consolidating servers using server virtualization technologies such as VMware ESX, Microsoft Hyper-V, or Xen. The capability to consolidate multiple physical machines typically running with low utilization into virtual machines that can run on a much smaller number of physical servers has significant cost benefits for the customer. In this phase, the virtual machine becomes the default application platform, with 60 to 80 percent of x86 applications running in a virtualized environment.

From a networking perspective, this increase in virtual machine density encourages a transition to 10 Gigabit Ethernet as the default mechanism for attaching servers since multiple virtual machines on a single server can quickly overwhelm a single Gigabit Ethernet link, and multiple Gigabit Ethernet links cease to be cost effective after a certain point (Figure 2). In this phase, storage traffic still tends to stay on separate transport. In this phase, both the Cisco Nexus 7000 and the 5000 Series can support the transition to10 Gigabit Ethernet-attached servers. With the Cisco Nexus 7000 Series, this simply means adding 10 Gigabit Ethernet I/O modules. The Cisco Nexus 2000 Series Fabric Extenders can provide support to the remaining Gigabit Ethernet attached servers while maintaining a consistent operating environment across the entire network. At this point, customers can also deploy the Cisco Nexus 1000V Switch if they are running VMware's ESX hypervisor. This capability gives customers operational consistency down to the individual virtual machine as well as policy portability, so network and security policy follows virtual machines are they move around the data center. The Cisco Nexus 1000V can be deployed wherever VMware ESX is currently running. It does not have any dependencies based on the server uplink speed or upstream access switch.

The third phase marks the transition to a unified data center fabric and generally has two triggers. The first trigger is the desire to continue to reduce TCO by simplifying infrastructure and eliminating the redundant elements (interfaces, cables, upstream switches, etc.) needed to support a separate LAN and storage network. The second trigger is the desire by customers to do more sophisticated things with their virtual machines, including use of technologies such as dynamic resource scheduling (DRS). These goals require all servers to have a consistent and ubiquitous set of network and storage capabilities, and one of the simplest and most efficient ways to accomplish that is to deploy a unified fabric. The shift to a unified fabric gives all servers, physical and virtual, access to the SAN, which yields further TCO reductions and efficiency gains for the customer by allowing more storage to be consolidated in the customer's SAN (Figure 3).

Since this phase focuses on consolidation of server I/O, the primary emphasis is on adapting the server access layer to support a unified fabric. To deploy Fibre Channel over Ethernet (FCoE), on the server side of the equation this goal involves either a new converged network adapter from companies such as Emulex and QLogic or, in the case of Intel, a new software driver for its 10 Gigabit Ethernet network adaptors. Note that FCoE is supported on VMware ESX 3.5 U2, and the Emulex, Intel, and QLogic interfaces are on the VMware 3.5 hardware compatibly list (HCL), so this phase encompasses both physical and virtual servers. On the network side, enabling FCoE is simply a matter of enabling FCoE features in the Cisco Nexus 5000 Series and installing either the Fibre Channel or Fibre Channel and Data Center Ethernet uplink modules. Any attached Cisco Nexus 2000 Series Fabric Extenders also inherit FCoE capability; however, because the uplinks are oversubscribed, care must be taken with traffic engineering. While the Cisco Nexus 7000 Series is FCoE capable, Data Center Ethernet-capable I/O modules (to provide reliable transport) will not be available until late 2009.

The goal of Phase 4 is to build on the functional benefits of the previous phase and increase the scalability, flexibility, and efficiency of the data center data fabric (Figure 4). This phase allows any data center asset to access any other asset. From a storage perspective, the data center fabric will support FCoE and iSCSI server access and FCoE, Fibre Channel, and iSCSI attached targets. Data Center Ethernet will expand from the access layer into the aggregation and core layers. One of the benefits of this expansion is simplified access to the Fibre Channel SAN, since Fibre Channel no longer has to be backhauled from the access layer over dedicated links.

As noted earlier, this phase builds upon the prior phase, so it is primarily concerned with completing the migration to a Cisco Nexus 10 Gigabit Ethernet infrastructure. The one new element is the addition of FCoE interfaces to the existing Cisco MDS 9000 family of director-class Fibre Channel switches to simplify access to the existing SAN.

The ultimate goal of this solution is a fully virtualized data center composed of pools of computing, network, and storage resources. Services such as security and Layer 4 through 7 processing (for example, load balancing) are also fully virtualized and can by implemented whenever they are needed. Underpinning this facility are automated management and provisioning capabilities. The result is the capability to dynamically create and tear down application environments based on both policy and real-time triggers. The net benefit to the business is improved cost efficiency and tighter coupling of IT to business needs.

The data center up to this point in its evolution has been confined within the walls of the data center. However, with the virtualized environment defined by Phase 5, where a workload actually runs becomes largely irrelevant; the resources needed to support can just as easily be supplied by an outside service provider, opening new means of innovation and optimization. This is the premise behind cloud computing. While some time is required before a fully mature cloud computing model is realized, the evolutionary path described in this document will prepare customers to take advantage of it as it emerges.

Cisco Data Center 3.0 provides a comprehensive vision of how to move the data center along a path that provides both business and technical benefits in a way that is granular and cost effective. Cisco has made Data Center 3.0 a practical reality by delivering the products and technologies needed to achieve the vision. Just as important, Cisco Data Center 3.0 represents a complete solution, including technology, professional services, financing, and a broad ecosystem of development and solution partners including APC, EMC, Emulex, HP, IBM, Intel, Panduit, QLogic, and VMware.