Data center transformation--Sophisticated switching drives fast, fat, and flat

Network virtualization technologies have evolved rapidly over the last few years, illustrating dramatic potential for evolution in longstanding network infrastructures. Combine this evolving technology with the growth of cloud computing, and the result is a surprisingly creative approach to improving data center performance and agility. Even enterprise networks firmly rooted in operational efficiencies are taking a fresh look at their data centers as a source for growth, competitive advantage and return on IT investment. 

The potential for transformation starts with the basic historical structure of a network--with physical servers once tiered for north-south data flow from access, distribution, and core layers to the wide area network and back again.  Today that network is moving more and more traffic patterns that are east-west in nature, inherent to modern distributed systems and applications with data that moves across racks and pods.  The result is swift adoption of fast, fat, and flat network topologies and network virtualization--creating deployments that rely on sophisticated switching technologies to deliver maximum scale and performance.

Initiating Network Transformation with VM-Aware Switching

Network usage models are indeed shifting rapidly, and network infrastructure designers must pay close attention to switching and virtualization requirements as part of this market's evolution.  Global data traffic is increasing exponentially, anticipated to reach 26 times today’s level within the next three years, and millions of minutes of video are expected to cross the network every second. Connected devices are at the heart of this growth, and projections place the number of devices expanding to two times the global population by 2015. The highly scalable, virtualized network deployments that manage these transactions must rely on sophisticated switching technologies, enabling the full potential of applications that are sensitive to network performance.

VM (virtual machine)-aware switching is essential in these environments, which increasingly require native OS-based server level performance, without performance penalties incurred by virtualized servers running multiple VMs.  Several standards-based switching technologies such as SR-IOV (Single Root I/O Virtualization by PCI Special Interest Group) and VEPA/EVB (Virtual Ethernet Port Aggregator/Edge Virtual Bridging by IEEE 802.1) have emerged to address this consideration, and ensure improved performance and scalability of applications that run in VMs. 

Effective VM switching means access layer switches in the data center network (e.g., top–of-rack switches, blade switches, or end-of-row switches) must support a large number of virtual switch ports (VSPs). In turn, VSPs in access layer switches must also support features including link aggregation, load balancing, traffic mirroring, and statistics counters similar to what is available for physical switch ports. Such features are essential to enabling VMs with the same level of reliability, performance, and monitoring as physical servers.

VM-aware switches are engineered specifically to meet current feature and scale requirements of private and public cloud networks, a new essential in virtualized networks.  Deployments can now feature virtual switch ports supporting high-level functionality, such as link aggregation, queuing, access control list (ACL), statistics, and mirroring services, that embodies many of the same services readily available in physical ports.

Smarter and More Flexible Deployments Lead the Market

Optimal switch solutions support live VM migration as part of the more prevalent east-west traffic pattern, allowing layer 2 (L2) networks to scale across pods, sites within the data center, or even across data centers. Live VM migration improves disaster recovery and is a fundamental consideration for increasing server utilization, in turn meeting the overall IT goal of implementing a dynamic data center infrastructure.

Figure 1. SmartSwitch technologies were developed to ensure that sophisticated network infrastructure design requirements can be implemented comprehensively, cost-effectively, and in volume scale.

Achieving live VM migration across servers, racks, or pods requires that they reside in the same L2 network segments, often referred to as a flat network.  L2oL3 overlay technologies further extend the benefits of these fast, fat, and flat architectures; this provides an essential means of combining L2 and L3 topologies and enabling network virtualization at cloud-scale. Consider the mega-scale or Internet-scale data center which commonly incorporates inexpensive L3 technologies for physical network infrastructure.  This historical approach follows the success and scalability of the Internet, built on IP and incorporating a scalable hierarchical addressing scheme as opposed to flat L2 addressing. When L2oL3 technologies are implemented, L3 networks have a natural mechanism to create a flat L2 network. This readily manages VM migration within and across data centers, and further enables scale and multitenancy.

L2oL3 overlay techniques are highly applicable to large-scale data centers currently relying on the proven scalability of L3 addressing and multipathing technologies. Physical L3 networks are implemented with multiple overlays or tunnels; each of these is a virtual L2 network that can then be assigned to a tenant. Virtual L2 networks can then enable live VM migration across pods, sites, or data centers that are connected via L3 networks. L2oL3 technologies provide substantial flexibility in network design, fueled in large part by the sophisticated network switches that support their inherent range of control plane features. The result is multiple bandwidth and port density configurations, creating flexibility for ideal placement in any location of the data center network.

Figure 2. Network topologies are changing with increases in rack-to-rack traffic patterns. Fast, fat and flat data center networks are required.