Consolidation of sites and resources demands major changes in WAN architecture and backup strategies. New technologies in the data center, such as virtualization, grid computing, and clustering dramatically affect information processing and therefore network requirements, leading not only to increasing demands for bandwidth, but also to fundamental changes in the way data flows across enterprise WANs.

At the same time, the shift to a data-center-centric model for enterprise computing continues to increase the importance of the reliability, scalability, and management of the underlying WAN infrastructure. Enterprise organizations must plan for these changes, and must concentrate on obtaining WAN services capable of meeting these changing requirements, including MPLS, Carrier Ethernet and VPLS.

The enterprise migration to data-center-centric computing has led to the creation of three distinct classes of enterprise WANs:

• ‑Data center-to-data center: comprised of high-bandwidth, low-latency services such as Ethernet metropolitan area networks (MANs), wave division multiplexing, and line-of-sight wireless.

• ‑Data center-to-branch: comprised of technologies such as Ethernet and MPLS.

• ‑Branch-to-branch: comprised of technologies such as MPLS, T1/E1 and fractional services, as well as emerging technologies such as digital subscriber line (DSL) and cable.

Each of these three networks has distinct requirements. For example, DC-to-DC networks require high and scalable bandwidth to support storage, replication, and perhaps virtualization and grid computing, while branch-related networks require application performance management and QoS to make the most of limited bandwidth.

Consolidation: More Eggs in Fewer Baskets

Most large IT shops are in the process of data center consolidation. Almost all of the IT executives that participated in Nemertes Research’s “New Data Centers” benchmark were in the process of consolidating data centers. Consolidation of data centers places greater importance on availability and the resiliency of existing and planned data center locations. Business demand for higher availability is particularly high in the financial services sector, and half of all New Data Center participants list availability as one of their top challenges. Correspondingly, almost half of the participants are setting up secondary data centers for continuity and disaster recovery, either by repurposing an existing data center, or by building new facilities into which they consolidate existing data centers.

The use of secondary data centers is not new to Wall Street. Unfortunately, the tragic events of 09/11 highlighted the fact that having a backup data center across the Hudson was not good enough. Similarly, since Hurricane Katrina, disaster recovery consultants and managers – across the country - have re-evaluated traditional assumptions about the potential geographic footprint of a disaster. As a result, secondary and tertiary data centers are often built more than 200 miles from the primary data center, leading to complex storage replication challenges. Latency, storage architecture, protocol and I/O characteristics all must be properly managed to maintain performance across distributed data centers.

The bottom line is that consolidating to one data center increases the demand on WAN availability and resiliency. Consolidating into multiple, geographically distant, data centers, combined with business continuity requirements increases the demand on the WAN, dramatically.

Solutions and Services

The changes in data center architecture are driving the emergence of “flatter” two-tier network architectures. In this new architecture, data center connectivity typically comes in two flavors: interconnections between data centers, and connections between the data centers and the rest of the organization.

As noted, although in some cases the common protocol is MPLS, more typically financial services organizations rely on higher-speed interconnects between data centers, such as dense wave-division multiplexing (DWDM) over dedicated fiber, with MPLS used between data centers and the rest of the company.

Enterprise IT executives who participated in the Nemertes Research benchmark “Building the Successful Virtual Workplace” reported an increasing trend toward MPLS for WAN services to support real-time applications, with 56% of enterprises now deploying MPLS (up from 42% a year ago). Drivers for MPLS typically include support for peer-to-peer traffic flows associated with voice and video, and the ability to prioritize real-time applications across the WAN.

How “fat” are the pipes into each data center? The mean bandwidth into the typical data center is approximately 90 Mbit/s, or roughly two T-3s (each T-3 is 45 Mbit/s). However, data rates vary widely by industry, with education leading the pack at 145 Mbit/s, followed closely by financial services with 96 Mbit/s.

In addition to MPLS and DWDM, Carrier Ethernet services show great promise as inter-data-center WAN solutions. Carrier Ethernet provides high bandwidth (typically 1 Gbp/s – 10 Gbp/s, with some providers offering interface speeds up to 20 Gbp/s) at relatively low latency. Traditionally, Carrier Ethernet has been considered more a metropolitan area network (MAN) technology than a WAN technology. However, the delineation between MAN and WAN is becoming blurred. There are now Carrier Ethernet providers that can provide 20 Gbp/s Ethernet coast-to-coast. And, as these services mature, convergence is occurring. Service providers are now extending Ethernet across the WAN, over MPLS. This technology is called Virtual Private LAN Services (VPLS). VPLS addresses some of the specific challenges with extending Ethernet across the WAN, most notably scalability and availability. Adoption of Carrier Ethernet services as a WAN and inter-data-center service is rapid. Nearly half of the companies that Nemertes works with say that they are, or plan to deploy some form of Carrier Ethernet service in the next 18 months.

Conclusions and Recommendations

Instead of the graduated, multi-tiered WAN architectures of yesteryears, WANs today are increasingly flat, with one type of interconnect (increasingly, MPLS) linking data centers to branch offices, and another type (increasingly, DWDM/fiber/Carrier Ethernet) linking data centers to each other. Moreover, consolidation increases the demand on data center network availability, driving the need for effective multi-homing (both for Internet and WAN connections).

How should the changing data center architecture be considered when looking at next generation WAN solutions? Some pointers:

• ‑Architect redundant WANS. Seek both physical and logical multi-homing, both for internal and Internet connections.

• ‑Assess the performance characteristics of your next generation WAN options. Develop an application-delivery performance management framework, to ensure predictable, consistent, and solid performance of the network and applications.

• ‑Deploy MPLS. Particularly if your organization is large (over $1B) or has global operations, if you haven’t rolled out an MPLS WAN, now is the time to do so. In another year, you’ll be behind the curve. Same is true for Carrier Ethernet.

• ‑Seek dark fiber. Although the fiber glut of the past few years has diminished, fiber is still available from a plethora of sources. Many high-end organizations are adopting a two-tiered architecture, with DWDM/dark fiber linking data centers to each other, and MPLS linking data centers to branch offices.

Ted Ritter is a research analyst with Nemertes Research (www.nemertes.com). Founded in 2002, Nemertes Research specializes in analyzing the business value of emerging technologies for IT executives, vendors, and venture capitalists. Recent and upcoming research topics include security and information protection, mobility and collaboration technologies, and outsourcing. For more information, please call 888-241-2685 or email research@nemertes.com.