Planning for High-Speed Data Center Migration

Matthew Melis, left, is Director of Eastern Sales Engineering for Equinix Inc., and John Schmidt is the Leader of CommScope’s Global Data Center Solutions Group.

Big Data, mobility and the Internet of Things (IoT) are generating an enormous amount of data, and data center operators must find ways to support higher and higher speeds. Many data centers were designed to support 1-gigabit or 10-gigabit pathways between servers, routers and switches, but today’s Ethernet roadmap extends from 25- and 40-gigabit up through 100-gigabit, and 400-gigabit and even 1-terabit Ethernet loom within a few years.  As a result, data center operators have an immediate need to migrate their Layer 1 infrastructure to support higher speeds, and that new infrastructure must also deliver lower latency, greater agility, and higher density. In this article, we’ll look at the challenges of moving to higher-speed cabling infrastructure, and how to plan for the future.

 Recent data center trends predict bandwidth requirements will continue growing 25 percent to 35 percent per year. A key impact of this sustained growth is the shift to higher switching speeds. According to a recent study by Dell’Oro, Ethernet switch revenue will continue to grow through the end of the decade, with the biggest sales forecasted for 25G and 100G ports. The shift to 25G lanes is well underway as switches deploying 25G lanes become more commonplace. Lane capacities are expected to continue doubling, reaching 100G by 2020 and enabling the next generation of high speed links for fabric switches. A number of factors are driving the surge in data center throughput speeds.

Migration Challenges

There are several aspects of data center design and the evolution of cabling that present challenges to those wishing to migrate to higher speeds.

Every data center is different. There’s no standard method of deploying cabling. While standards are continually being refined around fiber optic cable and connector technology, there’s no one roadmap for implementation that suits every or even many data centers.

The pace of change is accelerating. The move from 1G to 10G Ethernet took nearly a decade, for example, while migration from 10G to 25G and 100G will take half as long. A lot of legacy networks were designed with infrastructure that’s not as scalable as it needs to be; planners could anticipate an eventual move from 1G to 10G, for example, but in most cases cabling that was installed even a couple of years ago is now outdated. Data center managers are having to update fiber or add more fiber, and that fiber must support rapid advancements to 100G and beyond.

Standards are evolving. Many data centers use multi-mode fiber to connect servers and switches, but the state of the art in multi-mode fiber was OM3 or OM4 a few years ago. Last year, standards bodies approved the OM5 standard, which has four times the throughput of OM3.

Data centers are densifying. In multi-tenant data centers in particular, customers are reducing the size of their deployments by consolidating network gear into smaller footprints. As a result, they need to be able to expand their network capacity inside a smaller environment. Some older cable management systems and patch panels can’t support higher density.

Migration is costly and disruptive. Ripping and replacing cabling is disruptive enough, but when the data center also needs higher-density cable management systems and patch panels, it can be a real nightmare. In large enterprise data centers where there is often more space, migration can take place in sections to reduce disruption, but this is not an option in multi-tenant data centers.

Planning for Migration

The most important strategy for high-speed migration is to plan for the long term. Many data centers last upgraded their Layer 1 infrastructure to support the next generation of switches, routers and servers, but because the pace of change is accelerating, it’s best to plan for the longer term. Choose a point in the future (say, 400G), assume the data center will require more fiber strands than are available today, and buy the highest grade of multi-mode or single-mode fiber available to support future migration without ripping and replacing.

In addition, data center architects should adopt low latency designs. Low latency is important in financial trading applications today, but it will increasingly become a requirement to support IoT applications such as connected cars. Data center cabling and connectors using ultra-low loss components will offer the most flexibility in achieving low latency.

Also, architects should consider single mode fiber as well as multi-mode fiber. Single mode fiber delivers the highest throughput and reach and is important in larger data centers, while multi-mode fiber is less expensive and easier to deploy.

Finally, choose the right Layer 1 infrastructure solution provider. The largest providers have global operations, so they can deliver effective solutions throughout the world. These providers have teams of field application engineers that will come out to the data center and make appropriate recommendations about which products to install for long-term viability, and some offer guarantees that their infrastructure solutions will support any application.

Building for the Future

As the Ethernet roadmap extends to one terabit and beyond and data center applications demand higher transmission speeds, data center architects should plan today to support the future. With the right connectivity infrastructure, architects can deliver a solid foundation for high-speed migration.

Opinions expressed in the article above do not necessarily reflect the opinions of Data Center Knowledge and Informa.

Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.