Why Scalable Throughput Now Required in New-Gen Data Centers

By Chris Preimesberger  |  Posted 2012-10-02

An industry survey released Sept. 28 of 1,529 IT decision-makers across North America and Europe regarding issues with data center networks shows that IT departments—including CIOs and corporate executives—are genuinely concerned about how to keep pace with the demand for increased throughput on their networks.

As has been well-chronicled here in eWEEK and in other publications, the unprecedented upsurge in new data moving through networks is being driven by a perfect storm of trends that includes server virtualization, new machine-generated data, cloud computing, big data workloads, and the convergence of storage and data networks. IT managers know they need to think ahead on this or else face major traffic issues in the future.

Storage networking hardware and software maker Emulex Corp., which commissioned the study, has put together a closer look at the survey response, which underscores the need for a strong, stable and highly scalable I/O foundation.

Here is a list of highlights.

Half all IT departments will scale networks up to 100G bps by 2016: Eighty-one percent of survey respondents said network bandwidth demand is one of the most critical issues facing data centers, driven by increases in virtualization, cloud computing, big data and convergence. In fact, 54 percent of all IT departments are being asked to scale their networks to support speeds of up to 100G bps by 2016. Forty percent already have deployed 10 Gigabit Ethernet (10GbE), and in another four years, survey respondents say the majority of those networks will be operating at 100GbE. 

Network I/O—the most critical potential bottleneck: Seventy-six percent of IT leaders said network I/O is now the biggest bottleneck in the data center, making the need to improve data movement a high priority for enterprise IT. In fact, 70 percent of survey respondents said that improving I/O performance in their data centers is a "high" or "very high" priority. Additionally, more than half of survey respondents said that their need for network I/O increases by 60 percent or more annually.

Virtual machines per physical server will see a big-time increase: Eighty-five percent of survey respondents said network I/O was a "significant" or "very significant" gating factor in limiting how many VMs they could run on a single physical server. Seventy-two percent of respondents believe it will be practical to run as many as 100 virtual servers on a single physical server in just two years, which will significantly affect the way networks are designed and deployed. 

Cloud computing significantly increases bandwidth demand: Fifty-three percent of survey respondents reported that the advent of the cloud has driven the need for additional network bandwidth within their corporate data centers; 40 percent further said that the bandwidth expansion required to support migration to the cloud was 25 percent or more.

If your network can't scale, neither can your cloud: Nearly half of survey respondents said cost savings was a driver of cloud adoption in their organization, making it the No. 1 reason organizations are moving to the cloud. Scalability and speed of deployment are a close second and third as cloud catalysts. However, to have a cloud network that can truly scale—and do it quickly—to deliver capacity on demand, enterprises need to overcome the inherent limitations of current multitenant infrastructures. Overlay networking, which relies heavily on strong, stable, highly scalable I/O, can help.

Big data has arrived: More than a third (37 percent) of survey respondents manage 1 petabyte or more of data, and 11 percent said they currently manage more than 100PB of data. The big data apps contributing to this deluge—identified by 50 percent of respondents as managing at least 100 terabyte of data—are dominated by business intelligence applications (24 percent), financial applications (18 percent), security applications (11 percent) and scientific applications (11 percent).

Big data needs big bandwidth: Fifty-four percent of all survey respondents said that over the next two years, big data would increase their organization's need for network bandwidth in the data center by more than 50 percent.

Big data also requires big storage: Thirty-nine percent of survey respondents said that, to-date, big data applications have increased their organizations' need for storage by at least 50 percent.  Further, 20 percent of respondents said that over the next two years, big data applications will cause the amount of storage connected to data center networks to double.

Convergence is inevitable: Seventy-eight percent of survey respondents said they believed it was inevitable that data and storage networks would converge into a single, consolidated network. In fact, nearly two-thirds (64 percent) of IT organizations have either already moved to a single, converged network for data and storage, or have plans to do so in the future. This convergence has had a "very significant" or "extremely significant" impact on the need for increased I/O, according to 45 percent of respondents.

IT's perfect storm: The timely intersection in the last three years of higher bandwidth, virtualization of software and hardware, cloud computing services and infrastructure, big data workloads and more powerful yet lower-power processors had brought about this sea change in the data center. These factors need a scalable network.

Rocket Fuel