Emtec » Transforming IT

Home Page

Emtec Adviser- Unified Fabric

Reduce costs and increase flexibility with a single data center interconnect.

Pressures to reduce cost and improve agility during an era of unprecedented data growth are driving technology organizations to transform their data centers through consolidation and automation. Virtualization and cloud computing represent breakthrough technologies in this effort, but obstacles remain. A unified server I/O architecture could remove one of the more considerable barriers.

Today, the typical data center supports multiple networks — one for data and applications, one for storage, and perhaps another for server clustering. As such, servers must feature multiple network adapters that fulfill the I/O requirements of each function. What’s more, servers commonly have dedicated interfaces for management, backup or virtual machine live migration.

Supporting all of these interfaces contributes significantly to data center complexity and imposes significant costs related to cabling, rack space and upstream switches. In addition, the rat’s nest of different cables and connections required for each different function makes it harder to heat and cool the data center and contributes to skyrocketing power costs.

All Together Now

A unified fabric (UF) is designed to aggregate I/O and server resources with a “wire-once” interconnect — a single high-bandwidth, low-latency network cable running from server to switch to storage. In addition to reducing the number of server adapters, cables and switches required to support the data center, UF supports broader data center virtualization by providing consistent, ubiquitous network and storage services to all connected devices.

Although a wide range of companies (including Cisco, HP, IBM, Brocade, Qlogic, Intel, EMC, Sun, Mellanox, NetApp, Emulex and many others) now offer UF products or support, there is no consensus on the best interconnect for creating a unified fabric. The vendor community remains divided largely between two options — InfiniBand or Ethernet.

InfiniBand is well established as a consolidated, flexible, low-latency and high-bandwidth interconnect, and it is the dominant protocol in high-performance computing (HPC) environments. A March 2009 Tabor Research report noted that 60 percent of HPC organizations surveyed said they used InfiniBand as a system interconnect. And those HPC organizations that were considering a unified fabric strategy said they were more likely to consolidate on InfiniBand than on Ethernet.

“InfiniBand already possesses the attributes needed to successfully implement unified wire infrastructures, including compliance with standards, highly secured lossless fabric, high bandwidth and low latency,” said Jie Wu, research manager, High Performance Computing at IDC. “Among the benefits envisioned for these future infrastructures are simplified administration, higher application and storage performance, and lower total cost of ownership.”

Advantage: Ethernet?

However, recent enhancements could tip the scales in Ethernet’s favor. The development of Fibre Channel over Ethernet (FCoE) and Converged Enhanced Ethernet (CEE) are significant for organizations that already have a considerable investment in Ethernet and Fibre Channel, along with the management and operational structure to support them.

FCoE enables organizations to consolidate single-purpose storage networks based on FC with existing LAN and other storage traffic over Ethernet. CEE is an extension of the Ethernet standard that supports a unified Ethernet network and allows FCoE to share the same wire with multiple data traffic types — including NFS, CIFS and iSCSI — as well as the rest of the LAN. CEE allows prioritization of traffic types to maintain lossless transmission of the FCoE storage traffic while permitting lower service levels for applications that can afford temporary variability in transmission quality.

Cisco Systems has made FCoE the basis for the UF that underlies its Unified Computing and Data Center 3.0 strategies. That’s a marked strategy shift, considering Cisco spent $250 million on InfiniBand switching firm Topspin Communications in 2005, and upgraded its line of switches with Topspin technology the following year. Industry analysts say Cisco’s backing could push FCoE into a prominent role in the data center.

“FCoE will play a more prominent role in purchasing discussions with CIOs and data center managers,” said Bob Laliberte, senior analyst, Enterprise Strategy Group. “Given the challenging economic climate and that FCoE offers significant savings, these solutions have the potential for rapid adoption in the data center.”

Adviser Articles

Services

Federal Govt

Successes