Ethernet’s Role in Greening the Datacenter

By Dan Maltbie

March 21, 2008

According to a recent report by the U.S. Environmental Protection Agency (EPA), datacenters across the country consumed 61 billion kilowatt-hours of electricity in 2006 at a cost of $4.5 billion — twice as much as in 2000. That’s more power than is required to operate the nation’s 250 million television sets. The EPA predicts that by 2012, U.S. datacenters will consume 100 billion kilowatt-hours of electricity at a cost of $7.4 billion.

IT managers have long known they pay “twice” for electricity in a datacenter: first to power up the many racks crammed full of IT equipment, and then to cool off all those power-hungry, heat-generating systems. Power is getting expensive, too — up to half the cost of ongoing operations, according to some reports. And being wasteful is no longer politically correct given the tons of carbon dioxide being dumped into a warming atmosphere.

Some organizations face an even more pressing problem: they simply do not have the power or cooling capacity to grow. Gartner estimates that up to half of all datacenters will encounter this problem in this year. And datacenters are scaling ever larger to meet the growing demand for new services and to satisfy the need for more online storage, particularly given the increasing use of video content. Consolidation to fewer and larger datacenters is also driving growth as operators seek greater economies of scale. Many of these new “mega” datacenters are being located, quite consciously and prudently, near sources of readily-available and low-cost power.

The IT equipment in a typical datacenter consumes about half of the total power needed, with the balance going to transforming and distributing electricity, backup power sources, air conditioning, lighting and other requirements. Of the IT equipment, servers and storage are the major consumers of power at 40 percent and 37 percent, respectively, according to Gartner. The remaining 20 percent or so can be attributed to networking, and more efficient networking helps reduce total power demand both directly and indirectly. Direct network power consumption can be reduced substantially from current levels; the power required to network a single server, for example, can be cut from more than 50 Watts in some cases to less than 10 Watts, resulting in considerable savings for a datacenter with tens of thousands of servers. Indirect reductions in power consumption result from unifying the datacenter network to facilitate more efficient server virtualization and storage consolidation.

Mixed Motives

For some, the motivation for greening-up the datacenter is a noble effort to mitigate the effects of global climate change by reducing the facility’s carbon footprint. For others, the motivation has a different green incentive: capital and operational expenditures for the power and cooling that currently represent nearly half of a datacenter’s total cost of ownership, according to the Uptime Institute. Over a two year period, in fact, the total cost to power and cool servers can exceed the original cost to purchase those servers.

Reducing power consumption in datacenters is all about efficiency. And the best way to improve efficiency is to eliminate waste. For example, a dedicated server with direct attached storage (DAS) has a utilization of only 10-40 percent. Virtualized servers and consolidated storage can increase utilization rates to 80 percent or more.

Virtualization and consolidation, however, create new challenges satisfying the datacenter’s networking needs. At a minimum a modern, virtualized datacenter requires both a local area network and a storage area network (SAN). Many datacenters also require high-performance computing (HPC) clusters. Operating two or three networks — typically Ethernet, Fibre Channel and InfiniBand — is inefficient.

Perhaps the most inefficient way to accommodate these separate networking needs is the all-in-one “Swiss Army Knife” switch. These large, expensive and power-hungry general-purpose Layer 2/Layer 3 switch/router behemoths need to be full-featured to be able to support a full spectrum of applications, including many not applicable to the datacenter. The net effect of engineering for maximum application flexibility is power consumption growing to a factor of five or more over solutions engineered specifically — and exclusively — for the needs of the datacenter.

Unification via Ethernet Fabrics

A much more efficient way to unify datacenter networking can be found in a new class of Ethernet switches purpose-built for the datacenter — with nothing more, but certainly nothing less than what is required. These switches strip out the unnecessary and power-hungry functions to reduce physical footprint to one-third the rack space and the carbon footprint to one-fifth the power consumption of general-purpose switches or routers.

Ethernet fabric switches are designed for maximum data throughput, minimum latency, and minimum power consumption. To achieve these objectives, its design emphasizes sophisticated congestion management for lossless, non-blocking throughput, extremely low latency with cut-through switching, high scalability, flexible bandwidth provisioning, and, last but certainly not least, minimal power and space consumption.

Such a unified Ethernet fabric begins with top-of-rack switches that take advantage of Ethernet interfaces available for virtualized server and consolidated storage bays. For example, most servers today have a Gigabit-per-second Ethernet (GigE) interface built into the motherboard or the blade chassis, so the power consumed is trivial — and already accounted for in the server’s power supply. A top-of-rack switch with 48 GigE ports and four 10 Gbps Ethernet (10 GigE) ports, for example, can consume as little as 160 Watts, or 3 Watts per port. This single, power-efficient network can be used for both storage access and inter-process communications among servers.

The top-of-rack Ethernet switches feed a core 10 Gbps Ethernet fabric capable of unifying the LAN, SAN and HPC onto a single switched network. These core switches vary considerably in their capabilities and power consumption. Traditional Ethernet switches are rarely suitable for the core of the datacenter for a variety of reasons. The primary being the lack of scalability beyond the port density of the largest switch at the very core of the fabric. Even in smaller datacenters where 128 or 256 ports might be sufficient, the lack of effective congestion management, along with excessive latency from store-and-forward switching, normally renders traditional Ethernet switches unsuitable for SAN and HPC applications.

Ethernet switches purpose-built for the datacenter overcome these and other limitations to deliver non-blocking, lossless throughput with sub-10 microsecond latencies across a resilient, multipath Ethernet fabric supporting upwards of 4,000 10 GigE edge ports. Just as significantly, the better implementations of these Ethernet fabric switches consume as little as 17 Watts per 10 GigE edge port.

The power required to deliver non-blocking throughput on an Ethernet fabric amounts to only about 8 Watts per GigE-connected server for both storage access and inter-process communications. Impressive indeed. But also readily achieved. Here’s how. To realize fully non-blocking throughput, no server or storage array can be oversubscribed. Using the top-of-rack switch described above, 40 of the GigE ports would therefore be aggregated in the four 10 GigE uplinks, leaving eight of the GigE ports unused. In this configuration, the switch’s 160 total Watts of power yields a per-server consumption of just 4 Watts (160 Watts ÷ 40 servers). At the Ethernet fabric core, each ingress/egress port also requires a corresponding port on both the leaf and spine switches, resulting in a total power consumption of 51 Watts (17 Watts/port x 3 ports). A single 10 GigE ingress/egress port serving 10 servers, therefore, yields a power consumption of about 5 Watts per server (or 51 Watts ÷ 10 servers). In practice, of course, a small amount of oversubscription may be tolerated at the access layer, further increasing the power efficiency of the network.

Other Shades of Green

Not only is a unified Ethernet datacenter fabric green environmentally owing to its efficiency, it’s green economically for other reasons. GigE Ethernet switching is notoriously inexpensive on a per-port basis, and could be considered “free” at the server end where the interface is already built-in. The cost of 10 GigE switches has recently dropped dramatically, as well, to less than $2000 per-port for a datacenter-class switch. Next-generation 10 GigE adapters or NICs are also becoming increasingly more affordable.

The power savings associate with unifying datacenter networking on a purpose-built Ethernet fabric switch even helps defray the cost of the switch itself. Calculations comparing an Ethernet fabric switch with another popular datacenter switch revealed that, at 10¢ per kilowatt-hour (an average rate in the U.S.), the power savings alone over five years pays just over half the cost of the Ethernet fabric switch.

The real cost savings of Ethernet, however, derives from its remarkably low ongoing operational expenditures. The familiarity, simplicity and flexibility of Ethernet combine to make it the ideal choice for unifying datacenter networking onto a single fabric with a single network interface per server and storage array, a single set of network cabling and patching, and a single technology for the IT staff to master and manage. When all of these cost savings are combined, the conclusion is obvious: unified networking on an Ethernet fabric is the logical next step in the greening of the modern datacenter.

Of course, not all datacenters can go all-Ethernet at this time. Some have highly specialized HPC applications requiring continued use of InfiniBand. And many operators will want to preserve their significant investments in Fibre Channel — at least for a while longer. But for datacenters undergoing server virtualization and storage consolidation initiatives, network unification is now an integral part of going green with the advent of Ethernet fabric switches.Dan Maltbie of Woven

About the Author

Dan is a 27-year veteran of the networking industry focused on the design and development of new technology and products ranging from storage switches and IP routers to video encoding systems and Ethernet fabric switches. Before co-founding Woven Systems, Dan served as vice president of engineering at Sanera Systems, a provider of director-class Fibre Channel switches, which was later acquired by McDATA. Before Sanera, he was vice president of engineering for Caspian Networks where his team developed a telecom-class IP router. Previously, he was vice president of engineering at DiviCom, a video encoding system provider which is now part of Harmonic Lightwave. Earlier Dan held engineering and management roles of increasing importance at Hewlett Packard, where he represented the company in Ethernet IEEE 802 standards activities. Dan studied computer science and astrophysics at Indiana University.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

China’s Tencent Server Design Will Use AMD Rome

November 13, 2019

Tencent, the Chinese cloud giant, said it would use AMD’s newest Epyc processor in its internally-designed server. The design win adds further momentum to AMD’s bid to erode rival Intel Corp.’s dominance of the glo Read more…

By George Leopold

NCSA Industry Conference Recap – Part 1

November 13, 2019

Industry Program Director Brendan McGinty welcomed guests to the annual National Center for Supercomputing Applications (NCSA) Industry Conference, October 8-10, on the University of Illinois campus in Urbana (UIUC). On Read more…

By Elizabeth Leake, STEM-Trek

Cray, Fujitsu Both Bringing Fujitsu A64FX-based Supercomputers to Market in 2020

November 12, 2019

The number of top-tier HPC systems makers has shrunk due to a steady march of M&A activity, but there is increased diversity and choice of processing components with Intel Xeon, AMD Epyc, IBM Power, and Arm server ch Read more…

By Tiffany Trader

Intel AI Summit: New ‘Keem Bay’ Edge VPU, AI Product Roadmap

November 12, 2019

At its AI Summit today in San Francisco, Intel touted a raft of AI training and inference hardware for deployments ranging from cloud to edge and designed to support organizations at various points of their AI journeys. Read more…

By Doug Black

SIA Recognizes Robert Dennard with 2019 Noyce Award

November 12, 2019

If you don’t know what Dennard Scaling is, the chances are strong you don’t labor in electronics. Robert Dennard, longtime IBM researcher, inventor of the DRAM and the fellow for whom Dennard Scaling was named, is th Read more…

By John Russell

AWS Solution Channel

Making High Performance Computing Affordable and Accessible for Small and Medium Businesses with HPC on AWS

High performance computing (HPC) brings a powerful set of tools to a broad range of industries, helping to drive innovation and boost revenue in finance, genomics, oil and gas extraction, and other fields. Read more…

IBM Accelerated Insights

Leveraging Exaflops Performance to Remediate Nuclear Waste

November 12, 2019

Nuclear waste storage sites are a subject of intense controversy and debate; nobody wants the radioactive remnants in their backyard. Now, a collaboration between Berkeley Lab, Pacific Northwest National University (PNNL Read more…

By Oliver Peckham

Cray, Fujitsu Both Bringing Fujitsu A64FX-based Supercomputers to Market in 2020

November 12, 2019

The number of top-tier HPC systems makers has shrunk due to a steady march of M&A activity, but there is increased diversity and choice of processing compon Read more…

By Tiffany Trader

IBM Adds Support for Ion Trap Quantum Technology to Qiskit

November 11, 2019

After years of percolating in the shadow of quantum computing research based on superconducting semiconductors – think IBM, Rigetti, Google, and D-Wave (quant Read more…

By John Russell

Tackling HPC’s Memory and I/O Bottlenecks with On-Node, Non-Volatile RAM

November 8, 2019

On-node, non-volatile memory (NVRAM) is a game-changing technology that can remove many I/O and memory bottlenecks and provide a key enabler for exascale. Th Read more…

By Jan Rowell

MLPerf Releases First Inference Benchmark Results; Nvidia Touts its Showing

November 6, 2019, the young AI-benchmarking consortium, today issued the first round of results for its inference test suite. Among organizations with submissions wer Read more…

By John Russell

Azure Cloud First with AMD Epyc Rome Processors

November 6, 2019

At Ignite 2019 this week, Microsoft's Azure cloud team and AMD announced an expansion of their partnership that began in 2017 when Azure debuted Epyc-backed ins Read more…

By Tiffany Trader

Nvidia Launches Credit Card-Sized 21 TOPS Jetson System for Edge Devices

November 6, 2019

Nvidia has launched a new addition to its Jetson product line: a credit card-sized (70x45mm) form factor delivering up to 21 trillion operations/second (TOPS) o Read more…

By Doug Black

In Memoriam: Steve Tuecke, Globus Co-founder

November 4, 2019

HPCwire is deeply saddened to report that Steve Tuecke, longtime scientist at Argonne National Lab and University of Chicago, has passed away at age 52. Tuecke Read more…

By Tiffany Trader

Spending Spree: Hyperscalers Bought $57B of IT in 2018, $10B+ by Google – But Is Cloud on Horizon?

October 31, 2019

Hyperscalers are the masters of the IT universe, gravitational centers of increasing pull in the emerging age of data-driven compute and AI.  In the high-stake Read more…

By Doug Black

Supercomputer-Powered AI Tackles a Key Fusion Energy Challenge

August 7, 2019

Fusion energy is the Holy Grail of the energy world: low-radioactivity, low-waste, zero-carbon, high-output nuclear power that can run on hydrogen or lithium. T Read more…

By Oliver Peckham

Using AI to Solve One of the Most Prevailing Problems in CFD

October 17, 2019

How can artificial intelligence (AI) and high-performance computing (HPC) solve mesh generation, one of the most commonly referenced problems in computational engineering? A new study has set out to answer this question and create an industry-first AI-mesh application... Read more…

By James Sharpe

Cray Wins NNSA-Livermore ‘El Capitan’ Exascale Contract

August 13, 2019

Cray has won the bid to build the first exascale supercomputer for the National Nuclear Security Administration (NNSA) and Lawrence Livermore National Laborator Read more…

By Tiffany Trader

DARPA Looks to Propel Parallelism

September 4, 2019

As Moore’s law runs out of steam, new programming approaches are being pursued with the goal of greater hardware performance with less coding. The Defense Advanced Projects Research Agency is launching a new programming effort aimed at leveraging the benefits of massive distributed parallelism with less sweat. Read more…

By George Leopold

AMD Launches Epyc Rome, First 7nm CPU

August 8, 2019

From a gala event at the Palace of Fine Arts in San Francisco yesterday (Aug. 7), AMD launched its second-generation Epyc Rome x86 chips, based on its 7nm proce Read more…

By Tiffany Trader

D-Wave’s Path to 5000 Qubits; Google’s Quantum Supremacy Claim

September 24, 2019

On the heels of IBM’s quantum news last week come two more quantum items. D-Wave Systems today announced the name of its forthcoming 5000-qubit system, Advantage (yes the name choice isn’t serendipity), at its user conference being held this week in Newport, RI. Read more…

By John Russell

Ayar Labs to Demo Photonics Chiplet in FPGA Package at Hot Chips

August 19, 2019

Silicon startup Ayar Labs continues to gain momentum with its DARPA-backed optical chiplet technology that puts advanced electronics and optics on the same chip Read more…

By Tiffany Trader

Crystal Ball Gazing: IBM’s Vision for the Future of Computing

October 14, 2019

Dario Gil, IBM’s relatively new director of research, painted a intriguing portrait of the future of computing along with a rough idea of how IBM thinks we’ Read more…

By John Russell

Leading Solution Providers

ISC 2019 Virtual Booth Video Tour


Intel Confirms Retreat on Omni-Path

August 1, 2019

Intel Corp.’s plans to make a big splash in the network fabric market for linking HPC and other workloads has apparently belly-flopped. The chipmaker confirmed to us the outlines of an earlier report by the website CRN that it has jettisoned plans for a second-generation version of its Omni-Path interconnect... Read more…

By Staff report

Kubernetes, Containers and HPC

September 19, 2019

Software containers and Kubernetes are important tools for building, deploying, running and managing modern enterprise applications at scale and delivering enterprise software faster and more reliably to the end user — while using resources more efficiently and reducing costs. Read more…

By Daniel Gruber, Burak Yenier and Wolfgang Gentzsch, UberCloud

Dell Ramps Up HPC Testing of AMD Rome Processors

October 21, 2019

Dell Technologies is wading deeper into the AMD-based systems market with a growing evaluation program for the latest Epyc (Rome) microprocessors from AMD. In a Read more…

By John Russell

Rise of NIH’s Biowulf Mirrors the Rise of Computational Biology

July 29, 2019

The story of NIH’s supercomputer Biowulf is fascinating, important, and in many ways representative of the transformation of life sciences and biomedical res Read more…

By John Russell

Intel Debuts Pohoiki Beach, Its 8M Neuron Neuromorphic Development System

July 17, 2019

Neuromorphic computing has received less fanfare of late than quantum computing whose mystery has captured public attention and which seems to have generated mo Read more…

By John Russell

Xilinx vs. Intel: FPGA Market Leaders Launch Server Accelerator Cards

August 6, 2019

The two FPGA market leaders, Intel and Xilinx, both announced new accelerator cards this week designed to handle specialized, compute-intensive workloads and un Read more…

By Doug Black

When Dense Matrix Representations Beat Sparse

September 9, 2019

In our world filled with unintended consequences, it turns out that saving memory space to help deal with GPU limitations, knowing it introduces performance pen Read more…

By James Reinders

With the Help of HPC, Astronomers Prepare to Deflect a Real Asteroid

September 26, 2019

For years, NASA has been running simulations of asteroid impacts to understand the risks (and likelihoods) of asteroids colliding with Earth. Now, NASA and the European Space Agency (ESA) are preparing for the next, crucial step in planetary defense against asteroid impacts: physically deflecting a real asteroid. Read more…

By Oliver Peckham

  • arrow
  • Click Here for More Headlines
  • arrow
Do NOT follow this link or you will be banned from the site!
Share This