Light Reading
Facebook's load balancing is designed to keep its servers running at a moderate speed -- not too slow.

Facebook Slashes Data Center Power Consumption

Mitch Wagner
8/8/2014
50%
50%

Facebook is developing a new traffic management technology called Autoscale to optimize energy consumption by 10-15% at its data centers.

Facebook currently uses a traditional round-robin approach to load balancing, but found that was less than optimal, because servers running low-level loads use power more inefficiently than idle servers or servers running at moderate or greater loads, writes Qiang Wu, Facebook infrastructure software engineer, on the Facebook Code Engineering Blog.

Autoscale is designed to optimize workloads so that servers are either idling, or running at medium capacity. It tries to avoid assigning workloads in a way that results in servers running at low capacity, Wu writes.

An idle server consumes about 60 watts. It takes a big power hit, to 130 watts, when it jumps to low-level CPU utilization, for a small number of requests per second. But it only takes a small power hit, to 150 watts, when it goes from low-level to medium-level CPU utilization, Wu writes.

Therefore, from a power-efficiency perspective, we should try to avoid running a server at low RPS and instead try to run at medium RPS.

To tackle this problem and utilize power more efficiently, we changed the way that load is distributed to the different web servers in a cluster. The basic idea of Autoscale is that instead of a purely round-robin approach, the load balancer will concentrate workload to a server until it has at least a medium-level workload. If the overall workload is low (like at around midnight), the load balancer will use only a subset of servers. Other servers can be left running idle or be used for batch-processing workloads.

Though the idea sounds simple, it is a challenging task to implement effectively and robustly for a large-scale system.

Autoscale dynamically adjusts the size of the server pool in use, so that each active server will get at least a medium-level CPU load. Servers not in the active pool don't receive traffic.


Power up your data center knowledge on Light Reading's data center infrastructure channel


Optimizing both performance and power consumption was key in developing decision logic for traffic management: "On one hand, we want to maximize the energy-saving opportunity. On the other, we don't want to over-concentrate the traffic in a way that could affect site performance."

Results have been promising:

Autoscale led to a 27% power savings around midnight (and, as expected, the power saving was 0% around peak hours). The average power saving over a 24-hour cycle is about 10-15% for different web clusters.

Normalized power consumption for a production web cluster with and without Autoscale. Source: Facebook.
Normalized power consumption for a production web cluster with and without Autoscale. Source: Facebook.

Facebook is driving open source data center hardware design with its own Open Compute project. The project is self-serving -- Facebook runs among the most massive data centers in the world, and data center cost savings improves Facebook's bottom line. Facebook says it has saved $1.2 billion over three years using the Open Compute hardware designs it champions. (See Open Compute Project Takes on Networking.)

Earlier this week, Facebook bought PrivateCore, a security software company, to beef up its server security. (See Facebook Buys PrivateCore for Server Security.)

— Mitch Wagner, Circle me on Google+Follow me on TwitterVisit my LinkedIn profileFollow me on Facebook, West Coast Bureau Chief, Light Reading. Got a tip about SDN or NFV? Send it to wagner@lightreading.com.

(9)  | 
Comment  | 
Print  | 
Newest First  |  Oldest First  |  Threaded View
SachinEE
50%
50%
SachinEE,
User Rank: Light Sabre
8/12/2014 | 2:23:27 PM
Re: good news for Facebook
With the millions of users that Facebook has bagged over he years, it is good to know that they have found a way to improve their data centers. Facebook is used world wide and they make a lot of profit but this new occurrence will no doubt help them to cut costs and probably boost their earnings by a great percentage. It is a wonder that they did not do this sooner.
brooks7
50%
50%
brooks7,
User Rank: Light Sabre
8/11/2014 | 11:44:17 AM
Re: Not been done before ?
Dennis,

Note, that the applications that people like Amazon and Google do are diverse.  It may not be as easy to eliminate compute power in that environment.  Scaling computing and storage in those environments might have very different curves.

seven

 
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
8/11/2014 | 11:39:05 AM
Re: Not been done before ?
Maybe they have but choose not to crow about it. Or maybe they haven't. The point of the story was to highlight what FB is doing to improve its efficiency and margins.
Whatdoyouwant
50%
50%
Whatdoyouwant,
User Rank: Light Beer
8/11/2014 | 11:31:47 AM
Re: Not been done before ?
I meant by the industry in general.  Google, Yahoo, Amazon, etc..   none of these guys have figured this out ?
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
8/11/2014 | 10:36:51 AM
Re: Not been done before ?
When your growth is in triple digits, you don't worry so much about sweating down costs. That comes with maturity. Now that Facebook is a public company, it has to focus more on stuff like margins.
Whatdoyouwant
50%
50%
Whatdoyouwant,
User Rank: Light Beer
8/11/2014 | 9:46:42 AM
Not been done before ?
Is it me or does this seem like something that would have been done a long time ago ?
danielcawrey
50%
50%
danielcawrey,
User Rank: Light Sabre
8/9/2014 | 6:31:59 PM
Re: Something good from Facebook
I did not realize that the optimum power consumption for a server instance was at medium level capacity. This is really useful - something that hardware designers and cloud service providers should usel to optimize machines and cut costs.

More and more services are being powered through the cloud - which means there is huge opportunity to slash operating expenses. 
thebulk
50%
50%
thebulk,
User Rank: Light Sabre
8/9/2014 | 11:36:47 AM
Re: Something good from Facebook
Just think of the cost they are cutting with this! It really is impressive. 
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
8/9/2014 | 9:01:31 AM
Something good from Facebook
Congratulations to Facebook for figuring out how to make its data centers more efficient. Do you think there's a similar potential for energy cost savings for network operators that deploy virtualization?
Educational Resources
sponsor supplied content
Educational Resources Archive
Flash Poll
From The Founder
It's clear to me that the communications industry is divided into two types of people, and only one is living in the real world.
LRTV Interviews
The New Wave of IP + Optical Integration

11|21|14   |   04:29   |   (7) comments


At the Alcatel-Lucent Technology Symposium, Heavy Reading senior analyst Sterling Perrin talks about how SDN has reshaped the discussion around packet and optical integration.
LRTV Huawei Video Resource Center
Huawei Highlights at BBWF 2014

11|20|14   |   3:40   |   (1) comment


Broadband World Forum is one of the world's largest telecoms, media and technology events with over 7,800 senior executives from across the globe converging on Amsterdam every year to identify the Next Big Thing. BBWF is an exciting place to meet the entire industry under one roof and identify the latest in network innovation, service optimization and customer ...
LRTV Huawei Video Resource Center
How Will BCMS Stimulate Margin for Broadband Operators?

11|19|14   |   6:52   |   (0) comments


In BBWF 2014, Liu Shuqing emphasizes the value of FMC 2.0 based full service experience by throwing light on the BCMS solution. The underlying principle of this innovative technique is to create network robustness and driving network from connection oriented to ACE – BAND oriented infrastructure, in which applications, cloud, and user experiences will be an asset ...
LRTV Huawei Video Resource Center
SingleFAN3.0: Better Connected Experience

11|19|14   |   3:06   |   (1) comment


At the BBWF 2014, David Hu, the VP of Huawei Access Network Product Line, talked about the future of access networks – SingleFAN3.0: faster broadband, wider coverage, and smarter connection.
LRTV Interviews
Basil Alwan Interview: The Road to Cloud

11|19|14   |   09:09   |   (0) comments


Alcatel-Lucent's head of IP and Transport talks about the migration towards a web-like networking environment, the impact of the cloud, SDN and NFV, and the yet-to-be-announced FP4 chip.
LRTV Documentaries
FairPoint Makes a Fair Point About Analytics

11|19|14   |   1:56   |   (1) comment


The US-based communication service provider gets to grips with advanced analytics, tackling data and breaking down the silos within its own business.
LRTV Documentaries
Analytics Lets C Spire Get to Know Subs

11|19|14   |   3:01   |   (2) comments


It's all about the data for US operator C Spire as it uses analytics to personalize its customer service down to individual subscribers.
LRTV Interviews
Nuage Branches Out With SDN: CEO Interview

11|17|14   |   9:32   |   (0) comments


Sunil Khandekar, CEO of Alcatel-Lucent's SDN-focused unit Nuage Networks, talks about the opportunities and challenges of breaking out of the data center into wide-area networks.
Light Reedy
Telecom Analytics Grows Up

11|14|14   |   1:15   |   (4) comments


The big data analytics debate has moved on from a year ago, with some experts suggesting it's no longer a technology challenge.
LRTV Huawei Video Resource Center
Huawei Compass

11|14|14   |   3:17   |   (1) comment


At OpenStack Summit 2014, Shuo Yang, Huawei Principal Cloud Infrastructure Architect introduced Huawei Compass, the software tool for solving customers' problems on the journey of OpenStack Cloud.
LRTV Huawei Video Resource Center
Huawei's Cloud Strategy in European Region

11|14|14   |   2:56   |   (1) comment


At OpenStack Summit 2014, Dr. Gotz, CTO of Huawei IT in European Region introduced Huawei's cloud strategy in European region.
LRTV Huawei Video Resource Center
Huawei's Contribution on OpenStack

11|14|14   |   5:58   |   (0) comments


At OpenStack Summit 2014, Dennis Gu, Huawei Chief Architect of Cloud Computing introduced the relationship between OpenStack and cloud computing, and Huawei's contribution on OpenStack.
Upcoming Live Events
December 2, 2014, New York City
December 3, 2014, New York City
December 8-10, 2014, Reykjavik, Iceland
February 12, 2015, Atlanta, GA
April 14, 2015, New York City, NY
May 6, 2015, McCormick Convention Center, Chicago, IL
May 13-14, 2015, The Westin Peachtree, Atlanta, GA
June 9-10, 2015, Chicago, IL
Infographics
Irish Telecom outlines the rise of VoIP technology, including its adoption within businesses and their perception of its quality.
Hot Topics
Bell Labs Chief Slams 'Toy' Networks
Robert Clark, 11/19/2014
$38.3M: Ain't That a Kik in the SMS
Sarah Reedy, Senior Editor, 11/20/2014
Do You Have a 2020 Vision?
Dennis Mendyk, Vice President of Research, Heavy Reading, 11/21/2014
Google, AT&T, BT Unite on Network Data Models
Carol Wilson, Editor-at-large, 11/20/2014
The New Wave of IP + Optical Integration
Ray Le Maistre, Editor-in-chief, 11/21/2014
Like Us on Facebook
Twitter Feed