A presentation at International Conference on Massive Storage Systems and Technology (MSST) in in Santa Clara, CA, USA by erik riedel
DATA IMPACT ON THE ENVIRONMENT ERIK RIEDEL, PHD CHIEF ENGINEERING OFFICER FLAX COMPUTING MAY 2023 revision 7
Reduce carbon footprint Reduce cost footprint focus on efficiency & results via carbon / performance focus on efficiency & results via cost / performance scope 1 & 2 operational carbon; scope 3 embodied carbon capex, opex, people-ex
THANKS
ERIK RIEDEL AND GARTH GIBSON, “UNDERSTANDING CUSTOMER DISSATISFACTION WITH UNDERUTILIZED DISTRIBUTED FILE SERVERS” 5TH NASA GODDARD SPACE FLIGHT CENTER CONFERENCE ON MASS STORAGE SYSTEMS AND TECHNOLOGIES. COLLEGE PARK, MD. SEPTEMBER 1996.
SCALING STORAGE
PETER CORBETT, JEAN-PIERRE PROST, CHRIS DEMETRIOU, GARTH GIBSON, ERIK RIEDEL, JIM ZELENKA, YUQUN CHEN, ED FELTEN, KAI LI, JOHN HARTMAN, LARRY PETERSON, BRIAN BERSHAD, ALEC WOLMAN, RUTH AYDT, “PROPOSAL FOR A COMMON PARALLEL FILE SYSTEM PROGRAMMING INTERFACE” TECHNICAL REPORT CMU-CS96-193. PRESENTED AT THE INTERNATIONAL CONFERENCE ON HIGH PERFORMANCE COMPUTING AND COMMUNICATIONS (SUPERCOMPUTING ‘96). PITTSBURGH, PA. NOVEMBER 1996.
GARTH GIBSON, DAVID NAGLE, KHALIL AMIRI, JEFF BUTLER, FAY CHANG, HOWARD GOBIOFF, CHARLES HARDIN, ERIK RIEDEL, DAVID ROCHBERG, JIM ZELENKA, “A COST-EFFECTIVE, HIGH-BANDWIDTH STORAGE ARCHITECTURE” CONFERENCE ON ARCHITECTURAL SUPPORT FOR PROGRAMMING LANGUAGES AND OPERATING SYSTEMS (ASPLOS VIII). SAN JOSE, CA. OCTOBER 1998.
ERIK RIEDEL, GARTH GIBSON, CHRISTOS FALOUTSOS, “ACTIVE STORAGE FOR LARGE-SCALE DATA MINING AND MULTIMEDIA” 24TH INTERNATIONAL CONFERENCE ON VERY LARGE DATABASES (VLDB ‘98). NEW YORK, NY. AUGUST 1998.
OBJECTS
MIKE MESNIER, GREGORY R. GANGER, ERIK RIEDEL “OBJECT-BASED STORAGE” IEEE COMMUNICATIONS MAGAZINE 41 (8). AUGUST 2003.
CLOUDS
ERIK RIEDEL “EFFICIENT & CONVENIENT - HOW TO BUILD BIG STORAGE AS A CLOUD”, MSST CONFERENCE PACIFIC GROVE, CA. APRIL 2012.
SCALABLE STORAGE CLOUDS
$/TB $/TB/month $
ATMOS (2008) >1,000 CUSTOMERS >1 EXABYTE DEPLOYED ECS (2013) >>1,200 CUSTOMERS >1 EXABYTE DEPLOYED ATMOS OVER $2B LIFETIME CUSTOMER REVENUE 22.7 drives/U 816x drives 4x servers 2x switches 18x cables
SOSP 2003
CARBON
natural resources carbon footprint demand growth
TECHNOLOGY ADVANCES
Technology Constantly Advances Computing Technology has advanced in leaps & bounds. hardware, software, operations, applications, methods, methodologies Makes sense to regularly consider updates & refreshes.
Slide from my PhD thesis defense in 1999 Today 100G networking PB storage THz computing TB memory
1999 2022 8,640 TB storage 480 spindles 1,152 GHz compute 122,880 GB memory 76,800 W power $650,000 cost
CIRCULAR ECONOMY
https://www.youtube.com/watch?v=PuzEMYCGeVk
https://www.youtube.com/watch?v=PuzEMYCGeVk
https://www.youtube.com/watch?v=PuzEMYCGeVk
https://www.youtube.com/watch?v=PuzEMYCGeVk
$0.45 / $1 $0.35 / $1 $0.25 / $1 $1 / $1 $0.05 / $1 25,000 to 95,000 servers / month from “Open Uncomplicated” - presented at the OCP Virtual Summit in June 2020
REDUCE
Reduce carbon footprint Reduce cost footprint focus on efficiency & results via carbon / performance focus on efficiency & results via cost / performance scope 1 & 2 operational carbon; scope 3 embodied carbon capex, opex, people-ex
EXTEND
Why it works 3 years primary 3 years secondary 3 years tertiary 9 year design lifetime Recertified hardware approach – facilitate secondary and potentially tertiary use stages for technology assets in various forms
Why it works 3 years primary 3 years secondary STORAGE 9 year design lifetime 5 years primary 3 years secondary 5 years primary 3 years primary 2 yrs primary 3 years primary 4 years secondary 3 yrs secondary 4 years secondary 3 years tertiary 10-12 years In fact, anything that keeps technology running longer will be beneficial, as long as the technology is still useful for something by somebody, plus maintainable & serviceable. SO use those actual criteria to evaluate ALL technology: workload performance, ongoing maintenance complexity & ongoing service costs.
How it works (2) Step 3 - Optimize Step 2 - Quantify Step 1 - Audit Audit systems, servers, and applications. Match per-unit, perdevice carbon inventory, performance, and capacity. Report, review, and optimize - changes might be hardware, software, operations, or design. buy new hardware $$- extend life of existing hw $0 re-purpose existing hw $+ add recertified hardware $-
A WIDER LOOK
https://www.npr.org/transcripts/1048050024
GARBAGE WASTE
CLOSE THE CARBON LOOP
https://www.npr.org/transcripts/917060248 Tim & Gabe
Tim & Gabe Tradewater Refridgerant Finders
THANK YOU
Call to Action • Reach out to us to get involved • Engage us to evaluate / quantify your data center, systems, and server carbon footprints • www.flaxcomputing.com • Evaluate your own servers, share the results with us data @ flaxcomputing.com • If you have servers you don’t want any more, send to: Flax Computing, Suite A2 530 West Street Braintree, MA 02184 • If you want us to arrange a pickup instead, contact us at servers @ flaxcomputing.com Erik Riedel, PhD, Chief Engineering Officer, Flax Computing Twitter: @er1p, @RiedelAtWork email: erik @ flaxcomputing.com
CATCH US AGAIN SOON OCP REGIONAL SUMMIT – PRAGUE, CZ – 19-20 APRIL 2023 KUBECON EUROPE – AMSTERDAM, NL – 18-21 APRIL 2023 MASS STORAGE SYSTEMS & TECH – SANTA CLARA – 21-23 MAY 2023 STORAGE DEVELOPER CONFERENCE – 18-21 SEPTEMBER 2023 OCP GLOBAL SUMMIT – SAN JOSE, CA – 17-19 OCTOBER 2023 KUBECON NORTH AMERICA – CHICAGO, IL – 6-9 NOVEMBER 2023
This talk will show that using carbon footprint as a common metric to assess a piece of computing equipment allows straightforward comparison of technologies and designs on a “performance per carbon” basis, bringing together operational (energy use inputs) and scope 3 (production & materials inputs) carbon, along with workload-aligned performance metrics to compare technologies and systems. Our proposed methodology to apply “carbon points” to hardware components and systems can allow system-level, rack-level, and data-center-level quantification of detailed carbon footprints, which can then be optimized and reduced. You cannot improve what you cannot measure, and we believe that carbon footprint can be used today as a successful common metric for comparison. We will outline our database of footprint calculations and comparisons with real OCP systems, and we will review our success in bringing carbon-advantaged computing to OCP deployments in several real customer scenarios worldwide.
for free. You
can too.