Western Digital Corporation · NASDAQ:WDC
Western Digital — SC18 (Supercomputing 2018) Birds of a Feather session
26 slides from Western Digital's conference presentation, Nov 14, 2018
Slide 1: Exascale Archiving: Challenges and Opportunities
Exascale Archiving: Challenges and Opportunities Wednesday, November 14th, 2018, 12:15p to 1:15p D221
Slide 2: NG Archive – LANL, Gary Grider
NG Archive LANL Gary Grider LQ e Los Alamos NATIONAL LABORATORY EST.1943
Slide 3: How are archives changing given advances in technology and automated workflows like AI/ML and IOT?
How are archives changing given advances in technology and automated workflows like AI/ML and IOT? • More data read ( example read everything from beginning each year and reprocess while you grow ) • At LANL, single data set sizes ( only good if you have all of that data in memory at once) is approa
Slide 4: Thoughts on key archive requirements and cost considerations?
Thoughts on key archive requirements and cost considerations? • Scale out md with search but probably not flat/Sharded • 100s GB/s or more bw • If active devices dominate, solve random bathtub, spatial correlated, and non- spatial correlated failure efficiently • Extreme scale Erasure over non activ
Slide 5: What do you think long term data storage in future data centers will look like?
What do you think long term data storage in future data centers will look like? • Likely will have a bw tier or at least landing zone • Likely will have a lower less agile tier • Have all the above requirements met • Far more searchable and extendable metadata than today • Probable more automated su
Slide 6: Spectra Logic – Matt Starr
SPECTRA Spectra Logic Matt Starr
Slide 7: How are archives changing
How are archives changing • Active and “in-active” archives • Write once read maybe • 6-10x read to write ratio • Time to last byte • It is no longer how long it takes to get to the first byte, but rather to complete the whole job. • New data sets from new customers: • IOT, smart car, AI…..edge base
Slide 8: Next generation archives
Next generation archives • Object based • The whole idea of putting it away • Intelligent search • Customer generated • Auto Generated • Resilient • Distributed • Future proof • Open standard • Cloud technologies • Infinite scale (database, queues, users…) • Distributed, second site
Slide 9: Future archives
Future archives • New words • Zetta, Yotta, Xenotta • New Data sources • 20 Autonomous Vehicles generates 60 TB per week With full Diags on • High resolution microscope • Specialized cloud • Bio IT • HPC • M&E • Tape is not going away • Rotational Disk may be….
Slide 10: BOF SC18 – Stefaan Vervaet
BOF SC18 Stefaan Vervaet Sr. Director Solutions & Alliances November 14th 2018 ©2018 Western Digital Corporation or its affiliates. All rights reserved. 11/13/18 !10
Slide 11: Digital Transformation Increases the Value of Data
Digital Transformation Increases the Value of Data Requires Data be treated as a Competitive Weapon Data as a Record Data as Communication Data as Efficiency Data as Currency Richness Value
Slide 12: Machine Learning Driving New Insights
Machine Learning Driving New Insights +34% GPU-driven Application (#) is up +63% Deep Learning Institute +80% Developers (#) * of object storage is deployed on- 2012 – 2017* premises or in a hybrid form ** New Apps Require New Data STOP GPU cycles stranded by STOP STOP Local Flash Drives Infrastruct
Slide 13: Data Center Systems
Data Center Systems FAST: Low Latency at Scale BIG: Capacity at Scale OpenFlex F3000 Fabric Device and E3000 Fabric Enclosure ActiveScale Object Storage ActiveScale X100 ActiveScale P100 High-performance, low-latency Highly Durable and High-Throughput device for Fast Data: artificial System as a rep
Slide 14: Data Center & Server Evolution
Data Center & Server Evolution From Towards Data Center • Concrete floor (not raised) • Hot / Cold aisle separation • Spine / Leaf networks • Modular containers • Cooling centralization • Hydroelectric locations • Hyperscale (7.2M sqft 650 MW) Server Shared Composable Each Server Contains: Resources
Slide 15: Broadest Storage Leadership
BROADEST STORAGE LEADERSHIP Data Center Solutions ©2018 Western Digital Corporation or its affiliates. All rights reserved. Confidential.
Slide 16: ©2018 Western Digital Corporation or its affiliates. Investor Day 2018 all rights reserved.
©2018 Western Digital Corporation or its affiliates. Investor Day 2018 all rights reserved.
Slide 17: Bruce Gilpin – Versity
Bruce Gilpin VV Versity
Slide 18: The Future of Archiving Technology
The Future of Archiving Technology • Evolving software market dynamic • Accelerating commoditization of storage • Golden age of archiving software
Slide 19: Evolving Software Market Dynamics
Evolving Software Market Dynamics • Tightly coupled to loosely coupled • Closed to open
Slide 20: Accelerating commoditization of storage
Accelerating commoditization of storage • Driven by the size of archives and data growth rates • Driven by cloud and hybrid cloud options • Tape vs. object/cloud feedback loop
Slide 21: Golden age of archiving software
Golden age of archiving software • More data + more storage options = more to manage • Ever larger and more centralized collections of metadata = more to manage • Overall opportunity to create value for archive users is increasing
Slide 22: Jim Gerry – IBM / HPSS
Jim Gerry Senior Architect IBM / HPSS
Slide 23: How are archives changing given advances in technology?
How are archives changing given advances in technology? • Advances in data gathering technologies • Archives are growing FASTER as “data gathering technologies” become higher resolution, faster, and cheaper. • Advances in data storage technologies o HDD are being squeezed by NAND on one side and tap
Slide 24: How are archives changing given advances in automated workflows like AI, ML and IOT?
How are archives changing given advances in automated workflows like AI, ML and IOT? • Cause data storage costs to increase • Data has value • AI, ML and IOT will help you gain insight • Help reduce the growth rate of data • AI, ML and IOT will help you decide what data are important so you can thro
Slide 25: Thoughts on key archive requirements and cost considerations?
Thoughts on key archive requirements and cost considerations? • Access frequency • Ingest and recall rates – files/second and GB/s • Year to year capacity expectations • Redundancy • Data integrity and data revalidation • Expiration • Workflow and policy integration • Metadata
Slide 26: What do you think long-term data storage in future data centers will look like?
What do you think long-term data storage in future data centers will look like? • Tape for long term storage at scale with existing technologies to solve the more difficult requirements: • Tape striping • Tape striping with rotating parity (RAIT) • Efficient tape drive, tape media and tape library u
Western Digital — SC18 (Supercomputing 2018) Birds of a Feather session on 6ix