Data Storage for Neural Network Training

Company: Global Japanese Automotive OEM

Overview: A global leader in automotive design and production required a solution to store sensor data for the training of safety-critical neural network algorithms. The company required a solution with a low Total Cost of Ownership (TCO) as they did not want to expand their support resources. They also had limited space in their data center so they needed a small footprint solution. Ultimately, the company chose to invest in a Quantum storage solution via their Automotive Channel Panel AutonomoStuff LLC even though they had already invested in a Hadoop cluster further underscoring the value that they saw in Quantum.

Challenge:

  • Require quick access to recently collected data from vehicles
  • A desire to keep large datasets for long time periods at a lower cost per TB
  • Tight space requirements for a storage solution due to data center space constraints
  • Required a solution that did not mandate updating the data center HVAC
  • Support for a variety of communication protocols ie Ethernet, iSCSI, and fiber channel
  • Required high-performance support of engineering testing and analytics computing
  • A solution must have a low TCO, both operational and environment, due to constrained resources on site

Solution:

  • Quantum XcellisTM Workflow Director, powered by StorNext
  • 900 TB: StorNext QXSTM-5600 performance storage
  • 2.4 PB: StorNext AELI6 archive storage, 200 slots
  • Total Storage: 3.3PB (all fit in a single 29u rack space)

Customer Benefits:

  • Shared storage simplifies workflow, eliminating storage silos
  • Faster performance support for high-performance computing (HPC) analytics and accelerated development
  • Entire Quantum storage solution fits in 29u rack space
  • Virtually no additional heat in the data center due to tape usage
  • Variety of users and devices are able to access the shared data from several communication protocols
  • Performance and capacity can be added independently for future needs and budget control
  • The solution is future proof as there are always options to non-disruptively expand to the cloud and other storage technologies
  • The Quantum Storage solution was deployed with no additional onsite resourcing required (ie install and forget)
  • Total cost: 1/10 the cost of the alternative solution
advanced-driver-assistance-pedestrian-detection

Autonomous Vehicle Development

Company: Global automotive supplier

Overview: A global leader of electronics for next-generation vehicles, saw its research undergo rapid evolution as active safety systems were proving to reduce accidents and make cars safer. The company recently saw its progress in this important program threatened when data volumes dramatically increased, colliding with the need for faster testing and more sophisticated analysis.

Challenge:

  • Massive data growth created by higher sensor and camera resolutions and more devices per car
  • Data volume increasing from 1PB today to over 20PB in 10 years
  • Fast daily ingest from test vehicles—up to 20TB per car, per day—for engineering access
  • High-performance support of engineering testing and analytics computing

Solution:

  • Quantum XcellisTM Workflow Director, powered by StorNext
  • StorNext QXSTM-5600 performance storage
  • StorNext AEL6000 archive storage

Customer Benefits:

  • Shared storage simplifies workflow, eliminating storage silos
  • High-speed ingest and faster performance support high-performance computing (HPC) analytics and accelerate development
  • Automated data protection without backup software
  • Easier collaboration with international teams using common data sets
  • Cloud-ready—public and private
  • Performance and capacity can be added independently for future needs and budget control
  • Total cost: 10x less than the alternative solution
case-global

NASA Captures Massive Datasets for Science

Company: NASA

Overview: Every day, five of NASA’s initial Earth Observing System satellites beam down terabytes of scientific data about the earth, oceans, and atmosphere. This data is processed and made available to climate researchers around the world using NASA’s Earth Observing System Data and Information System (EOSDIS) Core System—one of the most ambitious data gathering and distribution systems ever conceived. The storage of this irreplaceable information, which spans multiple media types and locations, is entrusted to Quantum’s data management software.

Challenge:

  • Every Day:
    • The EOSDIS system receives 3 terabytes of new data from its five satellites
    • NASA distributes 2 terabytes of data from its four linked data centers to researchers around the world
  • NASA provides hundreds of thousands of global users access to more than 20 million files which together contain over a petabyte of information
  • This data bank is the richest set of information ever collected on the Earth’s climate, therefore it needs to be protected
  • By the end of the decade the system is expected to hold 10 petabytes of data

Solution:

  • Quantum’s Data Management Software
  • Tape Storage: Automated tape libraries
  • Server and Storage Environment: SGI IRIX servers

Customer Benefits:

  • Vendor support
  • High reliability and scalability
  • Support of multiple media types
  • High speed search and retrieval capabilities
  • To the users data appears as though it were stored in a single database
nasa_case-study