www.fugro.com
Data Data Everywhere,
Mike Liddell 13th March 2014
UUVs @ OI 2014
www.fugro.com Date
Contents Menu
BackgroundBig Data
What is Generating our Big Data Physical Management of Big Data Optimisation in Data Processing Techniques for Handling Data Conclusions
www.fugro.com Date
Contents Menu
Background
Big Data
What is Generating our Big Data Physical Management of Big Data Optimisation in Data Processing Techniques for Handling Data Conclusions
www.fugro.com
Current Status - Maturing Subsea Vehicle Market
AUV’s moving towards commodity– In most categories there is growing choice of supplier – More manufactures moving into the market
ROV’s at commodity status
www.fugro.com
Current Status - Sensor’s and Payloads
Commodity Market ( Just look around this show…)– Growing richness in range and suppliers of acoustic sensors. – Growing range of Optical and Laser sensors.
– Rapidly growing range Positioning and Navigation Sensors.
Some of the bigger players becoming acquisitive
Some of the manufactures starting to connect the dots in their portfolios from sensors to solutions.
Some vehicle providers tightly integrating sensors to provide a package solution
www.fugro.com
Current Status - Summary
Subsea Platforms are commodities ( just not cheap ones) Rich sensor / payload market
The toolbox is available
Operational Challenges Remain– Experienced People. – L & R ( AUVs)
– Autonomy
The big challenge for the Survey and Inspection contractors is the
efficient handling of all this data. Big Data.
www.fugro.com Date
Contents Menu
BackgroundBig Data
What is Generating our Big Data Physical Management of Big Data Optimisation in Data Processing Techniques for Handling Data Conclusions
www.fugro.com
What is Big Data
http://en.wikipedia.org/wiki/Big_Data
Big data is the term for a collection of data sets
so large and
complex
that it becomesdifficult to process using
on-handdatabase management tools or
traditional data processing
applications
. Thechallenges include capture,
curation, storage, search, sharing, transfer, analysis,
and visualization
. The trend to larger data sets is due to theadditional information derivable
from analysis of a singlelarge set of related data
, as compared to separate smaller setswith the same total amount of data,
allowing correlations to be
found
to "spot business trends, determine quality of research, prevent diseases, link legal citations, combat crime, and determine real-timewww.fugro.com
What Big Data Challenges do we face
Data Management.– Protect – Transport – Deliver
Efficient Data Handling (Processing of the Data) – Rapid Delivery of Results
• More Automation. Less Intervention – Integration of Data Sets
– Temporal Analysis
www.fugro.com Date
Contents Menu
BackgroundBig Data
What is Generating our Big Data
Physical Management of Big Data Optimisation in Data Processing Techniques for Handling Data Conclusions
www.fugro.com
What is our Big Data
Imagery– Digital Video -> HD Digital Video -> 4K – Photo Stills
New Survey and Inspection Sensors – Laser Profilers
– Synthetic Aperture Sonars – MBE Water Column Logging – Optical Sensors
Peripheral Sensors
– Inertial Navigation Sensors – Hydrocarbon Sniffers
– Magnetometers
– Forward Looking sonars – SVP/ CTD
www.fugro.com
Big Data Generators
VIDEO– Typical ROV Inspection 3-4 Camera
– Data Volumes SD per Month DVM 6TB – Data Volume HD per Month DVM 11TB – Growth in move from SD to HD to 4K??? – Move away from Video to Stills? Fast ROV
Photo Stills
– 172GB per day. ( 1 MB Image) DVM 3.4TB – Extensive Processing ( duplications) DVM 12TB
• Each process results in a new image
www.fugro.com
Unique Challenges of Big Data Offshore
Cruise lengths of 30-45 days not uncommon. Typically with limited bandwidth on vessels. Not commercially practical to stream all data to beach.
Port calls are expensive “down time” so need to turn vessel around fast.
Port calls are often in in remote locations. Arctic, West/East Africa. Shipping logistics can be difficult or time consuming.
Long duration uptime. 24/7 acquisition.
Complex data that needs to be rapidly accessible often with random access. – (Jump to the middle of a video file)
How and what data products we deliver to clients
www.fugro.com Date
Contents Menu
BackgroundBig Data
What is Generating our Big Data
Physical Management of Big Data
Optimisation in Data Processing Techniques for Handling Data Conclusions
www.fugro.com
Data Storage Optimisation
RAID array of disks for
redundancy and capacity
Academic Definition:
Redundant Array of
Inexpensive Disks
Manufacturers Definition:
Redundant Array of
Independent Disks
www.fugro.com
Data Storage Optimisation - Speed
Tiered Storage
approach for speed
optimisation.
Layers of disks
optimised for speed or
storage.
Optimised solutions fast
become expensive
www.fugro.com
Data Storage Optimisation - Backup
Solution for DR and
Archive
Replication to another
RAID.
LTO Tape Drives
6.25TB on a tape
www.fugro.com
Data Storage Optimisation - Office
Data Live on Vessel for Weeks or Months Data Live in Office for Longer. Final Report issued. Client Comments Office Data Storage may need to be significantly larger. 100-1000s TB
Disaster Recovery Solutions.
– What if the building burnt down…. – What if the building got flooded…. – What if the power went off….
How will this data be archived for the long term • Medium
• Duration
• (and who pays?)
www.fugro.com Date
Contents Menu
BackgroundBig Data
What is Generating our Big Data Physical Management of Big Data
Optimisation in Data Processing
Techniques for Handling Data Conclusions
www.fugro.com
Data Processing Optimisation
Data is Secure and Backed Up but we need fast access to this data
Often software vendors recommend storing data locally on SSD disks for fast access.
– Non starter in a large Survey Data Centre • Danger of data loss.
• Danger of sync issues. Where is the most recent data
• Single user access. What if it’s a large project and needs multiple user access
– Not a responsible method of managing data – But we still need fast access?
www.fugro.com
Data Access Optimisation - Network
Workstation PCs with
Fast Ethernet
connections to server
Minimised bottle necks
in network to maintain
available Ethernet
connection from PC to
Storage.
www.fugro.com
Data Access Optimisation - Software
Optimised Software
enables cluster
processing
Processing tasks
distributed across
multiple cores on multiple
PC.
Significant time savings
on computational
intensive tasks.
www.fugro.com Date
Contents Menu
BackgroundBig Data
What is Generating our Big Data Physical Management of Big Data Optimisation in Data Processing
Techniques for Handling Data
www.fugro.com
Delivery of Data to the Client
Generally there is a requirement to deliver significant number of data products to the client.
With Inspection activities that has meant dozens of hard disks or RAID arrays full of video files.
www.fugro.com
www.fugro.com
www.fugro.com
Video Deliverables
Both VHS Tapes and Digital Video Files are essentially RAW products Digital Video provides significant improvements for data review and QC
www.fugro.com
Video Deliverables
Digital Video integration with Survey Data tends to have a limited life span and user base.
Event Listing becomes “the product” that is circulated.
– Compact
There is a better way
– Convert the Digital Video into a Geo-referenced Image – Image can be loaded into GIS software
• Desktop or web based
– Product that can be shared within a company, access available to all that require it.
www.fugro.com
Video Processing
• INPUT: Digital Video and
USBL Positioning
• OUTPUT: Referenced Image
for GIS
www.fugro.com
Video Processing
As-Built photo record of a pipeline. Enables Temporal Comparisons
– Layers in GIS
– Automated Classification/ Comparison
Comparisons with other platform image data sets (AUV Stills Camera)
Benefits
– Compact deliverable (relatively) – Optimised for GIS
• Optimised for Web delivery
– Resource that can be widely distributed – Longer life.
• As-Built record
• Annual Comparisons – Spatial
www.fugro.com
www.fugro.com
AUV Image Handling
FliMap. Image and Laser Profiling Solution from Helicopter Very close analogy to AUV
– 50 knots v 4 knots
– 50m altitude v 4m altitude – Still images
– High resolution topographic profile. Very large image datasets
– 2Hz - 24 hours =172,000 Images Optimised Image Handling Tools
– Radiometric and Topographic Corrections GIS friendly outputs. Tiled Images
www.fugro.com
www.fugro.com
Image Handling
Convergence in Products - GIS Tiled Image – ROV – Video Input
– AUV - Still Image Input
Processing Flows that evolve towards minimal human interaction
Challenges Remain
– Computationally intensive.
www.fugro.com Date
Contents Menu
BackgroundBig Data
What is Generating our Big Data Physical Management of Big Data Optimisation in Data Processing Techniques for Handling Data
www.fugro.com
Conclusions
Big Data is here and its only going to get
BIGGER.
Rich sensor market of acoustic and now laser/optical sensors.
Tools are available for acquiring the data but the processing is the big challenge.
Investment in infrastructure for secure management of data is not insignificant. ( Should not be overlooked).
New approaches required for the handling of these large data volumes. Emerging techniques for improved handling of Video and creating new
geospatial products
www.fugro.com