petrophysics and big data by elephant scale training and consultin

23
Petrophysics and Big Data Presented at Annual SPWLA show in Houston, TX Dec. 6, 2016

Upload: elephantscale

Post on 21-Mar-2017

89 views

Category:

Technology


0 download

TRANSCRIPT

Petrophysics and Big Data

Presented at Annual SPWLA show in Houston, TXDec. 6, 2016

Presentation Overview

Part 1

– Cloud and Big Data in Petrophysics

Part 2

– How we teach this at Elephant Scale

Copyright © 2016 Elephant Scale. All rights reserved. 2

Motivation

Born out of last meeting of SPWLA in 2015

In cooperation with Antaeus

Chapter in “Guide to Big Data Applications” published Q1 of 2017

Copyright © 2016 Elephant Scale. All rights reserved. 3

Cloud and Big Data in Petrophysics

Cloud and Big Data in PetrophysicsHow we teach Big Data at Elephant Scale

(c) ElephantScale.com 2016. All rights reserved. 5

Data as oil replacement?

Challenges of technologies and Big Data

Technology is becoming a differentiator

Data is becoming a differentiator

O&G needs an overhaul

”If it ain’t broke don’t fix it” – may not work anymore

6Copyright © 2016 Elephant Scale. All rights reserved.

Predictions (for what they are worth )

Oil and oil-service companies will become

– more of technology companies

Companies like Antaeus will show the way

– (Pierre Jean is my good friend)

Knowledge and learning are paramount

– (We do the teaching/training)

7Copyright © 2016 Elephant Scale. All rights reserved.

Terms and pre-requisites Big Data

– More than can be stored on on computer– But also 3V (Volume, Variety, Velocity)

Cloud– NOT delivery through internet– YES – computing resources with unlimited scalability– Better name: elastic cloud (Price, Performance, Security)

Browser delivery– Ubiquitous, but not in O&G (Standard UI, Security)

O&G specific– Wi-Fi may be unavailable/intermittent/sub-par

8Copyright © 2016 Elephant Scale. All rights reserved.

Advantages Unlimited, centralized storage Modern technologies

– Not seen in O&G but well developed at Google, Facebook, etc. Machine Learning as the killer app for Big Data

Why did AI fail for O&G in the 1980’s?– Technology not deep enough– Outside consultants not knowledgeable in O&G

This should change through education– Side benefit – next slide

9Copyright © 2016 Elephant Scale. All rights reserved.

Side benefit of AI – being popular at parties

10(c) ElephantScale.com 2016. All rights reserved.

What’s going on in Houston

11Copyright © 2016 Elephant Scale. All rights reserved.

How we teach Big Data at Elephant Scale

Cloud and Big Data in PetrophysicsHow we teach Big Data at Elephant Scale

What’s there to learn?

1. How to store your data for archiving

2. How to use the data in real-time

3. How to learn from your data

4. What is deep learning

5. How to do it better in the cloud

13Copyright © 2016 Elephant Scale. All rights reserved.

How to store your data Hadoop

– Storage– Processing– Ecosystem

• Hive, Pig, etc. Benefits

– Centralized storage– Perfect online archiving

Courses for– Developers– Administrators– Business analysts– Architects

14Copyright © 2016 Elephant Scale. All rights reserved.

How to process your data in real-time NoSQL

– Scalable to billions of rows and millions of columns– Good for incomplete real-world data– Extremely fast reads and writes without lockups

Benefit– Perfect log/header store– Flexible data format

Courses– Developers/admins– NoSQL data modelers

15Copyright © 2016 Elephant Scale. All rights reserved.

How to learn from your data Machine learning tools

16Copyright © 2016 Elephant Scale. All rights reserved.

Machine learning and AI

Machine Learning "is an algorithm that learns from data" Usually improves its performance with more data.

Uses statistical / mathematical techniques to build a model

from observed data rather than relying on explicit instructions

“More data usually beats better algorithms”

– Anant Rajaraman said it first (?)

• Amazon Retail Platform (25% US transactions)

• WalmartLabs/Kosmix

• Etc.

17

What is deep learning?

– Neural networks with more than one hidden layer

Rebranded neural net with some twists

Reemerging due to cluster computing and GPU

Steps towards Artificial Intelligence (AI)

Examples (all world titles)

– Facebook Deep Face

– Google Translate

– Google DeepMind playing GO game

– IBM Deep Blue winning Jeopardy

Latest: Deep Learning

18(c) ElephantScale.com 2016. All rights reserved

Modeled loosely after the human brain Designed to recognize patterns Input comes from sensory data

– machine perception– labeling – clustering raw input

Recognized patterns– Numerical– Contained in vectors– Translated from real-world data

Images, Sound, Text, Time series Popular in 80s Fell out of favor in 90s in 2000s as statistical based methods

yielded better results Came back with a vengeance

Neural Networks

19

Deep neural network

20(c) ElephantScale.com 2016. All rights reserved

Our credentials

22Copyright © 2016 Elephant Scale. All rights reserved.

• Thousands of students

• Dozens of clients/channels

• A large variety of course

• Experts who do and teach

Our courses (www.elephantscale.com)

23Copyright © 2016 Elephant Scale. All rights reserved.