Download pptx - Big Data: IT does matter

Transcript
Page 1: Big Data: IT does matter

1

Big Data: IT does matter Dr. Sven Niedner, JettainereMail: [email protected], Twitter: S_N_I

London, November, 11th 2013

Page 2: Big Data: IT does matter

Still true?

Page 3: Big Data: IT does matter

Who we are and why IT matters for us

The leading ULD management companies Reliable partner for 13 airlines wordlwide 100% owned by Lufthansa Cargo

ULD fleet acqusition and lease back Savings potential enabled by intelligent asset tracking Over 60k air cargo ULDs – largest fleet in the world Active control concept (forecast-based) Zero offloads since …

Role of IT: IT products are a differentiator Jettware software for ULD management Used at 350+ stations worldwide Covers full ULD fleet lifecycle

Page 4: Big Data: IT does matter

Our customers

ULD management customers + JettLease short term lease

Our USP: 20% fleet size reduction due to our active control concept

Page 5: Big Data: IT does matter

Where are my ULDs?

Where do I have to reposition my ULDs?

How will the demand develop at an airport?

Will I survive the next holiday season without any

offloads?

Is my fleet to big?

Whom can I borrow ULDs from? (Smart Pooling!)

Business questions

Page 6: Big Data: IT does matter

EASY Where the hell is that overstock coming from? How many ULDs did not move in the last X days?

HARDER How did my mean time between repair develop? How did the demand develop in the last year before

christmas?

VERY HARD Has there been a similar situation as today occured

before and how was it resolved? I want to introduce this crazy new KPI, but what was

it‘s value last year?

Innovation is key for success – your Business questions will change

Access to real time data

Access to historic data

Link real-time data and historic

data

Page 7: Big Data: IT does matter

Jettware overview

TacticalStrategic Operational

Jettware: UCMs, LUCs and stock checks at stations

2005

UCD Cockpit: Stock monitoring & ULD positioning/move requests

2008

JettStart: Demand simulation and capacity planning

2011

Revised ULD Tracking Application & DWH reporting tools

2012

Management Cockpit: Overall performance monitoring

2012

JettApp: Digital ULD tracking for the ground handler

2013

Page 8: Big Data: IT does matter
Page 9: Big Data: IT does matter
Page 10: Big Data: IT does matter

<PARENTHESIS TOPIC=“mobile“>

Page 11: Big Data: IT does matter

Mobile devices collect more information than ever Not only the NSA, but also your company! Changing user beviour: Using application “on-the-go”, not only

“on premise”

People expect real time response & interactive UIs User expectations are set by Apple and Google Responses need to come within seconds You cannot satisfy your users with inferior quality

You think it’s different because your uses are your employees? Dream on! The pressure will come – from your users Think BYOD – consumer goods set the expectation level! These guys (Apple, Samsung, Sony, …) are fast and aggressive You will have to compete

Real world example: Mobile drives the hunger for realtime Big Data

JettApp device sends ULD messages from the ware house in real time

Increased data quality More information, more

events Tablet PC Running apps Build on Android

Page 12: Big Data: IT does matter

</PARENTHESIS >

Page 13: Big Data: IT does matter

How to do it?

Page 14: Big Data: IT does matter

How to do it

Keep all your data. Discard nothing.

Page 15: Big Data: IT does matter

Architectureis important

Page 16: Big Data: IT does matter

Challenges

l implement like:query = function ( all data )

Page 17: Big Data: IT does matter

http://www.flickr.com/photos/jblndl/3326092093/sizes/l/in/photostream/

Keep all your data. Forever.

Page 18: Big Data: IT does matter

30.000 ULD movements per day, 100.000 messages per day (mostly SITA)

Business is driven by data:

ULD inventories

stocks at stations

flight plans and changes

ULD histories back to 2005

20 TB+ (and growing)

>10.000.000.000 data sets

Input mostly SITA messages: varying data quality and compliance

… and our ULD control desk needs to act in real time

So where is “Big Data”?

Page 19: Big Data: IT does matter

Challenges for IT

It needs to be real big! (cargo ship) + 50 million USD

It needs to be real fast! (drag boat) + 20.000 USD

Conflicting goals high cost (aircraft carrier): 9 Billion USD

Foto by Bengt Nymann, http://www.flickr.com/photos/bnsd/6139750090

Really fast: 30.000 $

Page 20: Big Data: IT does matter

Challenges for IT

It needs to be real big! (cargo ship) + 50 million USD

It needs to be real fast! (drag boat) + 20.000 USD

Conflicting goals high cost (aircraft carrier): 9 Billion USD

Foto by cuxclipper, http://www.flickr.com/photos/36288065@N04/4949196610

Really big: 100.000.000 $

Page 21: Big Data: IT does matter

Challenges for IT

It needs to be real big! (cargo ship) + 50 million USD

It needs to be real fast! (drag boat) + 20.000 USD

Conflicting goals high cost (aircraft carrier): 9 Billion USD

Foto by US Navy, http://www.flickr.com/photos/usnavy/5781645295

Really fast and really big: 10.000.000.000 $

Page 22: Big Data: IT does matter

BatchLayer

ServingLayer

Really clever: The Lambda architecture pattern

Speed layerINPUTData

Views,Reports,

Apps

RealtimeView

BatchView

OUTPUT

OUTPUT

Lambda architecture proposed by Nathan Marz,Twitter

Combining real-time and batch procssing:

Page 23: Big Data: IT does matter

Really clever: The Lambda architecture pattern

Design principles: Human fault tolerance Data immutability Recomputation

BatchLayer

ServingLayer

Speed layerINPUT Data Views,Reports,

Apps

RealtimeView

BatchView

OUTPUT

OUTPUT

query = function(all data)

Page 24: Big Data: IT does matter

JettStart: Demand simulation and capacity planning

UCD Cockpit: Stock monitoring & ULD positioning/move requests

Jettware: UCMs, LUCs and stock checks at stations

Management Cockpit: Overall performance monitoring

ULD Tracking Application & DWH reporting tools

JettApp: Digital ULD tracking for the ground handler

Focus: historic data

Different applications have different focus

Tactical OperationalStrategic

Focus: real-time

Recommender System

Focus: historic data

Page 25: Big Data: IT does matter

http://www.flickr.com/photos/alreadytaken/2136780794/

Build a mental model of you business domain

And then map it on your existing data structures

Page 26: Big Data: IT does matter

Think beyond the obvious

Go for second order effects: Network effects Correlations non-linear phenomenons

Page 27: Big Data: IT does matter

Enhanced Business domain insight

Accept unreliable and inconsistent data as a fact

You cannot fix the quality of historic data

Statistics will help you to detect or cancel out errors

Page 28: Big Data: IT does matter

Have realistic expectations regarding the Data Scientist role

Nice idea in theory, but (almost) impossible to find in reality

Better kiss the frog now than wait for the prince forever

http://www.flickr.com/photos/fcstpauli/2067908929/sizes/l/in/photostream/

Page 29: Big Data: IT does matter

Don’t loose your business focus.

For us it’s all about…

1. Demand vs. Supply

2. Overstock vs. Understock (balance network)

3. Demand forecast and tactical ULD movements

4. Strategic movements

Page 30: Big Data: IT does matter

<BACKUP >

Page 31: Big Data: IT does matter

<PARENTHESIS TOPIC=“agile“>

Page 32: Big Data: IT does matter

The agile manifesto:

Individuals and interactions …… over processes and tools

Working software …… over comprehensive documentation

Customer collaboration …… over contract negotiation

Responding to change …… over following a plan

Implement lean and agile development tactics

Page 33: Big Data: IT does matter

</PARENTHESIS >