Debs 2013 tutorial : Why is event-driven thinking different from traditional thinking

Why is event-driven thinking
different from traditional thinking
about computing?
Presenters: Opher Etzion and Jeffrey Adkins

A year ago, Roy Schulte from
Gartner published a personal
blog entitled “does anybody
care about event processing”
He admitted that his
predictions about the
actual size of the event
processing market is
smaller than predicted
His observation is that
95% of the event
processing market is not
visible since it is home-
built and not labeled as
EP
In this tutorial we will
discuss what is event
driven thinking and how
it is possible to help
organizations to help
themselves in exploiting
events

© 2013, IBM Corporation
Agenda I
Introduction –
Brief History of
Event Processing
in practice
II
The major
differentiation
factor of event-
based thinking
III
The Ontology of
event and event
influence
IV
Anatomy of
reactive systems
V
Pragmatics :
A business
oriented
approach
VI
Summary
6/30/2013 3

Topic I – Introduction & a brief history
of event processing in practice

Event Processing History
First start-ups
Descendents
of academic projects
Apama acquisition
By Progress
Around
2000
2007
TIBCO and
Oracle announce
products
Streambase
Coral8
2005
EPTS
Established
Hitting
the analysts
hype…
2008
IBM
Joins
2012
New players:
SAS. Yahoo,
Twitter
2013
EP at the height
of BIG DATA hype
Cycle
M&A:
TIBCO/Streambase
Software AG/Apama
6/30/2013 5

Events, as “data in motion” is one of the
fundamental ingredients in big data:
event -driven analytics
Event -driven services / making events part of
SOA
Event-based decision making and event
driven Optimization
Event -driven processing as a backbone
of next generation systems: event-based
robotic, autonomic vehicles, human
enhancement technology…
Where are event used today? Virtually everywhere
6/30/2013 6

Event Processing in 2013
There is now an accelerated development of new event-based systems – many
of the current trends in computing are event-driven
Internet of Things
6/30/2013 7

Relatively new players in event and stream processing
Storm
S4
Google
Intelligent events
IFTTT
ON {X}
6/30/2013 8

Big Data Hype Cycle 2012 Source: Gartner publication
G00235042, July 31, 2012
6/30/2013 9
Event processing
Again in the hype
Cycle – in different
context

I want to know about it
immediately and react in the
best possible way
Detect Derive DoDecide
Awareness ReactionSituation
Event Driven Applications follow the 4D paradigm
6/30/2013 10

This is how the event-driven application market looks
This is how the current event processing market looks like…
Source:
Event processing
Manifesto
manual
Build your
Own Use COTS
New segments
Source:
Event processing Manifesto
6/30/2013 11

EPMM – Event Processing Maturity Model
0: unused
2: implicit
3: islands
4: integrated
1: manual
5: strategic
No event
awareness
Subscription to
some events,
manual handling
Events are stored in
databases and are
processed as part of
process oriented
Explicit event processing for some applications as islands;
instrumentation and actions are hard coded and sporadic
Event processing is integrated
with main business processes;
instrumentation and actuators
are well established
Strategic view
of event processing
across the enterprise
6/30/2013 12

Lack of standards:
SOA took off only when WS standards were accepted
Lack of sufficient awareness and good ROI understanding:
Need entry points and methodology about benefits to
individuals, enterprises, packaged applications providers.
Luck of understanding of what is event-based thinking
and how to translate it to implementation
Lack of skills – current tools require
highly skilled developers to do
tricky programming
6/30/2013 13
Barriers for Wider Adoption

Major Gap: Products in this area are geared towards IT Developers
A comprehensive user survey shows
that 84% of the users wish that
event rules could be defined by
business users
There is a
gap
Current models:
Implementation
oriented
Business analysts oriented
Modeling
6/30/2013 14

What makes it difficult to express requirements?
Develop correct application with the right semantics
Observation:
A substantial amount of effort is invested today in many
of the tools to workaround the inability of the language
to easily create correct solutions
6/30/2013 15

Some Correctness Topics
The right interpretation
of language constructs
The right order
of events
The right classification
of events to windows
6/30/2013 16

A simple scenario to demonstrate complexity –
why “native implementation” does not work?
Bid scenario- ground rules:
1. All bidders that issued a bid within the validity interval participate in the bid.
2. The highest bid wins. In the case of tie between bids, the first accepted bid wins the auction
Race conditions:
Between events;
Between events and
Window start/end
6/30/2013 17

A simple scenario to demonstrate complexity –
why “native implementation” does not work?
Bid scenario- ground rules:
1. All bidders that issued a bid within the validity interval participate in the bid.
2. The highest bid wins. In the case of tie between bids, the first accepted bid wins the auction
===Input Bids===
Bid Start 12:55:00
credit bid id=2,occurrence time=12:55:32,price=4
cash bid id=29,occurrence time=12:55:33,price=4
Bid End 12:56:00
===Winning Bid===
Trace:
Race conditions:
Between events;
Between events and
Window start/end
6/30/2013 18

Positively looking: what is the main challenge to resolve in order to
accelerate the use of events?
6/30/2013 19
Advancing the technology is always helpful - it is happening,
yet it is not the main challenge.
Organizations are not interested in technology
for technology’s sake alone
The next frontier: change the thinking
start with the business need and then get to the IT side.

Summary of topic I:
In the business level – it is not well understood how to
think in events and how to utilize events
In the application level – the life-cycle of event-based
systems require skilled IT developers
Next - we explain what is different about event-driven
thinking
6/30/2013 20

Topic II – the major differentiation of
event-based thinking

Deposit Transfer abroad
A simple example: event-based Anti money Laundering
Based on events:
Identify suspicious
AccountsAn account is suspicious if any of the
following patterns are satisfied:
1. Frequent big cash deposit
2. Frequent cases of a big cash deposit
followed by transfer abroad
3. Lack of account activity
4. Increasing amounts of deposits
6/30/2013 22
Compliance
officer

Characteristics of Event-Driven Scenarios
Events trigger action
Events influence logic
for the results
There may be multiple
events whose combined
content influences the
results
Temporal contexts
(15 days) influence the
results
I want to know about it
immediately and react in the
best possible way
A suspicious account is detected whenever there are at least three
big cash deposits followed by transfers abroad in the last 15 days
The next section
provides an
introduction to
event processing
6/30/2013 23

Traditional Thinking
Insert the event into a database; use periodic or on-
demand queries to process the events
The processing may not
be efficient – many of the
requests will not yield results
be effective – the time to
react may be missed
6/30/2013 24

Process Oriented Thinking (EPMM Level 2)
6/30/2013 25

Difficulties in expressing such scenario in traditional thinking
The event-driven vs.
request-driven nature
Effectiveness and
Efficiency issues
The temporal
oriented behavior
The hidden state
handling
6/30/2013 26

Efficiency and effectiveness issues
be efficient – many of the
requests will not yield results
be effective – the time to
react may be missed
6/30/2013 27

Request driven vs. event driven thinking
6/30/2013 28

In daily life we often react to events..
6/30/2013 29

Traditionally programmers are trained to think in a request driven way
Searching the web, database queries, use of web
services, use of mobile applications
6/30/2013 30

What are the differences in thinking?
6/30/2013 31
Question Response Driven Event Drive
Why is an action being taken? As a response to a specific
request
Triggered by the fact of a
specific situation
When is an action being taken? When the request is being
processed
Determined based on the
context of the situation
What happens when the
request / event occurs?
A response is always
produced
The event can be ignored,
increment the state, trigger
an internal derive event, or
trigger a situation

Temporal consideration changes everything
The logic is sensitive
to timing of events
A delivery should be
confirmed by the
deadline
to the order of events
The winner in the bid is
the first one who made
the highest bid
Determination by timing
considerations
Driver ranking increase
and decrease are
determined every 20
assignments
What?Why? When?
6/30/2013 32

Temporal consideration changes everything
to timing of events
A delivery should be
confirmed by the
deadline
to the order of events
The winner in the bid is
the first one who made
the highest bid
Determination by timing
considerations
Driver ranking increase
and decrease are
determined every 20
assignments
What?Why? When?
In traditional models
temporal functions are
hand-coded, adding
complexity
6/30/2013 33

Events during rush hour are of
interest, events outside rush hour are
not
Events that occur or don’t occur
relative to a deadline
Is the reported problem already
solved, or is it still open?
6/30/2013 34
Logic sensitive to the timing of events’ occurrences

Who arrived first?
Has the bid arrived while the auction
was still open?
6/30/2013 35
Logic sensitive to the order of events’ occurrences

Determine the status of a patient
based on blood pressure
measurements:
Every 8
measurements
Every 5 hours
6/30/2013 36
The nature of situation is determined by timing considerations

Event
Patterns
Pattern “event1 occurs after event2”
requires keeping state of all
unmatched instances of event1
Handling Hidden State
6/30/2013 37

Summary of topic II:
In many cases – event driven functionality is expressed
using the traditional request-response fashion
Fundamental differences exist between the two
paradigms, and benefits exist in using event-driven
modeling and implementation for certain applications
Next – drilling down to the essence of event driven
thinking
6/30/2013 38

Topic III – The ontology of events and
event influence

What is an event – three views
An event is anything that happens, or is contemplated as
happening.
The happening view
The state change view
An event is a state of change of anything
The detectable condition
view
An event is a detectable condition that can trigger a notification
6/30/2013 40

Events in Linguistics thinking
“Friends, you and me...
you brought another friend...
and then there were three...”
6/30/2013 41
Events that we want to know, can, and
should, be first worked through as
done as a sentence.

Events in Linguistics thinking
6/30/2013 42
Consider the story and write it down.
This captures the essence of the event.

© 2013, IBM Corporation43
Ancient Criterion of Change
6/29/2013
An Object, x, changes if and only if
i. 𝑡ℎ𝑒𝑟𝑒 𝑖𝑠 𝑎 𝑝𝑟𝑜𝑝𝑒𝑟𝑡𝑦, 𝑃,
ii. 𝑡ℎ𝑒𝑟𝑒 𝑖𝑠 𝑎𝑛 𝑜𝑏𝑗𝑒𝑐𝑡, 𝑥,
iii. 𝑡ℎ𝑒𝑟𝑒 𝑎𝑟𝑒 𝑑𝑖𝑠𝑡𝑖𝑛𝑐𝑡 𝑡𝑖𝑚𝑒𝑠, 𝑡 𝑎𝑛𝑑 𝑡’ 𝑡 ≠ 𝑡’ , 𝑎𝑛𝑑
iv. that 𝑥 ℎ𝑎𝑠 𝑃 𝑎𝑡 𝑡 𝑎𝑛𝑑 𝑓𝑎𝑖𝑙𝑠 𝑡𝑜 ℎ𝑎𝑣𝑒 𝑃 𝑎𝑡 𝑡’
(Lombard)

© 2013, IBM Corporation44
Quality Space
6/29/2013
Quality Spaces are sets (S) of simple, static properties {P1,
P2, … Pn} which meet conditions:
(i) If at any time, t, any object, x, has 𝑃𝑖 ∈ 𝑆 then
(ii) If at any time, t’, x doesn’t have 𝑃𝑖, 𝑖𝑡 𝑤𝑖𝑙𝑙 ℎ𝑎𝑣𝑒 𝑃𝑗
∈ 𝑆 𝑤ℎ𝑒𝑟𝑒 𝑖 ≠ 𝑗
We call this a Dimension. An object can have multiple
dimensions. Could be represented by an ERD, enumeration

The Goal
It is the MOVEMENT along this quality space that
constitutes an event.
The Goal:
To become aware of these events,
so that we may react to them.
6/30/2013 45

Awareness Boundary
What
knowledge
is accessible
Awareness
Boundary
The Awareness boundary represents the boundary of an ecosystem’s knowledge of events /
situations. This is the knowledge that an ecosystem uses to understand situations, decide on a
course of action and perform that course of action.
6/30/2013 46

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
There are situations that occur outside of an ecosystem’s awareness boundary to which a
reaction inside the ecosystem is warranted.
6/30/2013 47

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
Sensors
Some situations’ occurrences can be directly detected by sensors or come into an ecosystem
through data feeds or some other instrumentation.
6/30/2013 48

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
Sensors
System is
now aware.
Once detected, a virtual representation of the situation, called an event, exists within the
ecosystem. These event can be used as part of the work of the ecosystem.
6/30/2013
49

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
Sensors
Because we
can’t directly
detect it
we must
derive it
Other situations that occur outside our awareness boundary can’t be directly detected, so we
have to derive that it occurred.
6/30/2013 50

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
Because we
can’t directly
detect it
we derive it
These two situations are indicators that the third situation has occurred and since we have
knowledge of their occurrence, we can use it to derive the third, non-detectable situation.
Indicators
Indicators
6/30/2013 51
Indicated

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
Derive
Mechanism
Derive
Because we
can’t directly
detect it
we derive it
Propose calling relationship
between real world event and
derived awareness of event a
Luckham Relationship
There is a relationship between the real world situation’s occurrence and the virtual
representation. We propose calling it “a Luckham relationship”.
Indicators
Indicators
6/30/2013 52
Indicated

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
Derive
Mechanism
Derive
There are still other situations that cannot be directly detected nor derived. These situations
require a human to observe it and enter it in the system.
Indicators
Indicators
6/30/2013 53
Indicated

Awareness Boundary
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
out side
awareness
Situation occur
outside of
awareness
Derive
Mechanism
Derive
This is actually the most common way situations become known. These too should generate an
event inside the ecosystem to separate out awareness from reaction.
6/30/2013 54

These two situations, when occurred together in a pattern, indicated that the situation on top
has occurred.
Situation of
Concern
6/30/2013 55

There are different criteria of change that may play a part in this pattern indicating the situation
on top occurred.
Situation of
Concern
6/30/2013 56

This pattern also has a probability associated with it which indicates the confidence that the
situation on top occurred.
Situation of
Concern
6/30/2013 57

Luckham Relationship Situation of
Concern
, ,
6/30/2013 58
A particular situation may have more than one pattern that indicates to a certain level of
confidence that it has occurred.

Object – Nouns of our story
6/30/2013 59
Nouns
The things we manipulate, collect, buy, sell, talk with….
It has state and dimensions
It has relations with other THINGS

All participates in the process.
• Actor (hero) is the one who
performs the change
• Helper assists “hero”
• Obstacle inhibits the
process.
These can help tell about
what is happening.
Actors – The Hero, The Helper, and The Obstacle
6/30/2013 60

Observer – Reports what is seen
6/30/2013 61
Observer is not part of the process
There are potentially multiple observers
of the same events
Observer might have a subjective
inaccurate perspective
Observers may not be familiar with the
context

Processes
Some
Noun of
Importance
This noun is
something
we care
about
6/30/2013 62
We have some noun / thing that is
important to the business and we
want to know when it changes.

Processes
Some
Noun of
Importance but
different
A process is
needed to alter
the noun
The noun is altered
Process
6/30/2013 63
Some
Noun of
Importance
This noun is
something
we care
about
A process consumes that noun and
transforms it into something more
useful.
Transforms

Processes
Some
Noun of
Importance but
different
A process is
needed to alter
the noun
The noun is altered
Process
6/30/2013 64
Some
Noun of
Importance
This noun is
something
we care
about
When a noun is changed, it gives of
indicators that can be sensed.
Transforms
Gives off indicators of
the change

Processes
Some
Noun of
Importance but
different
A process is
needed to alter
the noun
The altered noun.
Process
6/30/2013 65
Some
Noun of
Importance
This noun is
something
we care
about
An event is created inside the
ecosystem that is the logical
representation of indicators
Transforms
Gives off indicators of
the change Event
“noun has
changed”
The event is a logical
representation of the
indicators

Process Ownership
666/29/2013
Nature
•Seemly Random
but patterns
emerge
•Impact small to
momentous/
catastrophic
Regulatory
•Due Process
•Rules by Law &
Procedure
•Impact
significant
Competitor
•Little visibility
until in open
market
•Actively hiding
info
•Share Market
Space
Supply Chain
Partner
•Visibility to a
point
•Vested interest
in mutual
success
•Other Customers
Self
•Own Process
•Easy
instrumentation
•Have
Understanding
Processes understood
And available indicators
Processes hidden
And opaque indicators
Process can be owned by one of the five group below.
Which have several characteristics. The spectrum of each
shows how much of the process and indicators are
available or obscured.

Data POV vs. Process POV
676/29/2013
Nature Regulatory Competitor
Supply Chain
Partner
Own
Processes Understood
And Available indicators
Processes hidden
And opaque indicates
Instrumentation Sensors, Market Intelligence
Collectors, Industry
Process, Object self-publish,
transactions, feeds
Tooling Streaming, Big Data, Master
data management
Processing / Transaction flow
D
a
t
a
P
O
V
P
r
o
c
e
s
s
P
O
V
When processes and indicators are understood and
available, we take a Process POV approach. When
they are hidden and opaque, we take a data
oriented approach.

Awareness Engineer
686/29/2013
Nature Regulatory Competitor
Supply Chain
Partner
Own
Starting Point 1. Identifying situations, streams /
datasets available
2. Determine patterns that indicate
situation
3. Iterate over intermediate
patterns until you get to raw
data/events
1. Identifying process triggers in terms
of situation
2. Identify significant nouns, its states
and properties
3. Map situation to noun-state change
4. Instrument in processes code
significant state changes (detect)
5. For all situations that does not
directly map, iterate top-down or
bottoms-up until you connect
(derive)
Data POV Process POV
The Awareness Engineer job is to figure out how we
become aware. Below are starting points for the
awareness engineer based on the Point of View.

Awareness of events – what are the main reasons?
Business
IT
Consumers
6/30/2013 69

Reaction types
Decision first: A decision is needed in order to make reaction; the decision can be simple,
or complex (requiring OR methods)
Actuators: Automatic activation of actuator
Notification in various ways:
Activation of process/workflow/task – manual or automatic
6/30/2013 70

By 2015, 80% of all available data will be uncertain
GlobalDataVolumeinExabytes
Multiple sources: IDC,Cisco
100
90
80
70
60
50
40
30
20
10
AggregateUncertainty%
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
2005 2010 2015
Data quality solutions exist for
enterprise data like customer,
product, and address data, but this
is only a fraction of the total
enterprise data.
By 2015 the number of networked devices will be
double the entire global population. All sensor data
has uncertainty.
The total number of social media accounts
exceeds the entire global population. This data
is highly uncertain in both its expression and
content.
6/30/2013 71

Representative sources of uncertainty
Uncertain
input data/
Events
Source
Malfunction
Thermometer
Human error
Malicious
Source
Fake tweet
Sensor disrupter
Projection
of temporal
anomalies
Wrong hourly sales
summary
Source
Inaccuracy
Sampling
or
approximation
Propagation
of
uncertainty
Visual data
Rumor
Wrong trend
Inference based on
uncertain value
6/30/2013 72

Types of uncertainty in event processing
Runtime
Engine
Definitions
Detected
Situations
Event
Sources
Run Time
Build Time
Events
Rules / Patterns
Situation Detection
Authoring
Tool
Actions
incomplete event
streams
insufficient event
dictionary
erroneous event
recognition
inconsistent event
annotation
imprecise event
patterns
Uncertainty in the event input, in the composite event pattern, in both
6/30/2013 73

Uncertainty handling
Two main handling methods:
Uncertainty propagation
The uncertainty of input events is propagated to the
derived events
Uncertainty flattening
Uncertain values are replaced with
deterministic equivalents; events may be
ignored.
Traditional event processing needs to be enhanced to account for uncertain events
6/30/2013 74

Patternmatching: Sequence(1/3)
Crime report matching
Pattern: Sequence [Suspicious observation,
Crime report]
Context: Location, Crime type
Certainty 0.8
Occurrence time Uni(9:45AM,10:05AM)
Id ‘John Doe’
…..
......
Suspicious Observation
Certainty 0.9
Occurrence time 10:02AM
Id NA
…..
......
Crime report
6/30/2013 75

Pattern: Sequence
Certainty 0.8
Id ‘John Doe’
…..
......
Certainty 0.9
Id NA
…..
......
Crime report
Certainty 0.612
Id ‘John Doe’
…..
......
Matched crime
obs.certainty * crime.certainty * Prob{obs.time<crime.time}
obs.time | obs.time<crime.time
The ‘uncertainty propagation’ approach
Pattern matching: Sequence (2/3)
6/30/2013 76

Pattern: Sequence
Certainty 0.8
Id ‘John Doe’
…..
......
Certainty 0.9
Id NA
…..
......
Crime report
Certainty 0.72
Id ‘John Doe’
…..
......
Matched crime
Occurrence time → percentile(occurrence time, 0.5)
The ‘uncertainty flattening’ approach
Pattern matching: Sequence (3/3)
6/30/2013 77

Summary of topic III:
The ontology of event-based systems is based on
awareness of events – either directly or by derivation
from other events
The awareness is enabler for reactions
Next - we describe the anatomy of event-based systems
6/30/2013 78

Topic IV – Anatomy of reactive systems

Two separate but connected goals:
Awareness and Reaction
Awareness Reaction
Event
Detect Derive Decide Do
6/30/2013 80

Reactions are Events too
Becoming aware of an event and then doing something about it.
Derive
Mechanisms
Single Event
Multiple
Events
Ancillary
Info
May need
multiple iterations
May require addition
reference / state
information
Something we
want to react
to
A Situation
Detect
Mechanism
Feedback
Events
Decide
Mechanism
Do
Mechanism
Event
of
Interest
Order
Should be
indicating
Entity State
Change
Should be
indicating
Decisions
Derived Event Trigger
6/30/2013 81

Detect
Some
Noun of
Importance but
different
The act of bringing into a system’s sphere of understanding knowledge about an event.
A person recognizes the change and enters it into some system. This is
the classic case! This is the most flexible because humans are ingenious.
6/30/2013 82

Detect
Some
Noun of
Importance but
different
A sensor senses the indicators and creates the corresponding event in
the system.
6/30/2013 83

Detect
Some
Noun of
Importance but
different
A data feed or systemic interface allows events to be published into the
system.
6/30/2013 84

Detect
Some
Noun of
Importance but
different
SwimLane
TriggerEvent
Activity
State
Change
When the processes under the system’s control makes changes to nouns
it should published these changes as events.
6/30/2013 85

Detect
Some
Noun of
Importance but
different
SwimLane
TriggerEvent
Activity
State
Change
As connected things are becoming more self-aware of their inner-
workings, they can publish their own state changes.
6/30/2013 86

Derive
The act of becoming aware of events that are not directly detectable by bringing together
events with other events, data, patterns and publishing the observation as a derived event.
Raw
events
Raw
events
Raw
events
A Person recognizes the pattern and enters the derived event or just
reacts to it directly. Shown a lot of time by dashboards and analysis.
6/30/2013 87

Derive
Raw
events
Raw
events
Raw
events
A Neural network processes the various inputs and determines a new
situation expressed by a derived event.
6/30/2013 88

Derive
Raw
events
Raw
events
Raw
events
A software applies pattern matching over multiple events and data to
find derived events.
6/30/2013 89

Derive
Raw
events
Raw
events
Raw
events
The most common place is hidden inside of every day system’s code.
6/30/2013 90

Decide
The act of determining the course of action to do in response to the situation. This
includes the background information needed to be collected to make the decision.
Pass through: Sometimes there is no decision. There is only one course
of action.
6/30/2013 91

Decide
Manual Decision: Many times the ecosystem asks a person to decide
the course of take.
6/30/2013 92

Decide
Automated Decision: Algorithmic decision via a decision management
system.
6/30/2013 93

Decide
Automated Goal Oriented: Algorithmic decision via a decision
management system that seeks a optimizing quantitative goals.
6/30/2013 94

Do
The act of performing the course of action that was decided upon.
Notification: Sending a signal of sort to either a person or system. This
would include calling a web-service or subscription to alerts.
6/30/2013 95

Do
Manual Action: This is an order for a human to go do an action.
6/30/2013 96

Do
Applying Actuator: cause a action or setting change on an actuator.
6/30/2013 97

Do
Trigger process: Execute a process or potential a single action.
6/30/2013 98

Richard Hackathorn’s Response Time Latency
Value in terms of Competitiveness decreases
6/30/2013 99
Source: Richard Hackathorn – Active data warehouse, from nice to
necessary, Teradata Magazine, 2006

4D Version of Response Time Latency
Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Latency Decide latency Do latency
Queue ActQueue Act
6/30/2013 100
Converting Richard Hackathorn’s flow
to the 4D Prospective.

Detect Latency
Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Queue ActQueue Act
6/30/2013 101
The time it takes for a ecosystem to
detect either the business event or
indicators that can be used derive the
event.

Derive Latency
6/30/2013 102
Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Queue ActQueue Act
The time necessary to combine the
indicators, events, ancillary data or
human analysis to derive situations
that can’t be directly detected.

Decide Queue Latency
6/30/2013 103
Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Queue ActQueue Act
The time waiting for someone
(manual decisions) or something
(automated decisions) to make a
decision.

Decide Act Latency
6/30/2013 104
Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Queue ActQueue Act
The time it takes to decide the
course of action in reaction to the
situation.

Do Queue Latency
6/30/2013 105
Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Queue ActQueue Act
The time waiting for someone
(human actor) or something
(machine actor) to start executing or
orchestrating the course of action.

Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Queue ActQueue Act
Do Act latency
6/30/2013 106
The time it takes to execute the
course of action.

4D Version of Response Time Latency
Value
Investment
Value
Realization
6/30/2013 107
Until the course of action is completed, everything is a value investment. After the course
of action is completed, value can be realized.
Business Event
Detected
Derived
Decide Started
Decide Completed
Do Started
Do Completed
Value
Time
Detect
Latency
Derive
Queue ActQueue Act

Business velocity – a key competitive edge
6/30/2013 108

Summary of topic IV:
An event-driven system consists of the 4D architecture
Latency reduction is the key to business velocity
Next we’ll present basic ideas about the business oriented
approach
6/30/2013 109

Topic V – pragmatics – a
computational independent model for
event-based systems

ROI
Business
Goals
Application
characteristics
Current state in the maturity model
Approaching Event Processing in an Enterprise
6/30/2013 111

Business
Predictability
Enable the support of event-driven adaptive business
processes
Business Agility
Business
Optimization
Make faster and better (manual or autonomic)
decisions based on timely multi-source information;
ROI
Business
Goals
Application
characteristics
operational
support
proactively seeking and adapting to patterns that
might indicate an emerging event (threats and
opportunities)
Increasing level of automation and thus increase
productivity (e.g. of back office) and reduce cost
Continuous audit ensures timely handling of
violations.
Quantify the impact of:
Compliance with
Regulation
Business
Velocity
Increase business velocity by faster response to
opportunities and early detection of threats
Calculating the benefits of event processing
6/30/2013 112

Situation
awareness
Enables the business logic to be context
sensitive
Context
sensitive
Real-time
dissemination
Taking advantage of information whose value
decreases in time.
ROI
Business
Goals
Application
characteristics
Fast change
The application includes sense and respond to
events, and situation awareness is key
requirement
Enables fast deployment of new versions when
the business logic dynamically changes
Situation
determination
complexity
Complexity stems from one or more of: event
rate, quantity of event sources, state and
context handling, event order sensitivity..
Quantify the level of relevance to each candidate application:
Return on Investment
6/30/2013 113

EPMM: Event Processing Maturity Model (IT View)
0: unused
3: islands
4: integrated
1: manual
5: strategic
No event
awareness
Subscription to
some events,
manual handling
Events are stored in databases and
are processed as part of process
oriented
Explicit event processing for some applications as islands ;
instrumentation and actions are hard coded and sporadic
Event processing is integrated with main
business processes; instrumentation and
actuators are well established
Strategic view
of event processing
cross the enterprise
2: implicit
6/30/2013 114

Next Step: Develop Multi-Step Method to Select Entry Points
Improve Business
Agility
Improve Business
Optimization
Select Goals Prioritize Goals
1 Improve Business
Agility
2 Improve Business
Optimization
Analyze the possible entry point
Applications based on their characteristics
Select applications that best fit
these goals
Produce plan of action
Based on the maturity model
By entry point, map both
satisfaction of business
goals and fitness
characteristics for best
ROI.
Liquidity
Management
Anti Money
Laundering
. . .
6/30/2013 115

Vision:
Shift Governance from Programmer to Knowledge Worker
Governance occurs
through development
and maintenance of
program code.
Governance occurs
through development
and maintenance of
event models.
TODAY TOMORROW
CODE LEVEL
6/30/2013 116

Process Oriented View
6/30/2013 117

Next step – event processing language (ESPER)
// Big cash deposit
insert into BigCashDeposit
select * from Transaction where amount > 100,000 and
transaction_cash_deposit_indicator =’Y’
// Frequent (At least three) big cash deposits
create context AccountID partition by accountId on Transaction;
Context AccountID
Insert into FrequentBigCashDeposits select count(*) from
BigCashDeposit having count(*)>3;
// Transfer abroad
insert into TransferAbroad select * from Transaction where
transferabroad_indicator =’Y’
// Frequent cash deposits followed by transfer abroad
Context AccountID
insert into SuspiciousAccount select * from pattern [
every f=FrequentCashDeposit -> t=TransferAbroad where timer.within(15
days)]
6/30/2013 118

Modeling according to the concept
computing principles
The application logic should be expressed
by a semantically declarative, directly executable,
implementation independent, and rigorously
structured knowledge model
Knowledge
Model
Automatic
translation to
code in regular
or specific
engine language
Free of
implementation
assumptions
Rigorous
verifiable
structure with
all
connections
Represented as
a collection of
tables
The term was
coined by Mills
Davis in 2012
6/30/2013 119

TEM Logic Specification
Frequent big cash deposits
Row # Context Event Logic
When Partition By Regular Multiple events Conclusion
Conditions Conditions
Expression Start End Account ID Count(Big cash
deposit)
Frequent big cash
deposits
1 last 15 days same > 3 is Derived
6/30/2013 120
Lack of account activity
Row # Context Event Logic
When Partition By Regular Multiple events Conclusion
Conditions Conditions
Expression Start End Account ID Transaction Lack of account activity
1 last 10 days same is Absent is Derived

Next step – model view
Big cash deposit
______________________________________
- - - - - - - - - - - - - -
transaction cash deposit indicator
transaction amount
(customer threshold)
______________________________________
Always Customer ID
______________________________________
Big cash deposit
- - - - - - - - - - - - - -
______________________________________
Last 10 days Account ID
Frequent deposits followed by transfers abroad
______________________________________________
Deposit followed by transfer abroad
- - - - - - - - - - - - - -
_________________________________________
___________________________________________
- - - - - - - - - - - - - -
Transaction is absent
___________________________________________
Suspicious account
________________________________________________
Frequent deposits followed by transfers abroad
- - - - - - - - - - - - - -
________________________________________________
Always Account ID
Derive Suspicious
account
Deposit followed by transfer abroad
______________________________________
Big cash deposit
- - - - - - - - - - - - - -
Transfer abroad
___________________________________________
deposit
occurrence Account ID
Bank
transaction
system
Transaction
Transfer
abroad
Transaction
Compliance
officer
6/30/2013 121

Concepts of
Facts
Actors
Events
States
Event Derivation
Logic Transitions
GoalsIT elements
Glossary Logic
Computation
Logic
6/30/2013 122

Summary of topic V:
Systematic approach in an enterprise based on ROI and
maturity models -the business world is not interested in
technology but in business outcomes
Business user oriented modeling as a key solution point
Next we’ll present basic ideas about computing
independent model
6/30/2013 123

What is the main take away from this tutorial – 1/3?
6/30/2013 125
The exploitation of events is a game changer in the
universe
In the business level – it is not well understood how to
think in events and how to utilize events
In the IT level – most of what can be functionally
thought as event processing is implemented using the
traditional thinking within application code

An event-driven system consists of the 4D architecture
Latency reduction is the key to business velocity
6/30/2013 126
The ontology of event-based systems is based on
awareness to events – either directly or by derivation
from another events

Systematic approach in an enterprise based on ROI and
maturity models -the business world is not interested in
technology but in business outcomes
Business user oriented modeling as a key solution point
6/30/2013 127
Next step: help the organizations help themselves with
realizing the benefits of events

Presenters
Opher Etzion Jeff Adkins
IBM Research
Haifa, Israel
opher@il.ibm.com
IBM GBS
Washington, DC
jmadkins@us.ibm.com
Thanks for
listening

Debs 2013 tutorial : Why is event-driven thinking different from traditional thinking

More Related Content

Similar to Debs 2013 tutorial : Why is event-driven thinking different from traditional thinking (20)

More from Opher Etzion (20)

Recently uploaded (20)

Debs 2013 tutorial : Why is event-driven thinking different from traditional thinking