More Related Content
Similar to Event Processing - The Next Generation; March 2009 (7)
More from Opher Etzion (20)
Event Processing - The Next Generation; March 2009
- 2. IBM Haifa Research Lab – Event Processing
' (
I am wearing three
different hats today
The industrial hat: IBM
(which pays my salary and
in return occupies most of
my time)
Once a week I am
wearing the Professor
hat and teach an event
processing course
The EPTS chair hat – a
catalyst in establishing
this community
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 3. IBM Haifa Research Lab – Event Processing
)* + ,-./
0 12
- ) 3
4 5 6
5 4 6
'# # )
! #) # # #
4
quot; 4 # 7
7# ) 4 4 7 4 ! 8
' 2 ! #& 4 ) !4
#
# 4 # #
0 $ $ $# 3
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 4. IBM Haifa Research Lab – Event Processing
4
What is event processing ? What is it used for?
A quick Introduction second generation
What are the current Research Challenges ?
Survey of recent projects
Summary
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 5. IBM Haifa Research Lab – Event Processing
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 6. IBM Haifa Research Lab – Event Processing
# ) !4
Every business person reacts to events, from time
to time
Well… there are all sort of reactions…
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 7. IBM Haifa Research Lab – Event Processing
We answer
phone calls
everywhere
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 8. IBM Haifa Research Lab – Event Processing
We have to make arrangements when our
flight is delayed
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 9. IBM Haifa Research Lab – Event Processing
7 4
4
I closed the deal with the
I closed Australians
the deal with the
Canadians
We closed two huge deals in a single day,
It is a good opportunity to send all the team to
Las-Vegas
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 10. IBM Haifa Research Lab – Event Processing
' (
9 0 )
3 0 ) 3 4 0
! 4$ 3 )
4 # 0 4 )
5 63 #
) 4 +
:# # !
; < 8 7
7
& # !
9 # !
9 )5!4 4 6 #!)
) 07 4 4 #!) 3$
2 0 3$ # ) # #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 11. IBM Haifa Research Lab – Event Processing
9 4
? $ $ $ 4 $
$#
Event Processing
Event Event
Producer Consumer
Generate and Consume and
may publish react to either
perform operations on
events raw or complex
events
events
' 4 #
8
&=4 2& #4 $
# !4 #
# + 4
CEP
# # )# >
8 7
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 12. IBM Haifa Research Lab – Event Processing
4 4 * 4
Getting the
right
information
in the right
granularity BAM
to the right Information Detect Quick observation into exceptional business
Dissemination
person at Decide behavior and notification to the appropriate
the right Respond people.
time
Diagnose Predictive Mitigate or
Active Diagnostics Processing
problems eliminate
based on predicted events
RTE
symptoms and
Reactions to events are done as part of business
resolve them
transactions – achieving low latency decisions,
and quick reaction to threats and opportunities
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 13. IBM Haifa Research Lab – Event Processing
? )
&
4 ! !4 # 4 ?
# ## #! 8
# ) 4 ! Solution:
$ 2 8 # 8 0# 3
Configurable aggregation rules
# 4 # ) 4 4 #
@ •Thresholds by currency
•Thresholds by time
•Thresholds by
# 4#! # 4# 7 counterparty/client
# 4#! # 4# 4
Transactional Context Complex Event Processing Transactional Context
INPUT Filter and Persist CEP Routing, Output
Transactional Aggregation Database Transactional
JMS Input Queues Transform Orders to Aggregation,
Output Queues For
Temporal Logic Logic Update
Pub/Sub Logic Database execution
Aggregated Order Notification to CEP
Direct passthrough of selected messages
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 14. IBM Haifa Research Lab – Event Processing
; & 4 8 2 9
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 15. IBM Haifa Research Lab – Event Processing
% + #%
A
+
! #
4
4 )
7 , #)
Detect .-B
Decide
)
Respond
+
8 #4
4 0 .- ?
1$
32 ) #! #
4 4 ! # 4
4 4 # 4
> )7 # ) #)
# 7 +
4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 16. IBM Haifa Research Lab – Event Processing
& 9 +
# #4 9 +
,C
7 4 # #4
, 4 4 # ! 8
1-
#4 !
# .C
4
9 #4 # !
! 4
4 quot; 9 +
9; #4 7
! 8 )
4
9 ! 8
$ 4 ##
!
84 7 ,
4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 17. IBM Haifa Research Lab – Event Processing
9D 8
4 #4 , #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 18. IBM Haifa Research Lab – Event Processing
. #, #
. +
$ #$
#2 7
) 4 , # +
) )!4 7 ) ##
$ ! #
4 $ !# #
# $
) # ! #05 63$ * 4 D
&=4 $
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 19. IBM Haifa Research Lab – Event Processing
+
! 4 #
7 8
# #
#4
&4 ;
9 !
!
4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 20. IBM Haifa Research Lab – Event Processing
. C
. C
? + !
#
F
.,
# +
9 ) ., H
# 4 # ..
E .-
, +
, # ! ( F
/
.,
9 +
H
? ; .G ..
+ .1
1 ;,? ./
? ? 1
? ? .C
+
G 9 4
E
.F
,-
.-
+ .E
; 44 ?
.H /
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 21. IBM Haifa Research Lab – Event Processing
' 4
(
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 22. IBM Haifa Research Lab – Event Processing
4
:
?4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 23. IBM Haifa Research Lab – Event Processing
#;
7 5 4 6 7
# # $ @
4 0 # 3 +
#
# <
<
4 22
4 <
<
9
9
! ###
#
!# #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 24. IBM Research
Engineering Oriented Challenges
&4 #
# # 9
4 +
!
# 2 4
9
9
EPN
EPN
Event
EPA Event
Producer
EPA EPA Consumer
Pattern
Event Event
Producer Consumer
EPA EPA
Event Pattern Pattern Event
Producer Consumer
#
© 2008 IBM Corporation
- 25. IBM Haifa Research Lab – Event Processing
?4 #; +
#
2 % #
? # #
?4 # 4
; ! #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 26. IBM Haifa Research Lab – Event Processing
: #;
4#2
+
& 2
4
# $
+
4 $
$
#!4
4#
4
quot; 4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 27. IBM Haifa Research Lab – Event Processing
Business User
View Point
# ! 4 ##
Why a decision is needed?
• Event Occurred Islands of
• Event Pattern has detected tools
• Fact has changed
• An explicit request
• Related to other decision
Implementation
Which information is needed
driven tools
In order to make this decision?
• Current state
• Past events
• Past states
• Future states and events. Gaps
In coverage
How the decision should be
Done?
• Inference from facts and rules
• Match patterns on event history Ability to manage
• Apply optimization tools Large complex
Decision network
Dependent
Decisions
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 28. IBM Haifa Research Lab – Event Processing
#
)# # 4 +
+
) 0 3 EP
) 0 3$ Gaps
) # 0 3
+;4 0 & 3
+
I 7 44 0 3
8 7 0 & 3 BRMS
# # 09 ) 3 Analytics
Most Important
Connection
) 4 4 +
J # Business User Orientation:
J The user should be able to author,
J modify, understand, and audit individual
Computerized decisions, as well as the
J # Relationships among them
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 29. IBM Haifa Research Lab – Event Processing
?4 ;
&
% #
94 #
:
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 30. IBM Haifa Research Lab – Event Processing
9 4 ) & K
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 31. IBM Haifa Research Lab – Event Processing
% quot;
&
&
?4 : ! )
'L 4
) 9 )
4 2
: ) #
quot;
&
#4
4#
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 32. IBM Haifa Research Lab – Event Processing
9) M? 74 7 8
N &!
4
Stratum_1 Stratum_2 Stratum_S
Producer_1 Consumer_1
Agent_11 Agent_21 Agent_S1
Producer_2 Consumer_2
Agent_12 Agent_22 Agent_S2
. … .
. . .
. .
. . .
. .
. . .
Agent_1N1 Agent_2N2 Agent_SN2
Producer_P Consumer_C
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 33. IBM Research
RFID Store Scenario - 3 Stratum Levels
Stratum1 Stratum2 Stratum3
TagRead Filter
ProductAtCashier AtLeast5ProductsOf
Loc= “Cashier”
SameKindReturnedToShelf
ProductExitStore ProductLeftAtCashier
TagRead Filter
Tag Sequence Atleast
Loc = “Exit”
Tag
Reader
Tag
Reader TagRead
Tag Filter AtLeast10ProductsOf
Reader ProductInCart
Loc = “Cart” ProductReturnedToShelf
Reader SameKindSold
Sequence Atleast
Filter
TagRead Loc = “Shelf” ProductOnShelf
Levels Average Average Implementation Input events Derived
input events derived total events total
throughput events rate throughput rate
(event/s) (event/s)
Centralized 1,399 38
First stratum – 4 agents on 21,419 21,418
single node 3 Stratum Levels – second 21,419*4+69 4,485
stratum has the slowest 3*107+
Second stratum – 2 agents on 693 277 performance. Maximal flow 29,904 =
single node performance can be achieved 189,731
by combination of 1 node in the
Third stratum – 2 agents on 29,904 4,485 (upper limit)
third level, 107 nodes in the
single node
second layer, 4 nodes in the
Centralized – all the patterns 1,399 38 first level
are detected by multiple
112 nodes improvement 13562% 11802%
agents on same node (8
agents) Additional node improvement 121% 105.37%
© 2008 IBM Corporation
- 34. IBM Haifa Research Lab – Event Processing
2 9 2quot; 4 +
9 # +
+
quot; 4 ; 4 + 4 991
4 ;
# 8
quot; 4 2; 4 +
991
4 # # 9
@
quot; 4 2 2
; 4 + 9
9
# 2
4
+
@
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 35. IBM Haifa Research Lab – Event Processing
& 2O $
% quot;
&
4 4 # ) 7
! # 4 )# #
! & # )
2 !4 9! 4 4
# 4 ) # )
7 # 8 #7 # #
4 )
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 36. IBM Haifa Research Lab – Event Processing
& 9
% quot;0
& # ,- /
-3
&!
9
) +9 ) 7 8
# ! 0# $ $
4 4
! $ # $ #@3
) ) +9 4 ) $
$ # $ 4 # #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 37. IBM Haifa Research Lab – Event Processing
+ ' 84
0K 4 7 9 3
9 4 # 4 # 4
) +
& 4 ) #
4 4
9 0 3 4 ! ! )=4
; 4 !! # #! 4
)
# 4 # #
4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 38. IBM Haifa Research Lab – Event Processing
9 I
quot; # # 4 4
quot; 4 # ! 7 8 7
7 8 # #
! quot; ! #) #
4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 39. IBM Haifa Research Lab – Event Processing
? ?4 # 7 ?O
I 4
# : K
4 2
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 40. IBM Haifa Research Lab – Event Processing
! quot; # $ %
%
& #
& '
4 +P
4 ,- E
-
#!) # !4
0 7 3
4# 4# # 7 $ 4$
9 # M )
) ! # ) # # $
# M # ) # #
9 ) # 0!4 7 ! # #
4 3$ 8 4 # #
)
% !
# #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 41. IBM Haifa Research Lab – Event Processing
; 4
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 42. IBM Haifa Research Lab – Event Processing
' # 7 # #
quot; & !
' ) 4 # #(
05 4 &4 9 #6$ ! * % 3
4# )! #
7 # 4# # )07 @
3$ ## # #4
7 #! 8 =4 ) #
)$ ! $ # )$ 4 )
# # # 4
0 ) ## 3+ Dquot; 2 # # ! #
! )
0 ) ## 3+ 9 &Q #
4 0 $4 )
= $
4 ) 3@ # ! #
4
! #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 43. IBM Haifa Research Lab – Event Processing
# # 4
quot; 8 # # # # !
7
) =4 # # #+
# # 7 )
#
! ) # # 0 ! !) ' 2
<' 3
2 4 # #0 4 # #3
# # #
' ! # !) 4 # #
0 )3 ##
# #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation
- 44. IBM Haifa Research Lab – Event Processing
' ) +
) 4 !
0 7 7 4# #
R
3
0 ) $
3 5 6 4 )
8 ! 4 7 )@ #7 7 #
! 0 !4
4 #
# 7
# 44 3
' # S $ ! $ $
T #4 7 # !
4 $ #7 # # #
IBM Haifa Research Lab – Event Processing © 2008 IBM Corporation