dynamic real time complex event processing · 2019-11-26 · data warehouse pda data real-time...
TRANSCRIPT
IBM Software Group
© 2007 IBM Corporation
Dynamic Real Time Complex Event Processing
A Department of Defense (DoD)/Ministry of Defense (MoD) Perspective
Real Time Middleware Seminar 2007
Haifa Israel15 April 2007
������������� ������
© 2007 IBM Corporation�
IBM Software Group
Agenda� Defining the Challenge – Problem Space
� CEP Consideration � Static Case� Dynamic Case
� Introduction to Real–Time� Why real–time?� What is “real–time?” What does it mean? In what context?� How is it possible to characterize real–time systems?� Real–Time and a programming model
� Dynamic RT CEP� SCA Model� Architectural Consideration
� Use Case Scenario – Military� Review the Challenge
© 2007 IBM Corporation�
IBM Software Group
DoD/MoD Industry Strategic Goals� To Achieve Net Centric Warfare and Operations, four key domains must intersect
� Physical: Traditional Warfare Domain where forces are moved through time and space. It includes physical assets and IT systems
� Information: Domain where information is created, manipulated and shared. This facilities the sharing of information
� Cognitive: This is the mind of the Warfighter and what drive Effects based Operations
� Social: Is the necessary elements of human enterprise. It is where human interact, exchange information, form shared awareness and understanding to make collaborative decisions
Social DomainCollaborative Decisions
Cognitive DomainCognitive Advantage
Information DomainInformation Advantage
Shared Awareness
Physical DomainForce Advantage
Position advantagePrecisionForce
CompressedOperations
Speed and Access
Plan, Organize, d\Deploy,Employ and Sustain operations
ConveyedCommander’s
intent
NCWO
InformationAge
Warfare
© 2007 IBM Corporation�
IBM Software Group
DoD/MoD Industry Strategic Goals 2010 :Net–Centric Operations Warfare (NCOW)
� Has become the dominate initiative since September 2001� Resulting from changing battlefield; asymmetric warfare is the order of the day, not classical military to military clashes
� Processes and policies must be recast and redesigned to drive effects–based operations� Need to contain adversaries to prevent their ability to conduct operations.
� Underlying this is the need for diverse systems and military resources to have the ability to interoperate and share information, resources, and services in a secure and deterministic manner.
� The key governing principles are:� Fight First for Information Superiority
� Access to information: Shared Awareness
� Speed of Command and decision making
� Self Synchronization
� Dispersed Forces : Non–contiguous operations
� Demassification (use information to achieve desired effects with a smaller force in a specific geography)
� Deep sensor reach
� Alter initial conditions at higher rates of change
� Compressed operations
© 2007 IBM Corporation
IBM Software Group
Challenge – Defining the “Problem Space”� Inability to manage latency and determinism across widely dispersed nodes for predictable
end–to–end processing
� A highly fluid and dynamic ecosystem: fixed event–producing and event–consuming nodes coupled with transient nodes, have bounded availability
� How to first tolerate, then exploit the value of these transient nodes into the enterprise, achieving time critical operations / situations
• Dynamically discover nodes and services• Register (Pre-register at initialization, or leverage tooling for on–the–fly registration)• Prioritize with respect to other transient nodes• Ensure temporal and causal interlock• Compose and integrate into an appropriate CEP flow / domain• Dynamically reconfigure run–time behaviors to support the workflow• Recognize duration and availability of boundaries / limits• De-allocate transient node and dynamically reconfigure the network
� Challenge: How to think / view the problem at different levels of the stack (vertical versus horizontal)
© 2007 IBM Corporation
IBM Software Group
Rivet JointRivet Joint
AWACSAWACSGlobal Hawk
Multi-INTGlobal Hawk
Multi-INT
Sensors /Shooters
Sensors /Shooters
GPS+GPS+
SBR
SBIRS+SBIRS+
UCAVUCAVPredatorPredator
TankerTanker
RT Dynamic CEP – Integration in the LargeRT Dynamic CEP – Integration in the Large
MC2AMC2A
FIAFIA
AirliftAirlift
TALCE/MSTTALCE/MSTCAOCCAOC
AMEAME
AFFORAFFOR--FF
ARGUS ARGUS FAMILY UGSFAMILY UGS
ASOCASOC
CRC/BCSCRC/BCS
TACPTACP
WOC/EOCWOC/EOC
AF DCGSAF DCGS
SOFSOF
TACCTACC
SPACEAF SPACEAF AOCAOC NIMANIMA
NSANSASPACESPACE
C2C2
WeatherWeatherC2C2
AF DCGSAF DCGSCAOCCAOC
AFFORAFFOR--RR
IOSAIOSA
Joint Force with common situational awareness, common operating picture, informed decision making
Land ComponentLand Component
Maritime Component
Maritime Component
� Massive numbers of sensors of all kinds�Unstructured data
�Fixed as well Transient event producers and consumers
� Massive numbers of “actuators” of all kinds
� Massive (potential) connectivity�Many partly, or rapidly changing dynamically–connected edges
� Many human beings, organizations in decision loops
� A nested collection of classic control systems..
� An Internet-Scale Control System that is a large stochastic process
© 2007 IBM Corporation�
IBM Software Group
Agenda� Defining the Challenge – Problem Space
� CEP Consideration � Static Case� Dynamic Case
� Introduction to Real–Time� Why real–time?� What is “real–time?” What does it mean? In what context?� How is it possible to characterize real–time systems?� Real–Time and a programming model
� Dynamic RT CEP� SCA Model� Architectural Consideration
� Use Case Scenario – Military� Review the Challenge
© 2007 IBM Corporation�
IBM Software Group
Event Processing Agent : Basic Capabilities
EPASource Consumer
EPA
Source
Source
Support Components�Registration�Discovery�Monitoring�Linkage
Consumer
Event Producer or Sources: (Sensors, IT generated, Business Process/application output)
Event Consumers: Other systems, EPA’s, Application/Business processes
Event Processing Components: (EPA’s and endpoints that could be applications as part of an overall business process)
Supporting Components: Event Processing Network, External service for data enrichment, Monitoring and management (i.e. Event Broker)
� First generation event–driven systems: based on predefined events, node–to–node relationships
� Performed relatively simple event pattern detections with various transformation capabilities:
• Aggregation• Translation• Composing• Splitting
� Systems had fixed end–to–end Event Processing (Complex or Simple) workflows
• They were data deterministic• They were not necessarily latency deterministic
� Not standards based
• None were defined and vendors created their own� Were predefined as part of their respective registration and system development processes
© 2007 IBM Corporation�
IBM Software Group
Event Processing Agent : Second Generation� Second generation event processing technologies must be based on a set of industry defined and
adopted standards and definitions.
� Technology convergence has pushed the event processing community to realize Standards are critical.
� Will still be static albeit more capable and standards based.
� May or may not support real time ( Latency determinism).� Unable to dynamically discover, integrate, and enable event–producers and event–consumers
that are not fixed in a specific infrastructure/enterprise.
Invoke SOA service
EPA = Event Process Agent
EPAEvent ChannelSource
Situational Awareness
Business Sub-process
Activity
Service (Data Retrieval)
EPA
Source
Source
EPAComplex
Event ComplexEvent
ComplexEvent
Enterprise Service Bus
�Registration�Discovery�Monitoring�Validation�Mediation�Linkage
�Validation�Enrichment�Transformation�Routing�Error handling
Event Stream
Event ChannelEvent Channel
Event Channel
Event Channel
Point-to-PointPublish/Subscribe
© 2007 IBM Corporation��
IBM Software Group
Event Processing : The Future
Command & Control
FixedNode
TransientNode
FixedNode
FixedNode
TransientNode
TransientNode
Event Processing BasedSOA Enterprise
� Future EP enterprises will need to support data and latency determinism across a continuum of performance capabilities for fixed and transient event nodes.
� Abstract the complexity of Event Processing into a RT Enterprise Service Bus (ESB)
� Augment and enrich predefined fixed Complex Event Processing (CEP) flows with transient event producers and consumers via dynamic discovery
� Pre-register event sources for future use in a CEP process flow
� Enable and support rule–rich and processing–rich, intelligent–event–agent nodes that can recognize newly active event nodes and reconfigure rules engine
� Leverage caching, knowledge modeling to accelerate system response
� Couple node–to–node and end–to–end Quality of Service (QoS) policies tightly
� Support robust, end–user tools
• Enable on-demand, run–time generation and registration of new event nodes• Enable workflow composition
© 2007 IBM Corporation��
IBM Software Group
Basic Event Atomic Structure
MediationRules
Events(1)
Out Comes
Event Process(EP)
Events (n)
© 2007 IBM Corporation��
IBM Software Group
Event Processing Agent EPA
F(r)
CEP(i=1)
)],..1(),(),([ xnxmniE ��
)],..1(),1(),([ xnxmiE ��
CEP(i=n)
ArgumentsPassedxnxm
TrackHistoryCausalEventiz
IDEventi
ZuluStampTime
txExnxmiziE
dtRftxECEP
_),...1(___)(
_)()(_
),()],..1(),(),(,[
)(),(
�
�
�
�
�
��
�
�
��
Complex Event Processing Agent EPA Message Arguments
F(r) = EPA Rules
© 2007 IBM Corporation��
IBM Software Group
Eve
nt B
roke
r
Event Topic (1)
Example of Static Complex Event Processing Flow
Validation
QoSB � CLatencyJitter
FilteringA � BA ��B
Enrichment
ComputationalEnhancement
PatternRecognition
Transformation
Fusion
Aggregation(A + B +C)
Routing
Pass Through
Hold and Store
Event Driven ESB
F
Event Topic (2)
Event Topic (i)
CEPUsers
CEPUsers
EPA 1
EPA 2
EPA 3 EPA 4
F
F
© 2007 IBM Corporation��
IBM Software Group
F(r1)
EPA1; First Order EPA2: Second Order
)],..1(),(),1(),([ xnxmiziE ��)],..1(),1(),([ xnxmiE ��
)],..1(),(),([ xnxmniE ��
CEP(i=n)
CEP(i=1)
CEP(i=n+1)
F
F
T
T T
F
F(r1)
F
F(r2)
F(r2)
Select alternate Rules set
Based on Transient persistence
Select alternate Rules set
Based on Transient persistence
Example of Dynamic Complex Event Processing Flow
Dynamic RT Event Driven ESB
© 2007 IBM Corporation�
IBM Software Group
Dynamic RT CEP Process Flow
T=0
Pre-registration
Transient Event Node Pre-registration parameters
� Event ID number
� Events QoS Policy
� Interface requirements/protocols
� Assignment to a predefined Community of Interest (COI)
� Capability (process, transform, route, enrich)
� Relationship to other fixed or transient assets in COI (Inputs, Outputs)
� Discovery when asset interrogates network with proper authentication, access rights and security level
Discovery/registration
Dynamic� IP Address Interrogation ( Are you
there )
� IP Address Authentication ( Do you have permission )
� Search for pre-registration event component/dossier
� Gather Temporal Information (Geo-spatial, duration )
� Compute Area of Operation (AoR) assignment
� Start Duration clock
� Notify CEP Domain Community of Interest (COI) of transient node(s) availability
� Dynamic reconfiguration of node-to-node or end-to-end Complex Event Processing flow QoS with special attention to alternate rules set within Event processing agents
Manual� Manual Registration of undefined
assets:
� Authentication, Access privileges
� Interface/protocol definition
� Assignment to COI and relationship
� Capability
� Dynamic assignment to a COI’s
Active Transient Event(s) enablement
� Notify event agent of Transient event availability/activation
� Switch rules engine
Active Transient Event(s) dis-enablement
� Notify event agent of Transient event Termination/availability
� Switch rules engine
Reconfiguration CEP Composition
� Define the processing states for an event
� Define acceptable outcomes
� Define anomalies, and response scenarios
© 2007 IBM Corporation�
IBM Software Group
Real-Time Event Driven Component
State - (State full/Stateless)
Import/Export Parameters
Data Type
Function – Internal process or action F(r)
FundamentalCapability
New function for Dynamic Real-Time
�Event Flow
�Process Flow
�Mediation Flow
�Individual QoS�End to End QoS�EPA Node rule selection based on transient input activation period
Component Component
Component
Component
Real-Time Event Driven Composition of
Component
Real-TimeRun-time
Deployment
RT Event Driven Architecture SCA Model Abstraction
Realization
Persistence time
Function F(r) is selectable based on transient input discovery/activation period
Quality of Service (QoS) Policy
© 2007 IBM Corporation��
IBM Software Group
Agenda� Defining the Challenge – Problem Space
� CEP Consideration � Static Case� Dynamic Case
� Introduction to Real–Time� Why real–time?� What is “real–time?” What does it mean? In what context?� How is it possible to characterize real–time systems?� Real–Time and a programming model
� Dynamic RT CEP� SCA Model� Architectural Consideration
� Use Case Scenario - Military� Review the Challenge
© 2007 IBM Corporation��
IBM Software Group
ESB
Real-time in event-driven architectures Where does it fitEvent Driven Architecture• Augments transactional business process and
application server within SOA framework
• ESB enriched to gather and detect significant patterns (temporal, spatial) in event data
• Delivers detected events to relevant processing applications – in both push and pull modes
• Significant advance beyond current request /message at a time routing and delivery by ESB
Sensors, RFID readers…
Control nets,actuators
Application history, data warehouse
PDA data
Real-time Events
• EDA supports the integration of real-time, near real-time and enterpriseapplications
• Timeliness and dependability new,required QoS in nodes and in the ESB toprocess events in time
• Time domains differ across EDA layers
• e.g. industrial automation, voice/mediaswitching, network-centric warfare
< 10ms
1-2s
non-RT
Sensing Weapons
Tracking
Command and control
Intelligence Ship systems
Ship applications
Provision-ing
Force coordination
Each ‘domain’ has QoS attributes setting granularity e.g.
• ship applications – enterprise-like, seconds , transactional
• command & control, intelligence, ship systems – soft RT, 100ms-1s
• tracking, sensors, weapons – harder RT, 10-100ms
• leaf components – devices, sub 10s hard RT response
Track objects
Assess / authorize
© 2007 IBM Corporation��
IBM Software Group
Real Time Systems Cover A Broad Deterministic Continuum
Embedded
System
Enterprise
Ecosystem
Em
bedd
ed
RT
devi
ces
Onboard Server-class RT Command & Ctrl
Onboard Server-class RT Command & CtrlESB
Events are everywhere and their interaction is a function of your frame of reference ( Big picture enterprise to embedded) and your needs.
They can happen at multiple levels simultaneously
�Different time epochs�(non to hard Real Time)�System Type
�Closed- No external inputs�Open - Bounded�Open - Un-bounded
We are here today
We are building this
© 2007 IBM Corporation��
IBM Software Group
What do we mean by “Real–Time,” exactly?� Repeatable, consistent behavior
� Causes of non-determinism have been minimized or eliminated
� Is a property:� Guaranteed and predictable deterministic completion of an “operation” to event in situ
� Is not necessarily fast, normally an indication of low latency:� Often appears in conjunction with latency requirements� Entire environment has factors for additional consideration
� Recognize that the deadline is guaranteed with predictable results – key!� If not achieved, requires some other action be taken, that some exception be raised
� No single answer� May be “hard” in the sense that violation of a timing constraint is considered an application failure, or
“soft” if timing constraints are simply stringent performance goals
� Consequences for missing a timing constraint may vary in severity (e.g. miss on Service Level Agreement, lost revenue, damage to property, danger to human life)
� Timing windows may vary in magnitude (e.g. from microseconds to seconds)
� Not simply about better performance; About greater predictability of performance� Maxim: “Real–Fast is not Real–Time”� Corollary: “Real–Slow is not Real–Good”� Challenge: strike the right balance between predictability and overall performance
© 2007 IBM Corporation��
IBM Software Group
� Hard �The system must meet its timing constraint on 100% of operations
� A single failure is considered a total application failure
� Soft�Meeting a timing constraint within a predefined accepted performance
parameters (i.e. within 85% or better)
� Hard real time - historically a niche market �Specialized hardware, RT O/S, RT software stack running on dedicated
uniprocessors
�RT Application code such as Assembly, C, C++, Ada
�Typically low latencies requirements - microseconds
What do We Mean by “Real Time” Exactly? Cont’d
© 2007 IBM Corporation��
IBM Software Group
From Seconds to Nanoseconds…
10 s
1 s
100 �s
10 ms
100 ms
1 ms
10 �s
STRATEGY
TACTICS
COORDINATION
ACTUATION
SENSING
MODULATION
SIGNALING CUSTOM HARDWARE
PE
RC
EP
TIO
N
RE
AC
TIO
N
CO
GN
ITIO
N
Incr
easi
ng s
oftw
are
com
plex
ity
Software complexity vs. Time domain
© 2007 IBM Corporation��
IBM Software Group
What are Real Time Goals?
� Program must complete a task in or by a certain time deadline and with a predictable degree of certainty (and, latency is additive)
� Time deadline (Time resolution or granularity)� 300ms GUI (graphical user interface)
� 100ms Network-based transaction
� 10ms multimedia
� 1ms “Real Time” (often physically-based) tasks
� Degree of certainty� 90% Interactive response (Word processor)
� 99% Soft Real Time (Web service)
� 99.9% Firm Real Time (multimedia, telecom)
� 100% Hard Real Time (flight control, radar/sonar loop)
© 2007 IBM Corporation��
IBM Software Group
Metronome Programming Abstractions
10 s
1 s
100 �s
10 ms
100 ms
1 ms
10 �s
RealTimeThread with Heap
NoHeapRealTimeThreadwith Scopes
© 2007 IBM Corporation�
IBM Software Group
Using Java for Real-time Development
� Incremental path to real-time Java
RTSJ + (NHRT) Non-Heap Real-Time
RTSJ (heap) + Realtime garbage collection
Std Java + Realtime garbage collection
Std Java + Object Pooling
Standard Java
Realtime (<5ms)Realtime (10sec – 5ms)Non-Realtime
Direction of faster response time
© 2007 IBM Corporation�
IBM Software Group
WRT Architecture
Javajar
J9 VM Real-time GC
Metronome
Java SE 5.0 Libraries
Real-time feature (“-Xrealtime”) Standard J9 component
X86-32 Opertron (4-way SMP)
Real-time Linux (Based on RedHat 4.2)
Bind
Boundjar
AOT
Compile Real-timeJIT
Real-time
library
© 2007 IBM Corporation��
IBM Software Group
Websphere Real Time Product Components
J9 VM Metronome
Real-time GC
Java SE 5.0 Libraries
Standard J9 component
Real-time Linux (Based on RedHat 4.2)
Real-timeJIT
Real-time Java
Libraries
Java Run-time Environment (JRE)
Software Development Kit (SDK)
jxeinjar
RT- Aware
Debugging tools
RT Class
Source
© 2007 IBM Corporation
IBM Software Group
Linux and Open Source Matter to Raytheon
Key Benefits � Open Source Solution that is on track to be adopted by the Linux Mainline�Context switch latency under 25 µs (99.9999% under 20 µs) �Open Real Time Stack: RT Linux- RTSJ & RT GC RT Java- x86 Blades�Reduced Risk and Reduced Total Cost of Ownership
Solution�Real Time Linux – Led by the IBM Linux Technology Center and built on the work of Red Hat and Open Source Community�IBM System x and BladeCenter based solution�Fully preemptive kernel, reducing critical path latencies�Priority inheritance enabled kernel and userspace locking
Challenge�Build a Real Time Linux Operating System that would compliment the RT Java to meet the performance demands of the DDG-1000 program while working with the Linux Community to mainline the enhancements.
DDG 1000Zumwalt Class
© 2007 IBM Corporation��
IBM Software Group
IBM’s RT Run-Time Execution Stack for J2SE Today
RT Programming ModelC++ & WebSphere Real Time Java
NetworkUDP/IP TCP/IP
HardwareOpteron x86-32 4-way SMP Blades,
RTOSRT Linux (RHEL V4U2 with RT patches)
RT MessagingDDS V4.1
INFORMATION MANAGEMENT•ANTs IMDB
Development•Tuning Fork
© 2007 IBM Corporation��
IBM Software Group
Quality of Service Parameters
Defines maximum worst case acceptable latency delay in message delivery from a publisher to a subscriberLatency Budget
Define persistence of transmitted data/message availability to newly joining subscribers. Three possible setting are available:
�Volatile-No past data Kept
�Transient- System will keep certain number of samples determined by resource limits and History Parameters
�Persistent- System will store all past message to either in-in system memory or a database
Durability
Specifies no missed data so data must be retransmitted. It has two sub-parameters:
Reliable- Data retransmission will occur until data is received or until time out occurs
Best Effort- (Default) no data retransmission will occur
Reliability
Specifies how many data samples will be stored for later delivery and is tied to Durability QoS . Can be set to keep last N samples or AllHistory
Minimum rate at which a
�Data Writer or publisher will send data
�Data Reader or Subscriber is willing to wait for data
Deadline
Establishes RT Threads priorities from 1- 28 with 1 being highest:
For RT Java Thread Priorities are as follows:
�NHRT Threads- 1 to 8
�RT Java Threads- 9 to 20
�Regular Java Thread- 21 to 28
Determines if a message/data entity is still alive and is tightly tied to the Life Span QoS. One of three values can be set:
�Automatic: Alive signal sent automatically based on Life Span setting or default is infinite
�Manual by Participant: Entity is considered alive if other entities in Domain are
�Manual by Topic: Publisher is alive if at least one instance of its topic has been asserted by the application
Maximum duration that any one message is valid. After such time data is removed form caches, transient history queues or persistent history queues
Feature
Thread Priorities
Liveliness
Life Span
Destination order
Parameters
© 2007 IBM Corporation��
IBM Software Group
Agenda� Defining the Challenge – Problem Space
� CEP Consideration � Static Case� Dynamic Case
� Introduction to Real–Time� Why real–time?� What is “real–time?” What does it mean? In what context?� How is it possible to characterize real–time systems?� Real–Time and a programming model
� Dynamic RT CEP� SCA Model� Architectural Consideration
� Use Case Scenario - Military� Review the Challenge
© 2007 IBM Corporation��
IBM Software Group
High Level SOA Representation for Dynamic RT CEP
ProcessingRT ESB
Transport
Registry
Orchestration
Aggregation
TransformationMessaging
RT QoS Policies
Services Level Agreements
Interfaces
DataDynamic Discover
Admission Control
Create
Compose
Enrich
RetireRT SOA Based EDA
Monitor
&
Manage
SLA
Alerts
Netw
ork Devices
Multilevel
Security
Com
munities of Interest
Attachm
ent Policies
Label Policies
SOA Composed Business Fabric
Multilevel Secure Presentation Layer
SATCOM
Manual Event
On-Demand
Registration
Composition
Monitoring
Deploy
Terrestrial Wireless
© 2007 IBM Corporation��
IBM Software Group
Agenda� Defining the Challenge – Problem Space
� CEP Consideration � Static Case� Dynamic Case
� Introduction to Real–Time� Why real–time?� What is “real–time?” What does it mean? In what context?� How is it possible to characterize real–time systems?� Real–Time and a programming model
� Dynamic RT CEP� SCA Model� Architectural Consideration
� Use Case Scenario – Military� Review the Challenge
© 2007 IBM Corporation��
IBM Software Group
Moving from Deconfliction to Horizontal IntegrationMoving from Deconfliction to Horizontal IntegrationMoving from Deconfliction to Horizontal IntegrationMoving from Deconfliction to Horizontal IntegrationTime sensitive targets must be shared, distributed, Time sensitive targets must be shared, distributed,
and integrated horizontally and integrated horizontally …… to one and all simultaneously to one and all simultaneously ��
FindersFinders
ShootersShooters
DecidersDeciders
ConnectorsConnectors
© 2007 IBM Corporation�
IBM Software Group
���
Node: ISR1
���
���
���
Node: ISRn
���
���
���
Node: C21C
onne
ctor
Infr
astr
uctu
re
���
Node: C22
�����������
Node: S1
�����������
Node: S2
�����������
Node: S3
�����������
Node: Sn
Dee
p S
enso
r an
d H
isto
rica
l dat
a R
each
�� ���������������
������������
�����������
����������� � ���������
������������ �
Near Real-Time Soft Real-Time Hard Real-Time
© 2007 IBM Corporation�
IBM Software Group
Scenario Description:Unknown Aircraft enters NO-FLY ZONEPre-setup� Air Operations Center (AOC) controls several aircraft – fighters and an AWACS Early warning aircraft
(all are transient nodes)� Command has deployed a defensive missile battery (transient node) forward
� Discover, assimilate transient nodes into AOC domains, based upon persistence in Area of Operation (AOR)� Sensor� Track� Weapons event processing
� Pre–register parameter (s)
Track to Engage Process� Several sensors track unknown target, send track updates; Transients send their location updates to AOC
� Update and compute each transient node’s geo-spatial location for intercept
� Sensor sends:� Alert trigger event� Alert track event (containing the target’s location, altitude, speed and classification)
� Receive alert track and an IFF identification as Foe� System decides to attack the threat, and an intercept asset is sent
• Simple solution: Static assignment according to target location• Complex solution: Dynamic asset assimilation queuing–tasking
� When an intercept asset event and an alert confirmation are received, the intercept missile is fired
© 2007 IBM Corporation��
IBM Software Group
Air Operations Center (AOC)Air Operations Center (AOC)
No-Fly Zone
Unknown
�����!��� ������������� �������������������"���
��# �$�%��&� ���������� '�"�������
Sensor 1AWACS Aircraft
Sensor 2Forward deployed Radar site
Sensor 3Command Center Radar
CommunicationLink
InterceptAircraft 1
Intercept Point
T
T
F
F
T
InterceptAircraft 2
T
Mobile Weapon System 1Forward deployed
© 2007 IBM Corporation��
IBM Software Group
High level Dynamic RT Event Processing FlowUse Case
Sensor 1
Sensor 2Sensor 3
EP 1
TrackData
Sensor 1
Sensor 2Sensor 3
AlertTrigger
EP 2
A1orA2orA3
E[Alert trigger (Ax)]
���������� ���
)321( orAsorAsAsAT ��
)(ATIFFAT �E[Alert Track]
EP 4
Fire Intercept Missile
EP 3 EP 5
ASSET 1
ASSET 2
ASSET n
AssetsThat canIntercept
(FINDERS)
IFF(AT) Trigger
E[Alert Confirm]
Compute Intercept PointAssign asset
E[Alert Track]
E[Intercept Asset (Asn)]
(DECIDER) (SHOOTER)
Total Time from Detect to Engage = 10 seconds
Microsecond ESB
Millisecond ESB
Second(s) ESB
T
T
T
T
TF
F
FF
A2orA3
)32( SSAx ��
© 2007 IBM Corporation��
IBM Software Group
Boost to Terminal Phase Community of Engagement (COE)
Boost
Ascent
Separation
Mid-Course
MIRV
Terminal
ReentryVehicle
Decoy
Battle Battle Management/CommManagement/Command, Control & and, Control & Communication Communication (BM/C3)(BM/C3)
SBIR�HEO�LEO�GEO
Terminal High Altitude
Area Defense(THAAD)
Patriot Missile PAC-3
SM-3
KEITHAAD
PAC-3GBI
Aegis CruiserAegis Destroyer
BMEWSCobraDane
Ground Base Interceptor (GBI)FT Greenly Alaska
Vandenberg AFB CAKinetic Energy Interceptor
(KEI)
Airborne Laser(ABL)
Sea BasedX-Band
COE(Boost)
COE(MID)
COE(Terminal)
MobileX-Band
STSSDSP
© 2007 IBM Corporation��
IBM Software Group
Problem Domain at a High level
� The primary components of what we shall call a Community of Engagement (COE) consist of Sensors, Command and Control and Weapons. These can be further characterized and classified according to:
� Geospatial location
� Flight Profile that they are specifically designed to engage to missile (i.e. Boost & Ascent, Midcourse or Terminal phase)
� Fixed or Mobile (i.e. Aegis Ship, Low earth orbit/High earth Orbit Satellite, Mobile X-band Radar (Land or Sea based) or Missile battery) asset.
� Each of the assets have differing internal and external Quality of Service (QoS) capabilities which must be harmonized and choreographed together to form a COE that meets a specific Service Level Agreement based on the three core attributes above.
� Once a detection event has occurred, our Missile Defense IT Ecosystem must be able to dynamically create a series of COE that track the missile until it is destroyed passing information and getting ready the next COE based on Missile flight path phase as a backup in the event the current COE fails to destroy the missile. In this way COE’s are created, made ready, queued and tasked in event order so as to maximize the probability of kill.
© 2007 IBM Corporation��
IBM Software Group
Problem Domain at a High level - Continued
� When creating COE’s, they will consist of a number of both fixed COE’s where by the IT ecosystem has established QoS and SLA and mobileassets that need to be discovered and dynamically configured into a COE’s work flow process including QoS policies and SLA’s.
� Herein lies a critical architectural design point with regard to real-time latency determinism. Can a geospatially separated set of services some grouped into a COE – Fixed and others that are added dynamically be auto-configured and provide a deterministic latency performance/predictability that satisfies that COE’s flight profile. Typically the three critical inhibitors to determinism are:� Bandwidth/Communications
� Memory
� CPU processing
© 2007 IBM Corporation��
IBM Software Group
Problem Domain at a High level - Continued
� With these three key parameters (Bandwidth/Communication, Memory, CPU processing) how do we architect an SOA/ESB IT infrastructure that can meet the Deterministic latencies required to intercept the Missile.� However, if we examine the end to end Missile engagement process it is a very large
Stochastic Process whereby there is only a probability of achieving RT Determinism
� The most critical constraint is provision in the face of either high data ingest or faults
� How do we degrade gracefully in this situation…. How do we trade off Urgency of processing versus importance
� How will your system perform in the presence of overload and resource starvation ( Shed Load, lockup, etc.)
� It is not good enough to just reduce priorities or make something a higher priorities, Resource management and provisioning play crucial role here this goes for memory, CPU and BW). Also not all processes can be arbitrarily stopped some provide critical admin and support services that must continue to execute otherwise we run the risk of catastrophic system collapse
� This problem is not unlike a futures market whereby you ensure against risk by over provisioning
� We work well on a per single node bases but when executing multiple JVM’s on a single CPU resource provision and scheduling become more complex.
© 2007 IBM Corporation��
IBM Software Group
ESB deployment Patterns Which one is best for BMD
ESB
SRVC
SRVC
SRVC
SRVC
Global ESB/SOA
ESB
SRVC
SRVC
Directly Connected ESB’s/SOA’s
ESB
SRVC
SRVC
ESB
SRVC
SRVC
Brokered ESB/SOA
ESB SRVC
SRVC
ESB
SRVC
SRVC
ESB
ESB
SRVC
SRVC
Federated ESB’s/SOA’s
ESB
SRVC
SRVC
ESB
SRVC
© 2007 IBM Corporation��
IBM Software Group
Questions?