active technologies - hrl · active technologies - hrl © 2005 ibm corporation adaptive systems in...
TRANSCRIPT
Active Technologies - HRL
© 2005 IBM Corporation
Adaptive systems in the Finance Industry
Opher EtzionHaifa Research [email protected]
© 2005 IBM Corporation
The adaptive world
�A business situation occurs that requires change in the businessbehavior
�The business has to decide what is the best way to adapt�The decision may have real-time constraints�The decision should effect the running system
© 2005 IBM Corporation
Analyze
Adapt
Deploy
Model
Monitor Run
IBM Business Performance Management
© 2005 IBM Corporation
Some scenarios
� Highly customized subscription service
� Real-time decision systems
� Automation of exception management
Notify me when my entire stock portfolio is down 2% since the beginning of the day
When there is a significant change in the mix of web transactions in a contact center –recalculate the routing policies to agents
If a platinum customer issues outstanding credit request and there were no credit changes within the last month and the credit gap is less than 10%, then cascade (increase the customer credit)
© 2005 IBM Corporation
Enabling technology –Complex Event Processing
Domain
experts
ERP
FFOO
CO
CO
50403020100
1020304050
120100
80
0
20
20
40
60
6040 Dynamic routing
context sensitive decision support
Dashboard
© 2005 IBM Corporation
CEP embedded in a message flow (WBI- Message Broker)Build time:� Rules & conditions are composed
using a wizard-like editors
Runtime:� Events of different format and
from multiple sources are fed into the Broker and passed to the SituationManager node
� The CEP engine analyzes incoming events & situations based on current rules & conditions
� Upon detection the situations are sent out as messages
� Situations are also fed back into the CEP engine and treated as incoming events
© 2005 IBM Corporation
Situation Manager – Sample flowWhen situations are detected, a message is posted to an MQ queue
Detected Situations
o Message 1 arriveso Copy of Msg 1 loaded to
CEP Engineo Message 1 propagated
unchangedo Message 2 arriveso Copy of Msg 2 loaded to
CEP Engineo Situation Duplicate
message is detectedo Situation Message created
and sent to appropriate queue
o Message 2 propagated unchanged
© 2005 IBM Corporation
Intelligent Filter – Sample flow
Situation
o Message 1 arriveso Copy of Msg 1 loaded to
CEP Engineo Reply – “no situation
detected’o Message 1 propagated to
‘out 1’ terminalo Message 2 arriveso Copy of Msg 2 loaded to
CEPT Engineo Situation Duplicate
message is detectedo Reply – Situation detectedo Situation name and
attributes appended to message header
o Message 2 propagated to ‘out 2’ terminal
© 2005 IBM Corporation
Transaction Flow Monitoring(with Bristol Technologies)
Alerts:
� Acknowledgements is not consistent with the sending order.
� 15 minutes before a settlement has to close, a message is waiting to be handled by a compliance officer due to insufficient credit.
� No acknowledgement/reject received from the stock exchange within one hour from sending a message.Abnormal number of payments received from a specific bank (account) within the first two hours or business
� Three rejects received within a single working day, for purchase messages of the same platinum client.
� A request has been re-routed three time within different agents
Order Management System Execution Management
Open Order
23
SM CP
validate the security
validate the Party & account
4
Exchange
Transaction ManagementCalculationsFigurationAllocationSetlement Details*
AM = Account MasterCP = CounterpartySM = Security MasterFC = Fees and ChargesFR= Fees and charges rulesSI = Setlement Instructions
AM
Reference DataOperation &Transaction Data
Order1
Investment Manger
Open Order
updates open orders with fills
Request for ExecutionNotification of Completion
Block trade
Message/Transaction/Flow
5
6
7
8Allocations, Net proceeds, affirms , confirms
Confirmation and Allocation
9Fully figure trade
Purchase & SalesClearing
SM3 CP3AM3
Settlement
SI2
10Settlement Instruction and Payment
CP2 FC FRAM2SM2 SI1*
Front-Office
Middle-Office
Back-Office
Bird’s eyes view
© 2005 IBM Corporation
Some Facts About Amit - IBM’s First CEP Solution
� Already available as a product extension of IBM Websphere Business Integration Message Broker
� Also licensed as stand-alone and in OEM mode by multiple ISVs
� Written in pure Java – can run in any environment
� High throughput – can process thousands of events per second in typical applications
� Supports dynamic updates – meta-data is updated when the system is running and becomes effective immediately
� Synchronous and asynchronous modes
� Small footprint – suitable for embedded cases
© 2005 IBM Corporation
ARAD Overview�ARAD (Active Real-time Automatic Decision Engine) is a technology that:
� Automatically finds a set of business process, business level parameters and/or IT related configuration parameters so as to optimize high level business objectives
� Enables On-demand optimization by recognizing when changes occur in the environment that require finding new optimal configuration settings
�The business objectives can be specified using a wide variety of business models. Examples include:
� Business Process models
� Rule Based models
� Economic models such as CLV�The ARAD technology provides methodologies, processes, architectures and
algorithms to provide the above capabilities
© 2005 IBM Corporation
ARAD Existing Application Domains
� Automatic SLA parameter optimization scenario (Autonomic Computing)
� On demand optimization of financial messages in a messaging financial application based on messages (USA Financial Enterprise)
� On demand optimization of stock trading transactions in an eTrading site using different effectors (Autonomic Computing)
� On demand optimization of income generated by a network using network configuration parameters (IBM solution in the Telco industry)
© 2005 IBM Corporation
Scenario – optimization of priority in manual exception handling queue� Real-time decision system� Problem: determine the priority of error repairs by manual agents.� Transactions classified according to : customer type, operation type, $$$ value.Optimization criterion: minimize penalties.
� Approach:The optimization is done whenever a “significant change” in the mix of traffic occurs (“implicit
complex event processing”),Best solution is a given time framework (2 minutes
� The optimization results are compared with the following two policies:
� Uniform prioritization, i.e. assigning the same priority to all messages (MQSeries default)
� A manual assignment created by an intelligent operator, that assigns the priorities according to the penalty amounts (i.e. highest priority assigned to message class for which the highest penalty is paid)
Penalty Amounts
0
10000
20000
30000
40000
50000
60000
70000
80000
90000
Policy Types
Pen
alty
am
ount
s Uniform (default) policy
Manual intelligentsetting
Solution found by ARAD
© 2005 IBM Corporation
Automatic exception handling:
� The problem:� There are many different cases in which a business process behaves
in an “exceptional way”, this includes exceptions in data or process or getting into exceptional state.
� Currently exceptions are handled either manually or by hard-coded ad-hoc programming.
� Handling exceptions is a bottleneck in many systems, resulting in violation SLAs, and substantial manual work
�The solution is based upon the following observations:� In many cases, there is a small set of “thinking patterns” behind the
exception mitigation
� This decision pattern may be context-based.
© 2005 IBM Corporation
Scenario :� Violation : A customer exceeded its credit limit (specified as a constraint : customer.total-spending <=
customer.credit-limit
� Stabilizers interpretation :� Restrict : abort the transaction.
� Cascade : update the credit limit.
� Repair : update the amount to the maximal amount that will not violate the credit limit
� Forgive : record for manual resolution.
� Apply : call user defined function to resolve
� Applications: exception resolving in trade processing, data cleansing.
ExceptionDetector
MQ
Exception Q
Self-healer
Messages Resolved Messages
Unresolved Messages
Messages for Automatic resolution
© 2005 IBM Corporation
IBM Solutions for Event Detection and DistributionBusinessActivity
Monitoring
DB2 II &CICS Event Publishers
WebSphere BIAdapters
WebSphere BIConnect
WebSphereRFID & Telemetry
Solutions
WebSphere BIMonitor
Event Detection
WebSphere BIMessage Broker
WebSphere BIEvent Broker
CEP DetectorNodes
EventDistribution
Complex Event Processing Process Execution
WebSphere BIServer Foundation
© 2005 IBM Corporation
Conclusion
� Adaptive behavior is part of IBM’s “business performance management” strategy
� The IBM Research Lab in Haifa serves as the competence center for complex event processing and real-time business optimization, as such feedback from the field is essential for us.
� More details are available upon request