Information processing and analytics cannot be focused only on “store-first” or batch-based approaches. To provide maximum business value, information must also be analyzed closer to the source, and at the speed in which it is being created. Streaming analytics utilizes various techniques for intelligently processing data as it arrives at the edge or within the data center, with the purpose of proactively identifying threats or opportunities for your business.
TIBCO Innovation Workshop Series: Reducing Decision Latency with Streaming Analytics
1. Presented by: Nelson Petracek, Office of the CTO | Strategic Enablement Group
Q4, 2016
Innovation Workshop Series
Reducing Decision Latency with Streaming Analytics
4. BATCH ANALYSIS REAL-TIME ANALYSIS
PROCESS ZONE STRATEGIC TACTICAL OPERATIONS EXECUTION
TIME INCREMENT
IMPORTANCE OF
ALERTS
Quarters/
Years
Months/
Quarters
Hours/
Days/Weeks
Seconds/
Minutes
Not Important Important Necessary
The Evolution of Analytics
5. STREAMING ANALYTICS
VISUAL ANALYTICS
ADVANCED ANALYTICS
• Gain insights at the speed of thought
• Prepare data visually
• Access connected visualizations –
out of the box
• Build beautiful dashboards in minutes
• Collaborate with a click
• Put your insights on a map
(GeoAnalytics)
DATA DISCOVERY DASHBOARDS
• Simplify statistical modeling
• Prescribe your next actions
PREDICTIVE PRESCRIPTIVE
• Gain continuous awareness
• Automate actions and alerts
• Enable human interaction
with live data
REAL TIME ACTIONABLE
The Evolution of Analytics
27. Streaming Analytics: Declarative or Procedural
Sensor
Weather
Write to
DB
Shut Down
SplitUnion
Clean
Clean
Window
Normal
Event
Log
Take
Action
28. Streaming Analytics: Declarative or Procedural
Data Streams
e.g. Flight Status,
Check-ins,
etc.)
EntityState&
Relationships
In Memory, IMDG/DB, SSD, …
Rulesets
Apply Flight Status Rules
Against Passenger
Manifest,
Determine Notifications,
Update Passenger and
Flight State
Correlate
Inbound
Streams, Load
Associated
Business Entities
(Passenger,
Flight, Loyalty,
etc.
Write Changed
Business Entities
to Backend
State Store
Deliver
Outcomes &
Responses
Change in Passenger
State Automatically
Triggers Passenger
Loyalty Rules, Response
Determined
…
29. Streaming Analytics: Logic Examples
Simple and Complex Events
• If F(E) then A(E)
• If F(E1,E2) then A(E1,E2)
Spatial
• If x outbreak notifications are “close to” each
other, then …
Business Entities (Concepts)
• If Changed(Customer) then A(E)
• If Changed(Customer) & Changed(Product)
Event Routing
• Route on event contents, historical data, etc.
Events and Business Entities
• If NewOrder(Customer) & Changed(Product)
…
Event Correlation (Stateful)
• Correlate on events, business entities, time.
Temporal / Missing Events
• If E1 followed by E2 within 20 secs then …
• If No Status Update in 30 mins then…
Rules with Basic Math
• If slope of pressure > 5 PSI, then investigate
equipment
Aggregation
• If Volume within last hour > Avg Volume
then ...
Rules on Top of Analytical Models
• Predictive Model produces prediction
• Rule applied to prediction
• If Prob(Fraud) > 0.8, then investigate claim
30. Streaming Analytics: Machine Learning
Predictive
Attempt to predict
what is “going” to
happen, based on
what “has”
happened in the
past.
Prescriptive
Once the
predictive model is
found, prescribe
your next best or
specific action to
take.
Supervised
Infer a prediction
from a model
produced via set
of training data,
where the inputs
and outputs are
representative of
the real-world.
Unsupervised
Draw inferences
based on input
data with no pre-
determined
responses. Find
hidden patterns or
groupings.
31. Streaming Analytics: Graph Analysis
Graph databases transform a complex web of dynamic data into meaningful
(and understandable) relationships.
• Stream Analytics Contextual Data
• Efficient Entity Link Analysis Triggered by Event Arrival
• Leverage Connected Data during Stream Processing
• Eliminate complex joins and self-joins in a relational model.
Flexible Schema
Assumes objects and nodes are linked
by relationships; designed to
constantly evolve, without impacting
performance of existing queries and
app functionality.
Consistent Performance
Index-free adjacency negates
requirement for index lookups –
enabling query performance to
remain relatively consistent, even as
datasets grow.
Increased Value
Enables quick extraction of new insight
from large and complex databases.
Helps uncover unknown interactions
and relationships. Provides valuable
insight into semantic context.
44. TIBCO BusinessEvents
Build Event-Driven Apps for Contextual Control
Developer Highlights
• Integrated development environment
• Graphic editors and model-driven environment
• Non-linear programming model in the form of declarative rules
Business User Highlights
• Define and implement rules and logic in a web browser.
• Decision tables and rule templates
• Adjust system behavior with no downtime
Runtime Highlights
• Multi-protocol channel support
• Event-driven rule evaluation and execution
• Stateful for reasoning across time and space
• Multiple deployment topology options
• Horizontal scalability / Memory management strategies
• Missing Event Detection
• Data grid securityBusinessEvents WebStudio UI
46. LiveView Web: Active Visualization & Alerting
Zero install, streaming ready.
WebSockets-Powered HTML5
Highcharts-powered visualizations display
your data as it changes.
Real-time Visualizations
Fine-tune your experience by authoring or
downloading Javascript extensions from
the community exchange.
JavaScript Plug-ins