Real-time Business Intelligence - InterSystems Benelux
Real-time Business Intelligence - InterSystems Benelux
Real-time Business Intelligence - InterSystems Benelux
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
DeepSee Overview<br />
Joe Gallant<br />
Symposium 2010
Embedded + <strong>Real</strong> Time<br />
Delivered as an<br />
integral part of an<br />
application<br />
Access to current<br />
data needed for<br />
operational decision<br />
making<br />
Embedded <strong>Real</strong>-Time<br />
<strong>Business</strong> <strong>Intelligence</strong>
Competitive Comparison<br />
Embedded <strong>Real</strong>-Time BI<br />
Data accessed in<br />
place<br />
Current + historical,<br />
detail + summary data<br />
Any application user<br />
Traditional BI<br />
Data extracted and<br />
loaded periodically<br />
Historical summary<br />
data only<br />
Specialist users with<br />
tool + data schema<br />
expertise
DeepSee In Action @ QuadraMed<br />
Rules<br />
Rules<br />
Patient Registration<br />
Workflow<br />
“With DeepSee, you’re<br />
managing the<br />
configuration and<br />
performance of your<br />
business in real <strong>time</strong>.”<br />
Patient Registration<br />
Workflow with DeepSee<br />
• <strong>Real</strong>-<strong>time</strong><br />
monitoring of<br />
quality, throughput,<br />
backlog, …<br />
• Dynamic tuning of<br />
rules
Registration WorkList Dashboard
Actionable <strong>Business</strong> <strong>Intelligence</strong>
DeepSee Lessons Learned<br />
• People like DeepSee!<br />
– Tremendous customer interest<br />
– Embedded BI is actually useful<br />
• What people like about DeepSee:<br />
– Performance<br />
– Ability to handle complex data<br />
– Ease of use<br />
– Thin client<br />
• Finding the right use case is the key!
DeepSee Lessons Learned<br />
• What people would like improved in<br />
DeepSee:<br />
– Performance<br />
– Ability to handle complex data<br />
– Ease of use<br />
– Thin client
All DeepSee is divided in 3 parts<br />
Query<br />
Engine<br />
Portal<br />
Data Model
DeepSee Model Improvements<br />
• Well-defined data model<br />
– Simplifies the design of the data model<br />
– XML representation is available<br />
– Better integration with source control<br />
• Removes the need to modify existing<br />
transactional classes<br />
• Use of standard bitmap and bitslice indices<br />
• Object-based <strong>time</strong> dimensions for better<br />
flexibility
DeepSee Data Model<br />
• Cubes<br />
– Defines dimensions and<br />
measures<br />
• Subject Areas<br />
– “Views” on cubes<br />
– Provides automatic filtering<br />
• KPIs<br />
– Makes more sophisticated<br />
computations available to<br />
dashboards<br />
– Can make use of DeepSee, SQL,<br />
or custom logic<br />
Note: Lifting of cubes in<br />
thunderstorm will void warrantee
DeepSee Execution Engine<br />
• Cleaner, more streamlined engine based on<br />
native index structures<br />
• ResultSet API available to make DeepSee<br />
functionality available to applications<br />
• Support for MDX query language<br />
– Subset at first (reasonably complete)<br />
– More complete implementation as needed<br />
– Extensions for additional power
DeepSee Query Engine<br />
• Analyzer<br />
– Visual query construction<br />
– Queries can be saved<br />
• DeepSee ResultSet<br />
– Programmatic API<br />
• DeepSee Shell<br />
– Interactive queries<br />
Keep fingers away from gears<br />
Keep arrows away from eyes<br />
Keep magnifying glass out of strong sunlight
DeepSee Performance and Scalability<br />
• Multi-level, incremental caching to support<br />
large data models (100M+ facts)<br />
• Support for parallel execution of queries to<br />
exploit multi-core architectures:<br />
– Queries are split by # of facts<br />
– Queries are split by # of cells<br />
– Subqueries and joins<br />
• Logic for updates to Data Model is<br />
streamlined
DeepSee User Interface<br />
• Clear separation of DeepSee engine and<br />
DeepSee UI<br />
– APIs available to access DeepSee engine<br />
• Built with Zen technology<br />
– Supports multiple browsers<br />
– Standard Zen components<br />
– Uses the DeepSee APIs
DeepSee Portal<br />
• User Portal<br />
– Clean, easy way to see dashboards,<br />
reports, and saved queries<br />
• Dashboards<br />
– Display data from saved queries or<br />
KPIs<br />
– Data displayed in visual “widgets”<br />
– Dashboard “actions” can interact with<br />
application logic
DeepSee Overview<br />
Joe Gallant<br />
Symposium 2010