Today’s growing commercial, operational and regulatory pressures underscore the importance of information management. Access to critical business information at the right time is essential to maintain competitive advantage. Despite industry progress, Enterprise Information Management (EIM) professionals continue to face implementation challenges due to the ever increasing and innovative ways information is used throughout the enterprise. Common practical challenges:
- Varying degree of information structure
- Dynamically locating, accessing and trusting the information
- Facilitating navigation, analysis and information visualization
With our EIM experience and deep cross-industry knowledge, Knowledgent believes these challenges are an opportunity to innovate. xBI is a platform that enables Knowledgent to deliver industry-centric solutions that are turn-key, configurable and/or managed-services for their clients. These solutions address EIM challenges by:
- Declaratively creating solutions that quickly enable complex and flexible use of structured and unstructured information
- Collating data from multiple data sources into a single model making holistic processing across the typical enterprise information landscape possible
- Integrating deep traceability, from the data processing rules down to the data that enhances trust and visibility
- Scaling to data complexity and volumes of the largest enterprise
The xBI Difference
At a high level, xBI exploits the difference between programming and modeling. Programming/procedural approaches require predetermining the order of steps necessary to solve a given problem. In contrast, declarative modeling systems specify rules independently from the problem to be solved. It’s then up to the modeling system to analyze these rules to determine how to apply them in an order that satisfies all the conditions of the solution. A major challenge with this approach is providing a way to declare the rules that scales to enterprise solutions.
As an illustration of the contrast between declarative and procedural approaches, compare a business spreadsheet environment with the stack of tools typically used to create an analytical solution (ETL, BI, Stored Procedures, programming language, etc.). A tool like Excel™ is a declarative modeling system in that the user wants a particular cell in a spreadsheet to represent a particular value. It’s necessary only to specify the formula required to compute that value. More specifically, the user does not need to identify any other cells that must be computed before the desired value is computed. The ability to operate at a high-level (and not pre-specify computational order) is what gives spreadsheets much of their power and flexibility vs. a procedural approach.
xBI ‘rules’ are declarative specifications used to solve information management problems. xBI is a modeling system for specifying the solution to the entire problem, from the lowest level data access thru the business processing and presentation, in ways that are faster to build and scale more effectively than other approaches. The integration of a federated data access, declarative modeling of the business and data logic and slice-and-dice style reporting, makes xBI dramatically more powerful. Through this integration the declarative approach is leveraged across all aspects of any solution.
xBI Technical Overview
Although xBI has multiple add-on modules, the core is the integration of the data access, rule processing and presentation capabilities.
- Federated Data Access: Enables solutions that span multiple data sources without physically moving or changing data. The tables, columns and joins needed by the rules are modeled and managed via the federation. This forms the foundation for a virtual data hypercube warehouse.
- Rules Engine: Dynamically combines data and generates insights. Every variable, dimension, fact, hierarchy, calculation, time period, data element and exception is modeled as a declarative rule. The rules and federation are linked via data map rules that point to values in the data source or multiple data sources based on rule conditions.
- Presentation: Is comparable to traditional standalone Business Intelligence (BI) products, such as a web-based interface enabling users to create and view dashboards, reports and run ad-hoc and “what if” analysis. xBI’s presentation layer is distinguished by its ability to leverage not just data, but by presenting the output of arbitrary points in the layers of rules. Further xBI supports tracing through the rules all the way to the data level. This deep integration is a core xBI differentiator, exploiting the advantages of the declarative rule based approach across the entire stack.
xBI’s declarative rules engine works by providing an optimized way to describe and operate on the data in an online analytical processing (OLAP)-like cube.
At the base are the rules to obtain the data from the source and place it into the cube. From there rules are defined to realize all the additional data and behavior needed for the solution. The results of a rule can be based on data or the results of other rules. In standard warehouse terminology, the results of the rules form both facts and dimensions. Rules can be as simple as Profit = Revenue – Expense or they can perform complex calculation and data access logic.
xBI is differentiated from similar sounding solutions through its use of Rule-Order independence. This capability is enabled by powerful “meta rules” (rules for rules) that handle rule ordering, and provides scaling well beyond spreadsheet-class applications. (The application of these rules is patented under US patent #5,918,232.)
Meta Rules Principles
The meta rules are inspired by two basic and intuitive principles.
The first principle is that more specific business rules have a higher priority. For example, a rebate calculation for trade executions of blocks greater than 10K shares is “.01% * (price * volume).” But there is a more specific calculation for trade volumes greater than 10K and a spread greater than 1/16 for a particular customer. That more specific rule will override less specific rules, and is a key enabler of xBI’s enterprise-scale analytic solutions.
The second principle is the priority of using data directly from a database in dimensions. For example, a hierarchy in the time dimension, e.g., day, month, quarters, year, where each point of aggregation in the hierarchy can be associated with a business rule. The vast majority of the rules in a typical model actually come from the aggregations in these hierarchies. These rules have a lower priority than rules defined by the model builder, following the intuitive idea that the database-derived rules are the basic defaults. The model builder could test the effect of implementing a September fiscal year end by creating new rules to define the new year.
Beyond this, if a user wants to change a value computed by the model, that “what if” value takes precedence over other rules of equivalent specificity, otherwise, the user’s desire to test the effect of the changed value won’t be properly honored. (Note: the ability to change values as needed to support what if or to write-back to the source is another differentiator.)
The importance of rule ordering impacts accuracy of results and efficiency of result delivery. Lacking appropriate efficiency can often translate into no results if the computational effort required is too great.
Real-time Rules Management
Although the xBI rules approach is intuitively powerful, the real power of xBI is in its ability to apply the rules in real time. For example, the power of spreadsheets would be greatly diminished if making any change to cell computation required hours or even minutes to process before answers could be retrieved. Interactivity is the key benefit of an analytic model. xBI’s patented “active rules” architecture optimizes rules ordering and models to provide dramatic performance gains.
The rules optimization is accomplished by the rule set preparation process. xBI evaluates all the rules in combination, producing an internal, multidimensional representation of the model. This representation contains entries with a specification of a collection of cells in the cube and a rule which can be applied to compute the value of each of the cells. Once organized and a request for data is presented, the precise definition of the process for computing every single value from the entire business model can be optimally found and computed.
The xBI user interface further leverages the optimized rule organization to allow interactive exploration of a model’s rules. This makes it possible to determine instantly and precisely not only how any value in the model was derived, but to the ability to show the actual values or data sources used. Further, it’s possible to see at a glance what “downstream” computations any value might be used in.
It’s critical that actual retrieval of data from source systems and computation of cell values from those data be as fast as possible.
xBI incorporates a large number of optimizations in both database access and internal calculations to address the performance requirements. Computations, filtering and sorting operations are pushed into the underlying relational database where the hardware and software platform will generally be scaled to match the volume of data being analyzed. Only if the operation can’t be performed by the database will it be performed by xBI. This occurs relatively infrequently, and the frequency can be reduced further by the ability for the business analyst to individualize specific SQL operations as part of the model’s definition.
Other optimizations are designed to increase performance by not having the database do some operations. For example, if most but not all rows need to be returned from a table, instead of giving the database a large and time-consuming filtering, xBI will drop the database filter altogether, and discard the few unneeded rows itself.
Finally, xBI fully supports caching of parts of a model and has a multithreaded architecture optimized for use on symmetric multi-processor hardware platforms.
Evolution of any solution must keep pace with the constant change of modern businesses. Every change implies changes to the business models used to guide the business.
The combination of the xBI capabilities, most notably declarative rules rather than traditional fragile procedural approaches, enable solution evolution at the pace of business.
Additional plug-in functionality augments xBI
- Semantic Enrichment Module: Provides the capability to integrate structured and unstructured data. The module allows data from social media websites (e.g., Linked-In / Twitter), or internal text sources (e.g., call-center call notes), to be semantically analyzed. This analysis produces insights such as people, places, organizations, sentiment, etc., and stores them as structured data that can be merged into xBI rules-based models. This makes solutions like semantic enrichment of Master Data, discovery of customer affiliations, and visibility of trends within and outside of the enterprise attainable.
- Data Management Modules: Comprised of discovery, metadata, lineage and governance components that provide a detailed picture of the data structures and underlying relationships of the data sources used in a solution.
- Federation Plug-in: Extends the available data universe to almost any source beyond the RDBMS-centric approach immediately available via the xBI core. This aspect of the plug-and-play architecture of xBI is frequently used to integrate xBI in an enterprise that already has a data federation product in place.
- BI Tools Plug-in: Extends the output of xBI rule-based processing to feed existing traditional BI tool deployments. A powerful variation on this theme is integration BI tools that support the XMLA-MDX standards. xBI can deliver the results of rules to any XMLA-MDX compliant product making the output of xBI transparently available.