Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Lily 2.4 – What’s New Product Release Notes
WHAT’S NEW IN LILY 2.4 2
NGDATA © Copyright 2013 http://www.NGDATA.com
Table of Contents
Table of Contents ....................................................................................................... 2
Purpose and Overview of this Document ................................................................ 3
Product Overview ...................................................................................................... 4
General ........................................................................................................................ 5 Prerequisites ........................................................................................................... 5
The Lily Data Repository ........................................................................................... 6 Side Effect Processor ............................................................................................. 6 Multi-Repository Support ....................................................................................... 6 Metadata ................................................................................................................... 6 Record Type Inheritance ........................................................................................ 5 HDFS Name Node Support ..................................................................................... 8
The Lily Customer Database ..................................................................................... 9 Foundations ............................................................................................................. 9 Master Records ....................................................................................................... 9 Hive Integration ....................................................................................................... 9 Interaction Summary Attribute Calculation Engine ........................................... 10
Lily Customer Database Explorer ......................................................................... 10
User Interface ....................................................................................................... 11
Item Explorer ........................................................................................................ 13
Lily Customer Intelligence Applications ................................................................ 14 Rules - Based Recommendation Strategies ....................................................... 14 Knowledge Based Recommendations ................................................................ 14
Provisioning ............................................................................................................. 15
WHAT’S NEW IN LILY 2.4 3
NGDATA © Copyright 2013 http://www.NGDATA.com
Purpose and Overview of this Document
This document serves as a comprehensive, Customer-oriented overview of all
feature additions, changes and enhancements available in Lily Release 2.4. Lily 2.4
has been released the 22nd of July 2013 with the main release focus being updates
to the Lily Customer Database and the Lily Data Repository.
WHAT’S NEW IN LILY 2.4 4
NGDATA © Copyright 2013 http://www.NGDATA.com
Product Overview
Lily, the Customer Intelligence Platform, Release 2.4 is composed of the following
products:
• Lily Data Repository (DR)
• Lily Customer Database (CDB)
• Lily Customer Intelligence Applications (CA)
• Lily Enterprise tools:
o Cluster installation
o ETL tool connectivity
o Hive connectivity
The Lily Customer Database relies upon the Data Repository, and the Customer
Applications on the Customer Database.
WHAT’S NEW IN LILY 2.4 5
NGDATA © Copyright 2013 http://www.NGDATA.com
General
Prerequisites Lily 2.4 has been certified against Cloudera’s Distribution Including Apache Hadoop
(CDH4) and therefore requires the following:
• CDH4.2
o Supported Operating Systems:
http://www.cloudera.com/content/cloudera-content/cloudera-
docs/CDH4/4.2.0/CDH4-Requirements-and-Supported-
Versions/cdhrsv_topic_1.html
o Supported Databases: http://www.cloudera.com/content/cloudera-
content/cloudera-docs/CDH4/4.2.0/CDH4-Requirements-and-
Supported-Versions/cdhrsv_topic_2.html
o Support JDK: http://www.cloudera.com/content/cloudera-
content/cloudera-docs/CDH4/4.2.0/CDH4-Requirements-and-
Supported-Versions/cdhrsv_topic_3.html
WHAT’S NEW IN LILY 2.4 6
NGDATA © Copyright 2013 http://www.NGDATA.com
The Lily Data Repository
With Release 2.4 of the Lily Data Repository a number of key enhancements and
expansion of functionality has been included. By delivering these enhancements
NGDATA has strengthened and expanded the foundation of the Lily Data Repository,
for all users and continuing to provide the best-in-breed solution for all clients.
Side Effect Processor A new component, the Side Effect Processor (SEP) has been developed for the Data
Repository allowing for greater processing efficiency of asynchronous update
operations The SEP sits at the core of Lily’s triggering and indexing mechanism and
replaces parts of the existing RowLog engine. With this enhancement, Lily is three
times faster with indexing switched off, and five times faster when using indexing.
Also, maintenance of multiple indexes at the same time bears no additional
performance impact. The SEP is also being used for computing aggregates and
maintaining additional data indexes in the Customer Database.
Multi-Repository Support With this release, records retained within the Lily Data Repository can be stored
across multiple Lily Repositories within HBase, thereby supporting the logical
segregation of data providing additional performance gains and data security. It is
possible to host multiple data repositories in one shared installation, e.g. to operate a
shared Lily for multiple departments.
Metadata Lily has been enhanced with this release to support field level metadata. This
metadata may be comprised of key / value pairs where key is a string and a value is
a simple type, such as String, Integer, Long, Float, Double, Boolean, or byte.
Providing this functionality to the Lily Data Repository allows Users to identify
information such as Source, Creator, Timestamp, Quality, etc. without having to
predefine this field in the schema and therefore making the addition of this data by an
application simpler. Amongst others, field metadata may be utilized to store access
control information.
WHAT’S NEW IN LILY 2.4 7
NGDATA © Copyright 2013 http://www.NGDATA.com
Record Type Inheritance When creating a schema definition the Lily Data Repository functionality now
supports the concept of ‘Supertypes’. This allows a User to define a base record
type and then ‘Inherit’ the structure of that base type into another record type that
then extends it. As an example:
recordTypes: [
{
name: “person”,
fields : [
{ name: “name”, mandatory: true },
{ name: “address”, mandatory: true}
]
},
{
name: “Customer”,
fields: [
{ name: “Customer Number”, mandatory: true }
],
supertypes: [
{ name: “person” }
]
},
{
name: “employee”,
fields: [
{ name: “employeeNumber”, mandatory: true }
],
supertypes: [ { name: “person” } ]
}
]
WHAT’S NEW IN LILY 2.4 8
NGDATA © Copyright 2013 http://www.NGDATA.com
In this example, the Customer record will have fields Name, Address, and Customer
Number while the Employee record will have Name, Address, and Employee
Number. By utilizing the recordtype inheritance, schema design has become much
more elegant.
HDFS Name Node Support Lily 2.4 Enterprise ships with deployment configuration and instructions to support
high-availability of the Hadoop HDFS Name Node service. This allows enterprise
clients to run Hadoop in a fault-tolerant configuration and be able to maintain system
uptime in the event of server failure.
WHAT’S NEW IN LILY 2.4 9
NGDATA © Copyright 2013 http://www.NGDATA.com
The Lily Customer Database
Foundations At the foundation of the Lily Customer Database, with this release the two
enhancements are being delivered:
• Interaction Timestamps
o Lily will add a Timestamp to all Interactions, when the Interaction
Timestamp has not been specified on the input data, thereby ensuring
the Timestamp field will always contain a value
• Automatic Creation of Item or Customer
o Lily will add a new Item and/or Customer record when an Interaction is
logged for a non-existent Item and/or Customer ensuring 100% of the
Interactions are prepared for further analysis
Master Records An enhancement to the Lily Customer Database allows Users to specify that
Customer records from multiple sources are actually for the same Customer. When
this occurs, a Master Customer record is created to combine all of the data from the
different sources. Each Source Record is kept intact for auditing purposes, but a
Master Record is updated to reflect a combination of the most recent data. This is
beneficial because it allows Lily Users to get a single view of their Customers in a
central place based on data from different systems.
Hive Integration Expanding upon the existing Lily Hive integration, a number of enhancements and
added features have been developed and included in Release 2.4, which are
highlighted below:
• Generate SQL like Queries called HiveQL for searching data in the Customer
Database through a wizard on the Lily User Interface
o This User Interface allows a User to generate a Hive query that which
can be run offline
o This allows the query to be copied into and utilized inside popular BI or
data exploration tools supporting Hive such as Tableau, Toad4Cloud,
…
WHAT’S NEW IN LILY 2.4 10
NGDATA © Copyright 2013 http://www.NGDATA.com
• Search data in Lily via Hive using field level metadata
• Store data set statistics in the HDFS layer of Lily
By expanding the Hive integration NGDATA continues to improve the ability of Users
to do analysis on the data that is stored in Lily.
Interaction Summary Attribute Calculation Engine The Lily Customer Database has been updated with an engine for both real-time &
batch calculation of Summary Attributes that can be defined by the Client, such as:
amount of visits, average amount spent, time on website, etc. Lily supports the
following basic aggregators: Max, Min, Average, Count, Distinct Count, Last, Sum.
Summary Attributes are calculated values, which are based on the analysis of
incoming/processed interactions. Summaries Attributes are always stored on a single
Customer or Item record, which is comparable to an SQL query with an ‘Aggregation’
(e.g. SUM) and a GROUP BY Customer.
The real-time functionality of this ‘Variable Calculation Engine’ has been developed
as a SEP component to compute the Summary Attributes in real-time during
ingestion of the data. The calculated values can also be used in the Lily Customer
Database Explorer for Facet Search, allowing for a richer User experience in the
User Interface and a continued expansion of the functionality of Lily as a data
exploration tool. The usage of the Summary Attributes within the Faceted Search
functionality can be enabled/disabled on a Summary Attribute level.
In addition to the real-time functionality, Lily also supports a batch-based rebuild of all
variables using the parallel processing power of MapReduce.
Lily Customer Database Explorer The Lily Customer Database Explorer has been enhanced with the functionality to
support facet-based searches using graph widgets (bar chart, pie chart…). This type
of User Interface navigation provides the User visual insights into the Lily Customer
Database, allowing a step-by-step creation of Customer or Item target views. The
Lily Customer Database Explorer also makes it possible to select one or more
Segments to aid in segment analysis by utilizing interactive usage & selection within
the of the graph-based User Interface widgets.
WHAT’S NEW IN LILY 2.4 11
NGDATA © Copyright 2013 http://www.NGDATA.com
The graph-based widgets utilize faceted data as input, provided by the Lily Customer
Database and support the following types of facets:
• Aggregated Facets: These type of facets make use of interaction summary
attributes, like shopping frequency, click count, …
• Field Facets: These type of facets are not using summarized data but basic
field values like gender, region, …
User Interface With the delivery of Release 2.4, the Customer Database Explorer includes the
functionality to drill down to the individual Customer records based on a ‘Faceted
Search’ by the following areas of Customer classification:
• Identity data
• Profile and segmentation data
• Behavioral data aggregated from customer interactions
• Preferences calculated by scoring and learning engines
• Source selection
A User is able to filter on selected facet values through an Include/Exclude function,
allowing for easy audience or item selection through a “query-by-example” interface.
WHAT’S NEW IN LILY 2.4 12
NGDATA © Copyright 2013 http://www.NGDATA.com
Additionally, a User is able to get down to view the results of the filter in the screen
below. The resulting results will include the following information: ID, Identity,
Behavior and Preference for the individual record result.
The User is able to see ID, Identity, Behavior and Preference results on a individual
record result.
WHAT’S NEW IN LILY 2.4 13
NGDATA © Copyright 2013 http://www.NGDATA.com
Item Explorer With the release of Lily 2.4, a User will be able to explore Item data through a
‘Faceted Search’ as well. The User will begin the search through product categories,
filter facets, and delving in the data of individual Items.
WHAT’S NEW IN LILY 2.4 14
NGDATA © Copyright 2013 http://www.NGDATA.com
Lily Customer Intelligence Applications
Rules-Based Recommendation Strategies The Rules-Based Recommendation Strategies functionality has been enhanced to
support Strategies through dynamic business rules for Pre / Post processing of
Recommendations. This includes selection of recommendation engines, different
models within the Recommendation Engines, bypassing recommendation engine
consultation, as well as influencing scores based on decision tables. A variety of
Business Rules may be defined specific to a Clients needs and the calculations or
the activation / deactivation of individual rules may be made by a User without having
to redeploy a new version of the application.
Knowledge-Based Recommendations A new Knowledge-Based Recommendation Engine can utilize the knowledge about
Users and Items and reasons out what products meet the Users requirements. This
has been delivered with this release of Lily. It supports the configuration and
application of custom-made, domain-specific recommendation scores based on
aggregates and overall Customer interaction data. This functionality allows Users to
determine recommendations for a Customer based solely on their behavior and not
taking into account the interactions of other Customers, such as with the Lily
Collaborative Filtering Machine Learning Recommendation Engine.
WHAT’S NEW IN LILY 2.4 15
NGDATA © Copyright 2013 http://www.NGDATA.com
Provisioning
With the 2.4 release of Lily, a new provisioning framework has been created to assist
in the installation and upgrade of Lily clusters. This framework is a Python solution
based upon Fabric and is capable of installing an entire Lily cluster including
Cloudera Manager and CDH4. By utilizing this tool, it is possible to create a Lily
cluster on EC2 and have it running within 15 minutes. It replaces the existing Whirr-
based installer and allows for a variety of deployment automations.
For more information on Lily, please contact us at [email protected].