Data Warehousing Quick Guide
Data Warehousing Quick Guide
An operational database undergoes frequent changes on a daily basis on account of the transactions that take place. Suppose a
business executive wants to analyze previous feedback on any data such as a product, a supplier, or any consumer data, then the
executive will have no data available to analyze because the previous data has been updated due to transactions.
A data warehouses provides us generalized and consolidated data in multidimensional view. Along with generalized and consolidated
view of data, a data warehouses also provides us Online Analytical Processing (OLAP) tools. These tools help us in interactive and
effective analysis of data in a multidimensional space. This analysis results in data generalization and data mining.
Data mining functions such as association, clustering, classification, prediction can be integrated with OLAP operations to enhance the
interactive mining of knowledge at multiple level of abstraction. That's why data warehouse has now become an important platform for
data analysis and online analytical processing.
It possesses consolidated historical data, which helps the organization to analyze its business.
A data warehouse helps executives to organize, understand, and use their data to take strategic decisions.
An operational database is constructed for well-known tasks and workloads such as searching particular records, indexing, etc.
In contract, data warehouse queries are often complex and they present a general form of data.
Operational databases support concurrent processing of multiple transactions. Concurrency control and recovery mechanisms
are required for operational databases to ensure robustness and consistency of the database.
An operational database query allows to read and modify operations, while an OLAP query needs only read only access of
stored data.
An operational database maintains current data. On the other hand, a data warehouse maintains historical data.
Subject Oriented - A data warehouse is subject oriented because it provides information around a subject rather than the
organization's ongoing operations. These subjects can be product, customers, suppliers, sales, revenue, etc. A data warehouse
does not focus on the ongoing operations, rather it focuses on modelling and analysis of data for decision making.
Integrated - A data warehouse is constructed by integrating data from heterogeneous sources such as relational databases,
flat files, etc. This integration enhances the effective analysis of data.
Time Variant - The data collected in a data warehouse is identified with a particular time period. The data in a data
warehouse provides information from the historical point of view.
Non-volatile - Non-volatile means the previous data is not erased when new data is added to it. A data warehouse is kept
separate from the operational database and therefore frequent changes in operational database is not reflected in the data
warehouse.
Note: A data warehouse does not require transaction processing, recovery, and concurrency controls, because it is physically stored
and separate from the operational database.
Financial services
Banking services
Consumer goods
Retail sectors
Controlled manufacturing
Information Processing - A data warehouse allows to process the data stored in it. The data can be processed by means of
querying, basic statistical analysis, reporting using crosstabs, tables, charts, or graphs.
Analytical Processing - A data warehouse supports analytical processing of the information stored in it. The data can be
analyzed by means of basic OLAP operations, including slice-and-dice, drill down, drill up, and pivoting.
Data Mining - Data mining supports knowledge discovery by finding hidden patterns and associations, constructing analytical
models, performing classification and prediction. These mining results can be presented using the visualization tools.
2 OLAP systems are used by knowledge workers such as OLTP systems are used by clerks, DBAs, or database professionals.
executives, managers, and analysts.
5 It is based on Star Schema, Snowflake Schema, and Fact It is based on Entity Relationship Model.
Constellation Schema.
8 It provides summarized and consolidated data. It provides primitive and highly detailed data.
9 It provides summarized and multidimensional view of data. It provides detailed and flat relational view of data.
11 The number of records accessed is in millions. The number of records accessed is in tens.
12 The database size is from 100GB to 100 TB. The database size is from 100 MB to 100 GB.
Tuning Production Strategies - The product strategies can be well tuned by repositioning the products and managing the
product portfolios by comparing the sales quarterly or yearly.
Customer Analysis - Customer analysis is done by analyzing the customer's buying preferences, buying time, budget cycles,
etc.
Operations Analysis - Data warehousing also helps in customer relationship management, and making environmental
corrections. The information also allows us to analyze business operations.
Query-driven Approach
Update-driven Approach
Query-Driven Approach
This is the traditional approach to integrate heterogeneous databases. This approach was used to build wrappers and integrators on
top of multiple heterogeneous databases. These integrators are also known as mediators.
When a query is issued to a client side, a metadata dictionary translates the query into an appropriate from for individual
heterogeneous sites involved.
Now these queries are mapped and sent to the local query processor.
The results from heterogeneous sites are integrated into a global answer set.
Disadvantages
This approach is also very expensive for queries that require aggregations.
Update-Driven Approach
This is an alternative to the traditional approach. Today's data warehouse systems follow update-driven approach rather than the
traditional approach discussed earlier. In update-driven approach, the information from multiple heterogeneous sources are integrated
in advance and are stored in a warehouse. This information is available for direct querying and analysis.
Advantages
This approach has the following advantages:
The data is copied, processed, integrated, annotated, summarized and restructured in semantic data store in advance.
Query processing does not require an interface to process data at local sources.
Data Transformation - Involves converting the data from legacy format to warehouse format.
Data Loading - Involves sorting, summarizing, consolidating, checking integrity, and building indices and partitions.
Refreshing - Involves updating from data sources to warehouse.
Note: Data cleaning and data transformation are important steps in improving the quality of data and data mining results.
Metadata
Metadata is simply defined as data about data. The data that are used to represent other data is known as metadata. For example, the
index of a book serves as a metadata for the contents in the book. In other words, we can say that metadata is the summarized data
that leads us to the detailed data.
Metadata acts as a directory. This directory helps the decision support system to locate the contents of a data warehouse.
Metadata Repository
Metadata repository is an integral part of a data warehouse system. It contains the following metadata:
Business metadata - It contains the data ownership information, business definition, and changing policies.
Operational metadata - It includes currency of data and data lineage. Currency of data refers to the data being active,
archived, or purged. Lineage of data means history of data migrated and transformation applied on it.
Data for mapping from operational environment to data warehouse - It metadata includes source databases and their
contents, data extraction, data partition, cleaning, transformation rules, data refresh and purging rules.
The algorithms for summarization - It includes dimension algorithms, data on granularity, aggregation, summarizing, etc.
Data Cube
A data cube helps us represent data in multiple dimensions. It is defined by dimensions and facts. The dimensions are the entities with
respect to which an enterprise preserves the records.
The following table represents the 2-D view of Sales Data for a company with respect to time, item, and location dimensions.
But here in this 2-D table, we have records with respect to time and item only. The sales for New Delhi are shown with respect to time,
and item dimensions according to type of items sold. If we want to view the sales data with one more dimension, say, the location
dimension, then the 3-D view would be useful. The 3-D view of the sales data with respect to time, item, and location is shown in the
table below:
The above 3-D table can be represented as 3-D data cube as shown in the following figure:
Data Mart
Data marts contain a subset of organization-wide data that is valuable to specific groups of people in an organization. In other words, a
data mart contains only those data that is specific to a particular group. For example, the marketing data mart may contain only data
related to items, customers, and sales. Data marts are confined to subjects.
Windows-based or Unix/Linux-based servers are used to implement data marts. They are implemented on low-cost servers.
The implementation cycle of a data mart is measured in short periods of time, i.e., in weeks rather than months or years.
The life cycle of data marts may be complex in the long run, if their planning and design are not organization-wide.
Virtual Warehouse
The view over an operational data warehouse is known as virtual warehouse. It is easy to build a virtual warehouse. Building a virtual
warehouse requires excess capacity on operational database servers.
Data Warehousing - Delivery Process
A data warehouse is never static; it evolves as the business expands. As the business evolves, its requirements keep changing and
therefore a data warehouse must be designed to ride with these changes. Hence a data warehouse system needs to be flexible.
Ideally there should be a delivery process to deliver a data warehouse. However data warehouse projects normally suffer from various
issues that make it difficult to complete tasks and deliverables in the strict and ordered fashion demanded by the waterfall method.
Most of the times, the requirements are not understood completely. The architectures, designs, and build components can be
completed only after gathering and studying all the requirements.
Delivery Method
The delivery method is a variant of the joint application development approach adopted for the delivery of a data warehouse. We have
staged the data warehouse delivery process to minimize risks. The approach that we will discuss here does not reduce the overall
delivery time-scales but ensures the business benefits are delivered incrementally through the development process.
Note: The delivery process is broken into phases to reduce the project and delivery risk.
IT Strategy
Data warehouse are strategic investments that require a business process to generate benefits. IT Strategy is required to procure and
retain funding for the project.
Business Case
The objective of business case is to estimate business benefits that should be derived from using a data warehouse. These benefits
may not be quantifiable but the projected benefits need to be clearly stated. If a data warehouse does not have a clear business case,
then the business tends to suffer from credibility problems at some stage during the delivery process. Therefore in data warehouse
projects, we need to understand the business case for investment.
The prototype can be thrown away after the feasibility concept has been shown.
The activity addresses a small subset of eventual data content of the data warehouse.
The following points are to be kept in mind to produce an early release and deliver business benefits.
Limit the scope of the first build phase to the minimum that delivers business benefits.
Understand the short-term and medium-term requirements of the data warehouse.
Business Requirements
To provide quality deliverables, we should make sure the overall requirements are understood. If we understand the business
requirements for both short-term and medium-term, then we can design a solution to fulfil short-term requirements. The short-term
solution can then be grown to a full solution.
Technical Blueprint
This phase need to deliver an overall architecture satisfying the long term requirements. This phase also deliver the components that
must be implemented in a short term to derive any business benefit. The blueprint need to identify the followings.
History Load
This is the phase where the remainder of the required history is loaded into the data warehouse. In this phase, we do not add new
entities, but additional physical tables would probably be created to store increased data volumes.
Let us take an example. Suppose the build version phase has delivered a retail sales analysis data warehouse with 2 months’ worth of
history. This information will allow the user to analyze only the recent trends and address the short-term issues. The user in this case
cannot identify annual and seasonal trends. To help him do so, last 2 years’ sales history could be loaded from the archive. Now the
40GB data is extended to 400GB.
Note: The backup and recovery procedures may become complex, therefore it is recommended to perform this activity within a
separate phase.
Ad hoc Query
In this phase, we configure an ad hoc query tool that is used to operate a data warehouse. These tools can generate the database
query.
Note: It is recommended not to use these access tools when the database is being substantially modified.
Automation
In this phase, operational management processes are fully automated. These would include:
Monitoring query profiles and determining appropriate aggregations to maintain system performance.
Note: This phase should be performed separately, since it involves substantial efforts and complexity.
Requirements Evolution
From the perspective of delivery process, the requirements are always changeable. They are not static. The delivery process must
support this and allow these changes to be reflected within the system.
This issue is addressed by designing the data warehouse around the use of data within business processes, as opposed to the data
requirements of existing queries.
The architecture is designed to change and grow to match the business needs, the process operates as a pseudo-application
development process, where the new requirements are continually fed into the development activities and the partial deliverables are
produced. These partial deliverables are fed back to the users and then reworked ensuring that the overall system is continually
updated to meet the business needs.
In this chapter, we will discuss how to build data warehousing solutions on top open-system technologies like Unix and relational
databases.
Note: Before loading the data into the data warehouse, the information extracted from the external sources must be reconstructed.
For example, in a customer profiling data warehouse in telecommunication sector, it is illogical to merge the list of customers at 8 pm
on Wednesday from a customer database with the customer subscription events up to 8 pm on Tuesday. This would mean that we are
finding the customers for whom there are no associated subscriptions.
Note: Consistency checks are executed only when all the data sources have been loaded into the temporary data store.
Aggregation
within itself.
Transforming involves converting the source data into a structure. Structuring the data increases the query performance and decreases
the operational cost. The data contained in a data warehouse must be transformed to support performance requirements and control
the ongoing operational costs.
Aggregation
Aggregation is required to speed up common queries. Aggregation relies on the fact that most common queries will analyze a subset or
an aggregation of the detailed data.
For example, in a retail sales analysis data warehouse, it may be required to keep data for 3 years with the latest 6 months data being
kept online. In such as scenario, there is often a requirement to be able to do month-on-month comparisons for this year and last year.
In this case, we require some data to be restored from the archive.
ensures that all the system sources are used in the most effective way.
The information generated in this process is used by the warehouse management process to determine which aggregations to
generate. This process does not generally operate during the regular load of information into data warehouse.
A data warehouse provides us a consistent view of customers and items, hence, it helps us manage customer relationship.
A data warehouse also helps in bringing down the costs by tracking trends, patterns over a long period in a consistent and
reliable manner.
To design an effective and efficient data warehouse, we need to understand and analyze the business needs and construct a business
analysis framework. Each person has different views regarding the design of a data warehouse. These views are as follows:
The top-down view - This view allows the selection of relevant information needed for a data warehouse.
The data source view - This view presents the information being captured, stored, and managed by the operational system.
The data warehouse view - This view includes the fact tables and dimension tables. It represents the information stored
inside the data warehouse.
The business query view - It is the view of the data from the viewpoint of the end-user.
Bottom Tier - The bottom tier of the architecture is the data warehouse database server. It is the relational database system.
We use the back end tools and utilities to feed data into the bottom tier. These back end tools and utilities perform the Extract,
Clean, Load, and refresh functions.
Middle Tier - In the middle tier, we have the OLAP Server that can be implemented in either of the following ways.
By Relational OLAP (ROLAP), which is an extended relational database management system. The ROLAP maps the
operations on multidimensional data to standard relational operations.
By Multidimensional OLAP (MOLAP) model, which directly implements the multidimensional data and operations.
Top-Tier - This tier is the front-end client layer. This layer holds the query tools and reporting tools, analysis tools and data
mining tools.
Virtual Warehouse
Data mart
Enterprise Warehouse
Virtual Warehouse
The view over an operational data warehouse is known as a virtual warehouse. It is easy to build a virtual warehouse. Building a virtual
warehouse requires excess capacity on operational database servers.
Data Mart
Data mart contains a subset of organization-wide data. This subset of data is valuable to specific groups of an organization.
In other words, we can claim that data marts contain data specific to a particular group. For example, the marketing data mart may
contain data related to items, customers, and sales. Data marts are confined to subjects.
Window-based or Unix/Linux-based servers are used to implement data marts. They are implemented on low-cost servers.
The implementation data mart cycles is measured in short periods of time, i.e., in weeks rather than months or years.
The life cycle of a data mart may be complex in long run, if its planning and design are not organization-wide.
Enterprise Warehouse
An enterprise warehouse collects all the information and the subjects spanning an entire organization
The data is integrated from operational systems and external information providers.
This information can vary from a few gigabytes to hundreds of gigabytes, terabytes or beyond.
Load Manager
This component performs the operations required to extract and load process.
The size and complexity of the load manager varies between specific solutions from one data warehouse to other.
Perform simple transformations into structure similar to the one in the data warehouse.
Fast Load
In order to minimize the total load window the data need to be loaded into the warehouse in the fastest possible time.
It is more effective to load the data into relational database prior to applying transformations and checks.
Gateway technology proves to be not suitable, since they tend not be performant when large data volumes are involved.
Simple Transformations
While loading it may be required to perform simple transformations. After this has been completed we are in position to do the
complex checks. Suppose we are loading the EPOS sales transaction we need to perform the following checks:
Strip out all the columns that are not required within the warehouse.
Warehouse Manager
A warehouse manager is responsible for the warehouse management process. It consists of third-party system software, C programs,
and shell scripts.
The size and complexity of warehouse managers varies between specific solutions.
Backup/Recovery tool
SQL Scripts
A warehouse manager analyzes the data to perform consistency and referential integrity checks.
Creates indexes, business views, partition views against the base data.
Transforms and merges the source data into the published data warehouse.
Archives the data that has reached the end of its captured life.
Note: A warehouse Manager also analyzes query profiles to determine index and aggregations are appropriate.
Query Manager
Query manager is responsible for directing the queries to the suitable tables.
By directing the queries to appropriate tables, the speed of querying and response generation can be increased.
Query manager is responsible for scheduling the execution of the queries posed by the user.
Stored procedures
Detailed Information
Detailed information is not kept online, rather it is aggregated to the next level of detail and then archived to tape. The detailed
information part of data warehouse keeps the detailed information in the starflake schema. Detailed information is loaded into the data
warehouse to supplement the aggregated data.
The following diagram shows a pictorial impression of where detailed information is stored and how it is used.
Note: If detailed information is held offline to minimize disk storage, we should make sure that the data has been extracted, cleaned
up, and transformed into starflake schema before it is archived.
Summary Information
Summary Information is a part of data warehouse that stores predefined aggregations. These aggregations are generated by the
warehouse manager. Summary Information must be treated as transient. It changes on-the-go in order to respond to the changing
query profiles.
It needs to be updated whenever new data is loaded into the data warehouse.
It may not have been backed up, since it can be generated fresh from the detailed information.
Relational OLAP
ROLAP servers are placed between relational back-end server and client front-end tools. To store and manage warehouse data, ROLAP
uses relational or extended-relational DBMS.
Multidimensional OLAP
MOLAP uses array-based multidimensional storage engines for multidimensional views of data. With multidimensional data stores, the
storage utilization may be low if the data set is sparse. Therefore, many MOLAP server use two levels of data storage representation to
handle dense and sparse data sets.
OLAP Operations
Since OLAP servers are based on multidimensional view of data, we will discuss OLAP operations in multidimensional data.
Roll-up
Drill-down
Pivot (rotate)
Roll-up
Roll-up performs aggregation on a data cube in any of the following ways:
By dimension reduction
Initially the concept hierarchy was "street < city < province < country".
On rolling up, the data is aggregated by ascending the location hierarchy from the level of city to the level of country.
When roll-up is performed, one or more dimensions from the data cube are removed.
Drill-down
Drill-down is the reverse operation of roll-up. It is performed by either of the following ways:
Drill-down is performed by stepping down a concept hierarchy for the dimension time.
Initially the concept hierarchy was "day < month < quarter < year."
On drilling down, the time dimension is descended from the level of quarter to the level of month.
When drill-down is performed, one or more dimensions from the data cube are added.
It navigates the data from less detailed data to highly detailed data.
Slice
The slice operation selects one particular dimension from a given cube and provides a new sub-cube. Consider the following diagram
that shows how slice works.
Here Slice is performed for the dimension "time" using the criterion time = "Q1".
Dice
Dice selects two or more dimensions from a given cube and provides a new sub-cube. Consider the following diagram that shows the
dice operation.
The dice operation on the cube based on the following selection criteria involves three dimensions.
Pivot
The pivot operation is also known as rotation. It rotates the data axes in view in order to provide an alternative presentation of data.
Consider the following diagram that shows the pivot operation.
In this the item and location axes in 2-D slice are rotated.
OLAP vs OLTP
Sr.No. Data Warehouse (OLAP) Operational Database (OLTP)
2 OLAP systems are used by knowledge workers such as OLTP systems are used by clerks, DBAs, or database professionals.
executives, managers and analysts.
5 Based on Star Schema, Snowflake, Schema and Fact Based on Entity Relationship Model.
Constellation Schema.
7 Provides summarized and consolidated data. Provides primitive and highly detailed data.
8 Provides summarized and multidimensional view of data. Provides detailed and flat relational view of data.
Points to Remember
ROLAP servers are highly scalable.
ROLAP tools analyze large volumes of data across multiple dimensions.
ROLAP tools store and analyze highly volatile and changeable data.
Database server
ROLAP server
Front-end tool.
Advantages
ROLAP servers can be easily used with existing RDBMS.
Disadvantages
Poor query performance.
Points to Remember:
MOLAP tools process information with consistent response time regardless of level of summarizing or calculations selected.
MOLAP tools need to avoid many of the complexities of creating a relational database to store data for analysis.
MOLAP server adopts two level of storage representation to handle dense and sparse data sets.
MOLAP Architecture
MOLAP includes the following components:
Database server.
MOLAP server.
Front-end tool.
Advantages
MOLAP allows fastest indexing to the pre-computed summarized data.
Helps the users connected to a network who need to analyze larger, less-defined data.
Disadvantages
MOLAP are not capable of containing detailed data.
MOLAP vs ROLAP
Sr.No. MOLAP ROLAP
3 MOLAP is best suited for inexperienced users, since it is very ROLAP is best suited for experienced users.
easy to use.
4 Maintains a separate database for data cubes. It may not require space other than available in the Data
warehouse.
Star Schema
Each dimension in a star schema is represented with only one-dimension table.
The following diagram shows the sales data of a company with respect to the four dimensions, namely time, item, branch, and
location.
There is a fact table at the center. It contains the keys to each of four dimensions.
The fact table also contains the attributes, namely dollars sold and units sold.
Note: Each dimension has only one dimension table and each table holds a set of attributes. For example, the location dimension table
contains the attribute set {location_key, street, city, province_or_state,country}. This constraint may cause data redundancy. For
example, "Vancouver" and "Victoria" both the cities are in the Canadian province of British Columbia. The entries for such cities may
cause data redundancy along the attributes province_or_state and country.
Snowflake Schema
Some dimension tables in the Snowflake schema are normalized.
Unlike Star schema, the dimensions table in a snowflake schema are normalized. For example, the item dimension table in star
schema is normalized and split into two dimension tables, namely item and supplier table.
Now the item dimension table contains the attributes item_key, item_name, type, brand, and supplier-key.
The supplier key is linked to the supplier dimension table. The supplier dimension table contains the attributes supplier_key
and supplier_type.
Note: Due to normalization in the Snowflake schema, the redundancy is reduced and therefore, it becomes easy to maintain and the
save storage space.
The following diagram shows two fact tables, namely sales and shipping.
The shipping fact table has the five dimensions, namely item_key, time_key, shipper_key, from_location, to_location.
The shipping fact table also contains two measures, namely dollars sold and units sold.
It is also possible to share dimension tables between fact tables. For example, time, item, and location dimension tables are
shared between the sales and shipping fact table.
Schema Definition
Multidimensional schema is defined using Data Mining Query Language (DMQL). The two primitives, cube definition and dimension
definition, can be used for defining the data warehouses and data marts.
define dimension time as (time key, day, day of week, month, quarter, year)
define dimension item as (item key, item name, brand, type, supplier type)
define dimension branch as (branch key, branch name, branch type)
define dimension location as (location key, street, city, province or state, country)
define dimension time as (time key, day, day of week, month, quarter, year)
define dimension item as (item key, item name, brand, type, supplier (supplier key, supplier type))
define dimension branch as (branch key, branch name, branch type)
define dimension location as (location key, street, city (city key, city, province or state, country))
define dimension time as (time key, day, day of week, month, quarter, year)
define dimension item as (item key, item name, brand, type, supplier type)
define dimension branch as (branch key, branch name, branch type)
define dimension location as (location key, street, city, province or state,country)
define cube shipping [time, item, shipper, from location, to location]:
To assist backup/recovery,
To enhance performance.
To Assist Backup/Recovery
If we do not partition the fact table, then we have to load the complete fact table with all the data. Partitioning allows us to load only
as much data as is required on a regular basis. It reduces the time to load and also enhances the performance of the system.
Note: To cut down on the backup size, all partitions other than the current partition can be marked as read-only. We can then put
these partitions into a state where they cannot be modified. Then they can be backed up. It means only the current partition is to be
backed up.
To Enhance Performance
By partitioning the fact table into sets of data, the query procedures can be enhanced. Query performance is enhanced because now
the query scans only those partitions that are relevant. It does not have to scan the whole data.
Horizontal Partitioning
There are various ways in which a fact table can be partitioned. In horizontal partitioning, we have to keep in mind the requirements
for manageability of the data warehouse.
Points to Note
The number of physical tables is kept relatively small, which reduces the operating cost.
This technique is suitable where a mix of data dipping recent history and data mining through entire history is required.
This technique is not useful where the partitioning profile changes on a regular basis, because repartitioning will increase the
operation cost of data warehouse.
Suppose a market function has been structured into distinct regional departments like on a state by state basis. If each region wants
to query on information captured within its region, it would prove to be more effective to partition the fact table into regional
partitions. This will cause the queries to speed up because it does not require to scan information that is not relevant.
Points to Note
The query does not have to scan irrelevant data which speeds up the query process.
This technique is not appropriate where the dimensions are unlikely to change in future. So, it is worth determining that the
dimension does not change in future.
If the dimension changes, then the entire fact table would have to be repartitioned.
Note: We recommend to perform the partition only on the basis of time dimension, unless you are certain that the suggested
dimension grouping will not change within the life of the data warehouse.
Points to Note
Partitioning Dimensions
If a dimension contains large number of entries, then it is required to partition the dimensions. Here we have to check the size of a
dimension.
Consider a large design that changes over time. If we need to store all the variations in order to apply comparisons, that dimension
may be very large. This would definitely affect the response time.
This technique makes it easy to automate table management facilities within the data warehouse.
Vertical Partition
Vertical partitioning, splits the data vertically. The following images depicts how vertical partitioning is done.
Normalization
Row Splitting
Normalization
Normalization is the standard relational method of database organization. In this method, the rows are collapsed into a single row,
hence it reduce space. Take a look at the following tables that show how normalization is performed.
16 sunny Bangalore W
64 san Mumbai S
30 5 3.67 3-Aug-13 16
35 4 5.33 3-Sep-13 16
40 5 2.50 3-Sep-13 64
45 7 5.66 3-Sep-13 16
Row Splitting
Row splitting tends to leave a one-to-one map between partitions. The motive of row splitting is to speed up the access to large table
by reducing its size.
Note: While using vertical partitioning, make sure that there is no requirement to perform a major join operation between two
partitions.
Account_Txn_Table
transaction_id
account_id
transaction_type
value
transaction_date
region
branch_name
We can choose to partition on any key. The two possible keys could be
region
transaction_date
Suppose the business is organized in 30 geographical regions and each region has different number of branches. That will give us 30
partitions, which is reasonable. This partitioning is good enough because our requirements capture has shown that a vast majority of
queries are restricted to the user's own business region.
If we partition by transaction_date instead of region, then the latest transaction from every region will be in one partition. Now the
user who wants to look at data within his own region has to query across multiple partitions.
Metadata acts as a directory. This directory helps the decision support system to locate the contents of a data warehouse.
Note: In a data warehouse, we create metadata for the data names and definitions of a given data warehouse. Along with this
metadata, additional metadata is also created for time-stamping any extracted data, the source of extracted data.
Categories of Metadata
Metadata can be broadly categorized into three categories:
Business Metadata - It has the data ownership information, business definition, and changing policies.
Technical Metadata - It includes database system names, table and column names and sizes, data types and allowed values.
Technical metadata also includes structural information such as primary and foreign key attributes and indices.
Operational Metadata - It includes currency of data and data lineage. Currency of data means whether the data is active,
archived, or purged. Lineage of data means the history of data migrated and transformation applied on it.
Role of Metadata
Metadata has a very important role in a data warehouse. The role of metadata in a warehouse is different from the warehouse data,
yet it plays an important role. The various roles of metadata are explained below.
This directory helps the decision support system to locate the contents of the data warehouse.
Metadata helps in decision support system for mapping of data when data is transformed from operational environment to data
warehouse environment.
Metadata helps in summarization between current detailed data and highly summarized data.
Metadata also helps in summarization between lightly detailed data and highly summarized data.
Metadata Respiratory
Metadata respiratory is an integral part of a data warehouse system. It has the following metadata:
Definition of data warehouse - It includes the description of structure of data warehouse. The description is defined by
schema, view, hierarchies, derived data definitions, and data mart locations and contents.
Business metadata - It contains has the data ownership information, business definition, and changing policies.
Operational Metadata - It includes currency of data and data lineage. Currency of data means whether the data is active,
archived, or purged. Lineage of data means the history of data migrated and transformation applied on it.
Data for mapping from operational environment to data warehouse - It includes the source databases and their
contents, data extraction, data partition cleaning, transformation rules, data refresh and purging rules.
Algorithms for summarization - It includes dimension algorithms, data on granularity, aggregation, summarizing, etc.
Metadata in a big organization is scattered across the organization. This metadata is spread in spreadsheets, databases, and
applications.
Metadata could be present in text files or multimedia files. To use this data for information management solutions, it has to be
correctly defined.
There are no industry-wide accepted standards. Data management solution vendors have narrow focus.
There are no easy and accepted methods of passing metadata.
Note: Do not data mart for any other reason since the operation cost of data marting could be very high. Before data marting, make
sure that data marting strategy is appropriate for your particular solution.
Consider a retail organization, where each merchant is accountable for maximizing the sales of a group of products. For this, the
following are the valuable information:
As the merchant is not interested in the products they are not dealing with, the data marting is a subset of the data dealing which the
product group of interest. The following diagram shows data marting for different users.
Given below are the issues to be taken into account while determining the functional split:
The merchant could query the sales trend of other products to analyze what is happening to the sales.
Note: We need to determine the business benefits and technical feasibility of using a data mart.
There are some tools that populate directly from the source system but some cannot. Therefore additional requirements outside the
scope of the tool are needed to be identified for future.
Note: In order to ensure consistency of data across all access tools, the data should not be directly populated from the data
warehouse, rather each tool must have its own data mart.
Data marts allow us to build a complete wall by physically separating data segments within the data warehouse. To avoid possible
privacy problems, the detailed data can be removed from the data warehouse. We can create data mart for each legal entity and load
it via data warehouse, with detailed account data.
The summaries are data marted in the same way as they would have been designed within the data warehouse. Summary tables help
to utilize all dimension data in the starflake schema.
Network Access
Note: Data marting is more expensive than aggregations, therefore it should be used as an additional strategy and not as an
alternative strategy.
Network Access
A data mart could be on a different location from the data warehouse, so we should ensure that the LAN or WAN has the capacity to
handle the data volumes being transferred within the data mart load process.
The structure of configuration manager varies from one operating system to another.
The interface of configuration manager allows us to control all aspects of the system.
Note: The above list can be used as evaluation parameters for the evaluation of a good scheduler.
Some important jobs that a scheduler must be able to handle are as follows:
Data load
Data processing
Index creation
Backup
Aggregation creation
Data transformation
Note: If the data warehouse is running on a cluster or MPP architecture, then the system scheduling manager must be capable of
running across the architecture.
Note: The Event manager monitors the events occurrences and deals with them. The event manager also tracks the myriad of things
that can go wrong on this complex data warehouse system.
Events
Events are the actions that are generated by the user or the system itself. It may be noted that the event is a measurable, observable,
occurrence of a defined action.
Hardware failure
A process dying
The most important thing about events is that they should be capable of executing on their own. Event packages define the procedures
for the predefined events. The code associated with each event is known as event handler. This code is executed whenever an event
occurs.
Scheduling
Database awareness
Backups are taken only to protect against data loss. Following are the important points to remember.
The backup software will keep some form of database of where and when the piece of data was backed up.
The backup recovery manager must have a good front-end to that database.
Being aware of the database, the software then can be addressed in database terms, and will not perform backups that would
not be viable.
Load manager
Warehouse manager
Query manager
Perform simple transformations into structure similar to the one in the data warehouse.
Fast Load
In order to minimize the total load window, the data needs to be loaded into the warehouse in the fastest possible time.
It is more effective to load the data into a relational database prior to applying transformations and checks.
Gateway technology is not suitable, since they are inefficient when large data volumes are involved.
Simple Transformations
While loading, it may be required to perform simple transformations. After completing simple transformations, we can do complex
checks. Suppose we are loading the EPOS sales transaction, we need to perform the following checks:
Strip out all the columns that are not required within the warehouse.
Warehouse Manager
The warehouse manager is responsible for the warehouse management process. It consists of a third-party system software, C
programs, and shell scripts. The size and complexity of a warehouse manager varies between specific solutions.
Backup/Recovery tool
SQL scripts
Creates indexes, business views, partition views against the base data.
Generates normalizations.
Transforms and merges the source data of the temporary store into the published data warehouse.
Archives the data that has reached the end of its captured life.
Note: A warehouse Manager analyzes query profiles to determine whether the index and aggregations are appropriate.
Query Manager
The query manager is responsible for directing the queries to suitable tables. By directing the queries to appropriate tables, it speeds
up the query request and response process. In addition, the query manager is responsible for scheduling the execution of the queries
posted by the user.
Stored procedures
It stores query profiles to allow the warehouse manager to determine which indexes and aggregations are appropriate.
The data from each analyst can be summarized and passed on to management where the different summaries can be aggregated. As
the aggregations of summaries cannot be the same as that of the aggregation as a whole, it is possible to miss some information
trends in the data unless someone is analyzing the data as a whole.
Security Requirements
Adding security features affect the performance of the data warehouse, therefore it is important to determine the security
requirements as early as possible. It is difficult to add security features after the data warehouse has gone live.
During the design phase of the data warehouse, we should keep in mind what data sources may be added later and what would be the
impact of adding those data sources. We should consider the following possibilities during the design phase.
Whether the new data sources will require new security and/or audit restrictions to be implemented?
Whether the new users added who have restricted access to data that is already generally available?
This situation arises when the future users and the data sources are not well known. In such a situation, we need to use the knowledge
of business and the objective of data warehouse to know likely requirements.
User access
Data load
Data movement
Query generation
User Access
We need to first classify the data and then classify the users on the basis of the data they can access. In other words, the users are
classified according to the data they can access.
Data Classification
Data can be classified according to its sensitivity. Highly-sensitive data is classified as highly restricted and less-sensitive data
is classified as less restrictive.
Data can also be classified according to the job function. This restriction allows only specific users to view particular data. Here
we restrict the users to view only that part of the data in which they are interested and are responsible for.
There are some issues in the second approach. To understand, let's have an example. Suppose you are building the data warehouse
for a bank. Consider that the data being stored in the data warehouse is the transaction data for all the accounts. The question here is,
who is allowed to see the transaction data. The solution lies in classifying the data according to the function.
User classification
Users can be classified as per the hierarchy of users in an organization, i.e., users can be classified by departments, sections,
groups, and so on.
Users can also be classified according to their role, with people grouped across departments based on their role.
Let's have an example of a data warehouse where the users are from sales and marketing department. We can have security by top-
to-down company view, with access centered on the different departments. But there could be some restrictions on users at different
levels. This structure is shown in the following diagram.
But if each department accesses different data, then we should design the security access for each department separately. This can be
achieved by departmental data marts. Since these data marts are separated from the data warehouse, we can enforce separate
security restrictions on each data mart. This approach is shown in the following figure.
If the data is generally available to all the departments, then it is useful to follow the role access hierarchy. In other words, if the data
is generally accessed by all If the data is generally available to all the departments, then it is useful to follow the role access hierarchy.
In other words, if the data is generally accessed by all
Audit Requirements
Auditing is a subset of security, a costly activity. Auditing can cause heavy overheads on the system. To complete an audit in time, we
require more hardware and therefore, it is recommended that wherever possible, auditing should be switched off. Audit requirements
can be categorized as follows:
Connections
Disconnections
Data access
Data change
Note : For each of the above-mentioned categories, it is necessary to audit success, failure, or both. From the perspective of security
reasons, the auditing of failures are very important. Auditing of failure is important because they can highlight unauthorized or
fraudulent access.
Network Requirements
Network security is as important as other securities. We cannot ignore the network security requirement. We need to consider the
following issues:
Are there restrictions on which network routes the data can take?
These restrictions need to be considered carefully. Following are the points to remember:
The process of encryption and decryption will increase overheads. It would require more processing power and processing
time.
The cost of encryption can be high if the system is already a loaded system because the encryption is borne by the source
system.
Data Movement
There exist potential security implications while moving the data. Suppose we need to transfer some restricted data as a flat file to be
loaded. When the data is loaded into the data warehouse, the following questions are raised:
If we talk about the backup of these flat files, the following questions are raised:
Do these backups need to be made to special tapes that are stored separately?
Some other forms of data movement like query result sets also need to be considered. The questions raised while creating the
temporary table are as follows:
We should avoid the accidental flouting of security restrictions. If a user with access to the restricted data can generate accessible
temporary tables, data can be visible to non-authorized users. We can overcome this problem by having a separate temporary area for
users with access to restricted data.
Documentation
The audit and security requirements need to be properly documented. This will be treated as a part of justification. This document can
contain all the information gathered from:
Data classification
User classification
Network requirements
Application development
Database design
Testing
Application Development
Security affects the overall application development and it also affects the design of the important components of the data warehouse
such as load manager, warehouse manager, and query manager. The load manager may require checking code to filter record and
place them in different locations. More transformation rules may also be required to hide certain data. Also there may be requirements
of extra metadata to handle any extra objects.
To create and maintain extra views, the warehouse manager may require extra codes to enforce security. Extra checks may have to be
coded into the data warehouse to prevent it from being fooled into moving data into a location where it should not be available. The
query manager requires the changes to handle any access restrictions. The query manager will need to be aware of all extra views and
aggregations.
Database design
The database layout is also affected because when security measures are implemented, there is an increase in the number of views
and tables. Adding security increases the size of the database and hence increases the complexity of the database design and
management. It will also add complexity to the backup management and recovery plan.
Testing
Testing the data warehouse is a complex and lengthy process. Adding security to the data warehouse also affects the testing time
complexity. It affects the testing in the following two ways:
It will increase the time required for integration and system testing.
There is added functionality to be tested which will increase the size of the testing suite.
Backup Terminologies
Before proceeding further, you should know some of the backup terminologies discussed below.
Complete backup - It backs up the entire database at the same time. This backup includes all the database files, control files,
and journal files.
Partial backup - As the name suggests, it does not create a complete backup of the database. Partial backup is very useful in
large databases because they allow a strategy whereby various parts of the database are backed up in a round-robin fashion
on a day-to-day basis, so that the whole database is backed up effectively once a week.
Cold backup - Cold backup is taken while the database is completely shut down. In multi-instance environment, all the
instances should be shut down.
Hot backup - Hot backup is taken when the database engine is up and running. The requirements of hot backup varies from
RDBMS to RDBMS.
Hardware Backup
It is important to decide which hardware to use for the backup. The speed of processing the backup and restore depends on the
hardware being used, how the hardware is connected, bandwidth of the network, backup software, and the speed of server's I/O
system. Here we will discuss some of the hardware choices that are available and their pros and cons. These choices are as follows:
Tape Technology
Disk Backups
Tape Technology
The tape choice can be categorized as follows:
Tape media
Tape stackers
Tape silos
Tape Media
There exists several varieties of tape media. Some tape media standards are listed in the table below:
DLT 40 GB 3 MB/s
8 mm 14 GB 1 MB/s
Scalability
Consider the server is a 48node MPP machine. We do not know the node to connect the tape drive and we do not know how to
spread them over the server nodes to get the optimal performance with least disruption of the server and least internal I/O
latency.
Connecting the tape drive as a network available device requires the network to be up to the job of the huge data transfer
rates. Make sure that sufficient bandwidth is available during the time you require it.
Tape Stackers
The method of loading multiple tapes into a single tape drive is known as tape stackers. The stacker dismounts the current tape when
it has finished with it and loads the next tape, hence only one tape is available at a time to be accessed. The price and the capabilities
may vary, but the common ability is that they can perform unattended backups.
Tape Silos
Tape silos provide large store capacities. Tape silos can store and manage thousands of tapes. They can integrate multiple tape drives.
They have the software and hardware to label and store the tapes they store. It is very common for the silo to be connected remotely
over a network or a dedicated link. We should ensure that the bandwidth of the connection is up to the job.
Disk Backups
Methods of disk backups are:
Disk-to-disk backups
Mirror breaking
These methods are used in the OLTP system. These methods minimize the database downtime and maximize the availability.
Disk-to-disk backups
Here backup is taken on the disk rather on the tape. Disk-to-disk backups are done for the following reasons:
Speed of restore
Backing up the data from disk to disk is much faster than to the tape. However it is the intermediate step of backup. Later the data is
backed up on the tape. The other advantage of disk-to-disk backups is that it gives you an online copy of the latest backup.
Mirror Breaking
The idea is to have disks mirrored for resilience during the working day. When backup is required, one of the mirror sets can be broken
out. This technique is a variant of disk-to-disk backups.
Note: The database may need to be shutdown to guarantee consistency of the backup.
Optical Jukeboxes
Optical jukeboxes allow the data to be stored near line. This technique allows a large number of optical disks to be managed in the
same way as a tape stacker or a tape silo. The drawback of this technique is that it has slow write speed than disks. But the optical
media provides long-life and reliability that makes them a good choice of medium for archiving.
Software Backups
There are software tools available that help in the backup process. These software tools come as a package. These tools not only take
backup, they can effectively manage and control the backup strategies. There are many software packages available in the market.
Some of them are listed in the following table:
Networker Legato
ADSM IBM
Omniback II HP
Alexandria Sequent
Does the package have client-server option, or must it run on the database server itself?
Does the package support easy access to information about tape contents?
What tape drive and tape media are supported by the package?
It is very difficult to predict what query the user is going to post in the future.
Performance Assessment
Here is a list of objective measures of performance:
Scan rates
It is of no use trying to tune response time, if they are already better than those required.
To hide the complexity of the system from the user, aggregations and views should be used.
It is also possible that the user can write a query you had not tuned for.
Note: If there is a delay in transferring the data, or in arrival of data then the entire system is affected badly. Therefore it is very
important to tune the data load first.
There are various approaches of tuning data load that are discussed below:
The very common approach is to insert data using the SQL Layer. In this approach, normal checks and constraints need to be
performed. When the data is inserted into the table, the code will run to check for enough space to insert the data. If sufficient
space is not available, then more space may have to be allocated to these tables. These checks take time to perform and are
costly to CPU.
The second approach is to bypass all these checks and constraints and place the data directly into the preformatted blocks.
These blocks are later written to the database. It is faster than the first approach, but it can work only with whole blocks of
data. This can lead to some space wastage.
The third approach is that while loading the data into the table that already contains the table, we can maintain indexes.
The fourth approach says that to load the data in tables that already contain data, drop the indexes & recreate them when
the data load is complete. The choice between the third and the fourth approach depends on how much data is already loaded
and how many indexes need to be rebuilt.
Integrity Checks
Integrity checking highly affects the performance of the load. Following are the points to remember.
Integrity checks need to be limited because they require heavy processing power.
Integrity checks should be applied on the source system to avoid performance degrade of data load.
Tuning Queries
We have two kinds of queries in data warehouse:
Fixed queries
Ad hoc queries
Fixed Queries
Fixed queries are well defined. Following are the examples of fixed queries:
regular reports
Canned queries
Common aggregations
Tuning the fixed queries in a data warehouse is same as in a relational database system. The only difference is that the amount of data
to be queried may be different. It is good to store the most successful execution plan while testing fixed queries. Storing these
executing plan will allow us to spot changing data size and data skew, as it will cause the execution plan to change.
Note: We cannot do more on fact table but while dealing with dimension tables or the aggregations, the usual collection of SQL
tweaking, storage mechanism, and access methods can be used to tune these queries.
Ad hoc Queries
To understand ad hoc queries, it is important to know the ad hoc users of the data warehouse. For each user or group of users, you
need to know the following:
Points to Note
It is important to track the user's profiles and identify the queries that are run on a regular basis.
It is also important that the tuning performed does not affect the performance.
If these queries are identified, then the database will change and new indexes can be added for those queries.
If these queries are identified, then new aggregations can be created specifically for those queries that would result in their
efficient execution.
Unit testing
Integration testing
System testing
Unit Testing
Each module, i.e., procedure, program, SQL Script, Unix shell is tested.
Integration Testing
In integration testing, the various modules of the application are brought together and then tested against the number of
inputs.
System Testing
The purpose of system testing is to check whether the entire system works correctly together or not.
Since the size of the whole data warehouse is very large, it is usually possible to perform minimal system testing before the
test plan can be enacted.
Test Schedule
First of all, the test schedule is created in the process of developing the test plan. In this schedule, we predict the estimated time
required for the testing of the entire data warehouse system.
There are different methodologies available to create a test schedule, but none of them are perfect because the data warehouse is very
complex and large. Also the data warehouse system is evolving in nature. One may face the following issues while creating a test
schedule:
A simple problem may have a large size of query that can take a day or more to complete, i.e., the query does not complete in
a desired time scale.
There may be hardware failures such as losing a disk or human errors such as accidentally deleting a table or overwriting a
large table.
Note: Due to the above-mentioned difficulties, it is recommended to always double the amount of time you would normally allow for
testing.
Media failure
Instance failure
Security - A separate security document is required for security testing. This document contains a list of disallowed operations
and devising tests for each.
Scheduler - Scheduling software is required to control the daily operations of a data warehouse. It needs to be tested during
system testing. The scheduling software requires an interface with the data warehouse, which will need the scheduler to
control overnight processing and the management of aggregations.
Disk Configuration. - Disk configuration also needs to be tested to identify I/O bottlenecks. The test should be performed
with multiple times with different settings.
Management Tools. - It is required to test all the management tools during system testing. Here is the list of tools that need
to be tested.
Event manager
System manager
Database manager
Configuration manager
Testing the database manager and monitoring tools - To test the database manager and the monitoring tools, they
should be used in the creation, running, and management of test database.
Testing database features - Here is the list of features that we have to test:
Querying in parallel
Scheduling software
Overnight processing
Query performance
Note: The most important point is to test the scalability. Failure to do so will leave us a system design that does not work when the
system grows.
As we have seen that the size of the open database has grown approximately double its magnitude in the last few years, it
shows the significant value that it contains.
As the size of the databases grow, the estimates of what constitutes a very large database continues to grow.
The hardware and software that are available today do not allow to keep a large amount of data online. For example, a Telco
call record requires 10TB of data to be kept online, which is just a size of one month’s record. If it requires to keep records of
sales, marketing customer, employees, etc., then the size will be more than 100 TB.
The record contains textual information and some multimedia data. Multimedia data cannot be easily manipulated as text data.
Searching the multimedia data is not an easy task, whereas textual information can be retrieved by the relational software
available today.
Apart from size planning, it is complex to build and run data warehouse systems that are ever increasing in size. As the
number of users increases, the size of the data warehouse also increases. These users will also require to access the system.
With the growth of the Internet, there is a requirement of users to access data online.
Hence the future shape of data warehouse will be very different from what is being created today.
Q: Define metadata?
A : Metadata is simply defined as data about data. In other words, we can say that metadata is the summarized data that leads us to
the detailed data.
Q: Define dimension?
A : The dimensions are the entities with respect to which an enterprise keeps the records.
Q: What is Normalization?
A : Normalization splits up the data into additional tables.
Q: Out of star schema and snowflake schema, whose dimension table is normalized?
A : Snowflake schema uses the concept of normalization.
Advertisements