Discriminating Factors During Etl Phase of Data Warehouse


The extract transform load (ETL) is the trio of distinct functions that are integrated into a common programming tool, most when administering databases. The extort utility deciphers data from a particular resource and pulls out a preferred section of data which is converted into a preferred position. Lastly, the load utility is used to write the consequential data to a target database, which may or may not previously exist.

Surajit C. et.al (2006), ETL can be applied to achieve a transitory separation of data for reports or other reasons, or a more permanent data set may be realized for diverse objectives; including the population of a data mart or data warehouse; conversion from one database type to another and the migration of data from one database or platform to another. To be able to enhance the analysis of businesses, data warehousing should be loaded regularly.

To achieve this, data from one or more outfitted systems ought to be pulled out and pasted into the warehouse. Extraction, transformation, and loading (ETL) is the process by which data is pulled out from the source systems and brought into the data warehouse. The acronym ETL is rather too simplistic, in the sense that it excludes the transportation phase and implies that each of the other phases of the process is discrete.

ETL, therefore, refers to a wide-ranging process, and not necessarily three illustrated steps. The issue of integrating, rearranging, and consolidating data over many systems is one of the anathemas facing the Data warehouse atmospheres. This creates a new integrated information foundation for business acumen. Consequently, the data volume in data warehouse surroundings has propensities of bulkiness. Ronald Fagin, Phokion G.Kolaitis, Ravi Kumar, Jasmine Novak, D. Sivakumar, Anderw Tomkins (2005).

The Process

During the extraction, the preferred data is recognized and pulled out from many dissimilar bases; this comprises database systems and applications. Very often, it is not possible to classify a particular detachment of interest; this prompts the extraction of much more data than required. Classification of relevant data is normally done at a later point in time. Based on the system’s source competence i.e. the operating interface, most conversions happen during this drawn-out process. The dimensions of the data pulled out differ from hundreds of kilobytes up to gigabytes, in references to the source system and the business situation.

Zhimin Chen, Vivek R Narasayya (2005). The same is true for the time delta between two (logically) identical insertions; the period may vary between days/hours and minutes to near real-time. Web server log files for instance can easily become hundreds of megabytes in a very short period. Data is physically relayed after the extraction process has been achieved; it is transferred to the target system or an intermediate system for more processing. Most alterations are done during this process based on the chosen way of transportation. For instance, a SQL statement that authentically accesses a remote target through a gateway can concatenate two columns as part of the SELECT command. Zhimin Chen, Vivek R Narasayya (2005).

The emphasis in many of the examples

ETL Tools

The development and preservation of the ETL process are normally termed as one of the most intricate and reserve-demanding segments of a data warehouse project. Most data warehousing projects use ETL tools to manage this process. Oracle Warehouse Builder (OWB), for instance, offers ETL competencies and takes improvement of intrinsic database abilities. Other data warehouse builders create their ETL tools and processes, either inside or outside the database. Ralph Kimball and Margy Ross (2002).

However, the support of extraction, transformation, and loading other undertakings are fundamental for a victorious ETL accomplishment as part of the day-by-day maneuvers of the data warehouse and its shore up for further enrichments. The undertakings of designing the data warehouse and support and the data flow are characteristically addressed by ETL tools such as OWB. Oracle 9i for instance, is not an ETL tool and as such it offers a minimal solution for ETL.

Nevertheless, Oracle9i supports capabilities that are viable and are applicable in both ETL tools and customized ETL solutions. Oracle9i put forward performances for conveying data between Oracle databases, in support of converting large volumes of data, and for rapidly loading new data into a data warehouse. Ralph Kimball and Margy Ross (2002).

Daily Operations

The succeeding loads and conversions ought to be programmed and processed in a specific order. Based on the operational accomplishments or malfunctions, the outcome must be followed, and succeeding, unconventional procedures might be incorporated. Ralph Kimball and Margy Ross (2002).

Development of the Data Warehouse

Because a data warehouse is a living IT system, the basis and objectives might transform. It is obligatory therefore that these alterations are maintained and followed out through the lifespan of the system without overwriting or deleting the old ETL process information. To fabricate and maintain an echelon of dependence about the information in the warehouse, the process flow of each record in the warehouse can be restructured at any point in time in the future in a superlative case. We confirmed initially that Extract, Transform, and Load (ETL) is a course of action in statistics deport that is associated with pulling out of data from outside sources and converting it to fit business demands and consequently loading it into the end target. Surajit C. et.al (2006).


The primary ETL procedure is to pull out the data from the source systems. Most data warehousing projects consolidate data from diverse source systems. Independent systems may also employ diverse data formats, like relational databases and flat files; this may also include non-relational database structures such as IMS or other data structures such as VSAM or ISAM. Withdrawal transforms the data into a format for transformation processing. Conversely, the most profound stage of the extraction is the parsing of mined data, resulting in a make sure if data meets an anticipated prototype or composition, data is, however, discarded if it does not meet these requirements.


The conversion/ transformation phase applies a series of rules or functions to be extracted data from the source to derive the data to be loaded to the end target. Most data sources entail so minimally or even no manipulation of data. In other cases, one more of the following conversion types to meet the business and technological needs of the end target could be imperative. Choosing a particular column to load (choosing illogical editorials not to load) deciphering programmed values, thus, the source system stockpile 1 for male and 2 for female, but the warehouse stores M for male and F for female, this is well-known as computerized data rinsing out; no labor-intensive cleansing transpires during ETL. Surajit C. et.al (2006).

Indoctrinating of free-form values i.e. mapping ‘Male’ and ‘1’ and ‘Mr’ into M) deriving a newly computed value (i.e sale_amount =qty* unit_prive) embedding data from diverse sources (e.g. lookup, merge, etc). Rearranging or hinging (turning numerous columns into manifold rows or vice versa) splitting a column into several columns for instance putting of comma-disconnected list specified as a string in one column as individual values in different columns.

Applying any form of simple or complex data validation; if failed, a full partial or no rejection of the data, and thus no partial or all the data is passed over to the next phase based on the rule design and exception handling. The majority of the exceeding makeovers might result in outstanding i.e when a code-translation parses an unknown code in the extracted data. Ganesh R.et.al (2006).


The load phase loads the data into the end target, mostly being the data warehouse (DW). Based on the standards of the organization, this procedure ranges widely. Some data warehouses might weekly overwrite existing information with cumulative, updated data, while other DW (or even other parts of the same DW) might add new data in a historized form, e.g. hourly. The timing and dimension to substitute or append are strategic design choices dependent on the time available and the business needs. Additional intricate systems can maintain records and audit trails of all changes to the data loaded in the Data warehouse.

During the interfacing level with the database, the constrictions described in the database representation as well as in prompts activated upon data load apply, (e.g. distinctiveness, referential reliability, and obligatory fields) that as a result donates to the overall data eminence performance of the ETL procedure. Ganesh R.et.al (2006).


ETL processes can be quite intricate, and imperative operational anomalies might result from poorly designed ETL systems. The assortment of data values or data quality in an operational system could be beyond the expectations of designers at the time validation and transformation rules are specified. Data silhouetting of the source during data analysis is recommended to illustrate the data conditions that will need to be administered by transform regulations.

This leads to an amendment of validation rules explicitly and implicitly implemented in the ETL process. The data warehouse is typically fed asynchronously by a variety of sources which all serve a different purpose, resulting in e.g. diverse reference data. ETL is fundamental in transporting heterogeneous and asynchronous sources pulled out to a homogenous platform. Ralph Kimball and Margy Ross (2002).

The scalability of an ETL system across the lifetime of its application needs to be developed during the evaluation phase. This encompasses the volumes of data that will have to be processed within Service Level Agreements (SLAs). The time accessible to extract data from the source could easily be altered, which possibly denotes the equivalent quantity of data that may have to be developed in less time. Most ETL systems have to scale to procedural terabytes of data to update warehouses with tens of terabytes of data. Dong Xin, Zheng Shao, Jiawei Han, Hongyan Liu (2006).

Some ETL systems have to scale to process terabytes of data to update data warehouses with tens of terabytes of data. Increasing volumes of data may require designs that can scale from daily bunch to intra-day minute-batch to incorporation utilizing memorandum queues or synchronized transformed data, extended as of each day lot to intra-day micro-batch to amalgamation with communication queues or instantaneous amend data scope as of daily consignment to intra-day group assimilation capture (CDC) for uninterrupted conversion and reviews. Dong Xin, Zheng Shao, Jiawei Han, Hongyan Liu (2006).

Parallel Processing

Contemporary expansion in ETL software is the accomplishment of the corresponding dispensation. This has enabled myriad methods to progress the holistic performance of the ETL process when handling large quantities of data. Ralph Kimball and Margy Ross (2002).

ETL has about 3 implemented parallelisms

Data split

As a result of dividing a single chronological sleeve into less important data files to make available comparable access. Ralph K, et al (1998).


Ornamental the coincident scuttle of several components on the same data stream, for instance, would be looking up a value on record 1 at the same time as embedded two fields on the record.


The simultaneous running of multiple processes on different data streams in the same job. Sorting one input file while performing a de-duplication on another file would be an example of component parallelism. Three parallelisms are mostly integrated into on holistic whole, augmented to this complexity is making sure that the data being uploaded is virtually consistent. Because multiple source databases all have myriad update cycles, the ETL system may be obligated to inhibit definite data awaiting the coordination of all the sources. Ralph K, et al (1998).


While an ETL process can be created using almost any programming language, creating them from scratch is quite complex. Increasingly, companies are buying ETL tools to help in the creation of ETL processes. A good ETL tool must be able to communicate with the many different relational databases and read the various file formats used throughout an organization. ETL tools have started to migrate into Enterprise Application Integration, or even Enterprise Service Bus, systems that now cover much more than just the extraction, transformation, and loading of data. Many ETL vendors now have data profiling, data quality, and metadata capabilities. Michael D. M, Jignesh M P, William I.G (2006).

ETL Data deliberations

Apart from their implementation, several design considerations are common to all ETL systems.


Extraction, transformation load constitutes modular elements that perform discrete tasks. This encourages reuse and creates them easy to transform when implementing alterations in response to business and data warehouse changes. Monolithic systems could be avoided.


ETL structures ought to warranty uniformity of data at the point of loading it into the data warehouse. A wide-ranging data load should be delighted as a distinct rational business deal, either the entire data load is booming or the complete load is turn rounded back. However, the load is normally a single tangible transaction in most private schemes, while on the other hand, it is a series of connections. Regardless of the physical implementation, the data load should be treated as a single logical business deal. Michael D. M, Jignesh M P, William I.G (2006).


ETL systems should be developed to meet the needs of the data warehouse and to accommodate the source data environments. It may be appropriate to accomplish some transformations in text files and some on the source data system; others may require the development of custom applications. A variety of technologies and techniques can be applied, using the tool most appropriate to the individual task of each ETL functional element. Jens-Peter, Dittrich, D.K, Kreutz A (2005).


Speed is an integral component for reliable and ETL systems. In due course, the available time window for ETL processing is governed by data warehouse and source system schedules. a Quantity of data warehouse building blocks could have a huge processing window (days), while others may have a very limited processing window (hours) despite the consequences of the accessible time, it is significant that the ETL system performs as quickly as achievable. Ralph K, et al (1998).


RTL structures ought to be able to work with a myriad variety of data in various formats. An ETL structure that only works with a unified type of source data is inadequate.

Meta Data Management

ETL systems are arguably the single most significant source of Metadata about both the data in the data warehouse and data in the source system. Finally, the ETL process itself generates resourceful Meta that should be preserved and evaluated periodically. Ralph Kimball and Margy Ross (2002),


Ralph Kimball and Margy Ross (2002). The Data Warehouse Toolkit: The Complete Guide to Dimensional Modeling (Second Edition), Wiley.

Ralph K, et al (1998). The Data Warehouse Lifecycle Toolkit. Wiley.

Surajit C. et.al (2006). Data warehouse: Automated Sales Lead Generation from Web Data.

Ganesh R.et.al (2006). Industry, data warehousing, and analysis; C-cubing; Efficient Computation of closed cubes by Aggregation-based checking.

Dong Xin, Zheng Shao, Jiawei Han, Hongyan Liu (2006). Data warehouse: Detecting Duplicates in Complex XML Data.

Melanie Weis, Felix Naumann (2006). Poster, data integration, data mining, data warehousing.

Michael D. M, Jignesh M P, William I.G (2006). Efficient Skyline Computation: Data warehousing.

Jens-Peter, Dittrich, D.K, Kreutz A (2005). Bridging the gap between OLAP and SQL; industrial data warehousing.

Zhimin Chen, Vivek R Narasayya (2005). Efficient Computation by Multiple Group by Queries.

Ronald Fagin, Phokion G.Kolaitis, Ravi Kumar, Jasmine Novak, D. Sivakumar, Anderw Tomkins (2005), Efficient Implementation of Large –scale Multi Structural Databases.