Operational Data Store is not that much of a popular term as that of a Data Warehouse. However, it can be an extremely useful application. ODS is an integration platform of data from various Source Systems, which is used for operational purpose. This is unlike a Data Warehouse, which is used for non-operational/processing purpose. ODS is more of a concept than an architecture. A Staging Area can be used as an ODS.
Following are different kinds of Operational Data Stores possible:
Non-Dynamic, read- only, one way:
These Operational Data Stores are updated typically on overnight basis, and are used where the real-time updating is not required. This kind of ODS is typically used for taking a single view on an entity (Customer, supplier..),which is not possible from a single source system.
For example Customer service, OR direct marketing may need to have a single Customer view to assess all the transactions the Customer has done. As soon as a Customer calls-up, his single view from (operational data store) comes up to help the DM take a call on Sales pitch. The difference between a Data Warehouse and ODS is that in Data warehouse you can take out a report on single Customer view. However, a Data Warehouse will not be open to an online access for thousands of Direct marketing agents accessing hundred of thousands of Customers. The de-normalized structure of Data Warehouse is not suited for high volume standard queries. A normalized ODS is much more suited for the same.
Dynamic, Read-only, One Way:
This ODS is same as the previous type, with an only difference that it is updated on online basis. This is required for the applications where one can’t wait for an overnight refresh. For example lets take a high volume assets management and trading company. One always wants to maintain a cap OR an exposure to a sector, industry, fund OR a commodity. If the investment systems are different, an ODS is required for the user to check on an online basis for the exposure, before he/she commits any further investment.
Dynamic, Read-only, both ways:
This ODS is used as a platform for the systems to exchange data, by routing the same through the ODS.
For example a CRM system (used for up-sell/cross-sell) might be interested in finding out the risk profile of an existing Customer, which exists in the risk management system. However, due to various reasons it cannot access it directly, and the Customer code used in two systems are different. An ODS will be used to extract data from both the system, assign a common Customer code. Apart from giving a single Customer view, it can also update the risk profile of the Customer (as picked from risk management system) to the CRM system.
Dynamic with Write-Back facility:
This is the closest an ODS can get to being an OLTP. This kind of ODS accepts information from the online user and updates it in its database and also on the databases of the source systems. For example a Direct Marketing personnel brings up a single Customer view for a Customer, makes a Sales pitch and enters the outcomes of the Sales call on the ODS, which updates the relevant CRM/Sales systems.
NOTE- A part of what an ODS can do is claimed to be achieved by Enterprise Application Integration (EAI) and Enterprise Information Integration Systems (EII). However, ODS has not lost its appeal primarily because EAI/EII tools typically are not completely successful to transform & cleanse a highly disparate & non-standard data. An ODS can achieve the same by following standard Extraction & Transformation practices.
Please refer ODS placement for Data Warehouse for more perspective on ODS.
|
