- This topic has 1 reply, 1 voice, and was last updated 5 years, 7 months ago by
.
-
Posts
-
-
-
OLTP
OLTP is quite related to RDBMS, handling structured data and particularly implemented for production transaction based applications. In that normalization is important to avoid redundancy of data. Low-latency is also a key consideration for such transactional applications so that the users can write many times i.e. insert,update, etc. fairly quickly. Hadoop framework is focused on providing high throughput rather than low latency.
It is better suited for write once and read many times data handling. Nevertheless, It can work well with all types of data from disparate systems viz., structured, semi-structured and unstructured. However we have to remember if the transactions yield “small files” then it would also be not advisable to handle such transactions on Hadoop yet. Hence it is not feasible yet to implement OLTP on Hadoop as both have different requirements to fulfill.OLAP
OLAP tools such as Microsoft Analysis Services provide a means of creating Unified Dimensional Model (a Cube) that makes for faster analyses of “large” structured data. It is ideally developed on top of a dimensional model which again is quite RDBMS. The users have then the capability of creating quick reports with aggregations that are pre-calculated and stored in memory. Hadoop provides a base to create faster analyses of “very large” data however it can replace OLAP tools in providing multidimensional analyses. Again please note that big data perhaps zettabytes will be better handled with Hadoop whereas if the requirements demand multidimensional analyses of records in kb size it is better placed in some dedicated OLAP tool.It seems to me there can be a Business Intelligence environment where OLTP, Enterprise Data Warehouses and OLAP tools shall be thriving well in harmony with the Hadoop ecosystem – each fulfilling different requirements.
Having said this and as we are learning, one can only appreciate the power of Hadoop and how beautifully it hides the complexity of the underlying framework and makes distributed processing of ginormous Big data fairly simple!
-
- You must be logged in to reply to this topic.
