Provide data structure to Big Data to improve its performance

OLAP-on-Hadoop products bring data structure to big data, optimizing both performance and scalability, allowing users to query and analyze large volumes of information at the speed of thought.

OLAP seems like the antithesis of big data, as it recalls old school approaches to data management and analysis. But Some big data techniques can only be successful if everyone in an organization can benefit from them..

Some prefer to write Java or SQL against raw Hadoop data, and those who prefer to run SQL statements in Hadoop. But Imost prefer to access a data structure designed and built in advance by a data architect.

OLAP and Hadoop

It is a subset of analytical tools that seem to revive the old concept of online analytical processing. (OLAP) adapting it for Big Data. These tools achieve higher levels of performance and scalability than other solutions..

Products called OLAP-on-Hadoop size data and present it in a business-friendly format. Con OLAP, business users see metrics as common dimensions. For instance, executives can examine sales by product, region and time. With a click of the mouse, can exchange metrics, add or filter dimensions, pivot axes and explore from summary views of business performance to raw data. In other words, OLAP makes it easy for business users to analyze data presented the way they see the business.

To size the data, OLAP-on-Hadoop products require designers to model data to analyze, match, to integrate, clean and validate, before users consult it. Most OLAP products on Hadoop don't just pre-model data, but they materialize. Create new aggregate data structures that are loaded into memory or into high-performance columnar databases.. This is a writing scheme, what if you listen to the big data community, it is no longer in fashion, but certainly useful for querying big data.

Optimizing scalability and performance

Modeling, calculating and storing dimensional aggregates in advance, OLAP-on-Hadoop Products Achieve Scalability and Performance in a Big Data Environment. They solve the problem of scalability by keeping data on Hadoop where storage is cheap, what allows you Generate huge dimensional cubes with terabytes or more of data. And they solve the problem of performance by pre-aggregating data in high-speed data caching, providing analysis of speed of thought versus big data.

Many companies turn to OLAP-on-Hadoop products once they discover that other approaches to big data analytics don't work... Many want to use Hadoop to replace a data store, but they quickly find that the tools just don't respond when creating a board with a select list with 480 billion rows of data.

OLAP-on-Hadoop Product Dissection

There is a split in the OLAP-on-Hadoop community on the best way to size data in Hadoop:

• Physical cubes. Most of the products use MOLAP approach where they create a physically distinct dimensional data structure (or cube) inside Hadoop. Once the cube is created, performance is consistently fast. But IThe downside is that it can take hours to calculate the cube and the queries only run on the cube, not in the raw data that makes up the cube., which means that users are not getting the most up-to-date data possible.

• Virtual cubes. Other products create virtual cubes against raw data, which is more like a ROLAP approach. This provides users with access to real-time data and avoids the initial cost of building cubes. But potentially hurts query performance as you create dimensional views on the fly.

In addition to architectural approaches, OLAP-on-Hadoop products differ in other ways. Some have their own visualization tools, others don't, or give customers the option to use a third-party tool. Some materialize cubes in memory, while others store aggregates in a disk-based data structure.. Some store cubes in proprietary formats, while others use open source formats. Some have their own metadata repositories and analytics engines, while others use open source tools.