bigobject service brochure

2
US: (650)784-0609 [email protected] www.thebigobject.com BigObject® Store Store and Compute Billions Data Records on a Commodity Machine BigObject® is a database built on our unique In-place Computing Model, capable of storing and computing up to a few billions of data records in seconds on a one-thousand-dollar computer. It outperforms traditional RDBMS by two to three orders of magnitude and offers great cost-performance value for those who want to implement real-time analytics. In-Place Computing Model The foundation of BigObject® store, it is unconventional in two ways: First, it obviates data-retrieval, the most time-consuming part of data compute process. Secondly, it decomposes data complexity by organizing data objects into Macro Data Structures in the way macromolecules function in living cells. These two innovations contribute to the significant progress in performance of BigObject®. How BigObject Differentiates from Others Extended Relation Model Allows various shapes of data objects besides tablesand meanwhile supports algebraic operators besides joinas well as operator overloading. Set-based algebra and programming play a main role in BigObject® Store. Data Locality Optimizes memory utilization by using row-based manner for fact tables and column-based for dimension tables. Memory layout for macro data structures is arranged hierarchically and semantically. Efficient Transform Algorithm Adopts an efficient algorithm ‘trans-join’ for transforming a table object (fact table) into a tree object while joining with one or more dimension table objects. Data-centric Programing Sends tiny programs to attach to data elements in macro data structures instead of moving data around and simply computes the results right on the data elements. Data Compression Implements an efficient compression algorithm to reduce data to 20% of the original size.

Upload: bigobject

Post on 07-Aug-2015

1.512 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: BigObject Service Brochure

US: (650)784-0609 [email protected]

www.thebigobject.com

BigObject® Store Store and Compute Billions Data Records on a Commodity Machine BigObject® is a database built on our unique In-place Computing Model, capable of storing and computing up to a few billions of data records in seconds on a one-thousand-dollar computer. It outperforms traditional RDBMS by two to three orders of magnitude and offers great cost-performance value for those who want to implement real-time analytics.

In-Place Computing Model The foundation of BigObject® store, it is unconventional in two ways: First, it obviates data-retrieval, the most time-consuming part of data compute process. Secondly, it decomposes data complexity by organizing data objects into Macro Data Structures in the way macromolecules function in living cells. These two innovations contribute to the significant progress in performance of BigObject®.

How BigObject Differentiates from Others

Extended Relation Model

Allows various shapes of data objects besides ‘tables’ and meanwhile supports algebraic operators besides ‘join’ as well as operator overloading. Set-based algebra and programming play a main role in BigObject® Store.

Data Locality

Optimizes memory utilization by using row-based manner for fact tables and column-based for dimension tables. Memory layout for macro data structures is arranged hierarchically and semantically.

Efficient Transform Algorithm Adopts an efficient algorithm ‘trans-join’ for transforming a table object (fact table) into a tree object while joining with one or more dimension table objects.

Data-centric Programing Sends tiny programs to attach to data elements in macro data structures instead of moving data around and simply computes the results right on the data elements.

Data Compression Implements an efficient compression algorithm to reduce data to 20% of the original size.

Page 2: BigObject Service Brochure

US: (650)784-0609 [email protected]

www.thebigobject.com

BigObject® Service Library Building Blocks for Real-time Analytic Applications

As a software developer, do you have a headache analyzing vast data collected from your product in real time? BigObject® Service Library is the solution for you. This service library is an easy-to-use and well-defined programming library built on top of the BigObject® store, perfect for you to implement real-time analytic functions for your applications.

Figure 1: BigObject® Service Library Architecture

The current release includes three major service modules: Multi-Dimensional Analysis Service, Association Analysis Service, and Loader Service. The main purpose of Multi-dimensional Analysis Service is to perform hierarchical comparison, ideal for business intelligence applications. The key difference from most OLAP systems is that it does not require creating cubes offline or any pre-processing but dynamic mapping datasets into the BigObject® store.

The code fragment above shows how easy to implement a comparative analysis function for a retail application that collects vast transaction logs. The result displays aggregated sales by product category.

The Association Analysis Service is to define, calculate, and query correlation between any two elements within a

certain dimension. It calculates an entire matrix of association factors for all element pairs timely and then allows instant query in run time. This service is designed for applications such as recommendation mechanism and knowledge discovery.

Via Loader Service BigObject® supports standard API to import data from external data sources. For example, the loader for MySQL is an ETL program that connects to and extracts data from a MySQL database, and inserts data into the BigObject® store.

Figure 2: Group by, roll up, sum 100 million data records

BigObject® outperforms relational databases in two to three orders of magnitude and enables truly interactive analysis. The service is extensible with a data-centric programming framework where the developers can write their C++ classes that compute intermediate and final results directly on the big objects.

Analyze Your MySQL Data in Real-time

BigObject® presents a complete offer for MySQL to accelerate the analysis in a timely manner.

Figure 3: BigObject® ETL Solution for MySQL

With the BigObject® solution, you don’t need to create cubes for analytics but map the tables in two systems in real-time.