Aster Data FAQ

What is a massively parallel data-application server?

A data-application server manages both data and applications as first-class citizens. Like a traditional database, it provides a very strong data management layer. Like a traditional application server, it provides a very strong application processing framework. It is architected on the principle of co-locating applications with data, thus eliminating data movement from the database to the application server. At the same time, it keeps the two layers separate to ensure the right fault-tolerance models - bad data will not crash the application, and vice-versa a bad application will not crash the database.

What are the technical characteristics of a massively parallel data-application server?

Architected to serve the needs of 'Big Data' and their users and analysts, the DAS follows three key design principles that distinguish it from a traditional database:

  • Scalability: a DAS can scale to petabyte data management, data loading and peta-op analytics without encountering performance bottlenecks
  • Continuous availability: a DAS avoids unplanned downtime of both applications and database tasks with fault tolerant features and eliminates planned downtime and performance degradation for routine administration
  • In-database processing: a DAS provides programmable analytics that analyzes and transforms data in-place without off-loading to a separate application processing platform

Why is the traditional data architecture experiencing performance problems?

The 20-year old data management legacy architectures or 'data pipelines' are inherently unfit for today's 'Big Data' applications because they rely on the movement of data from the place where data is stored to the place where data is processed. The traditional 3-tier architecture of Databases and Data Warehouses (to store data), Application Servers and Analytics Servers (to process data) and UI Servers (to present data) are under severe strain as applications pull terabytes to petabytes of data through the data pipeline for consumption. Moving large volumes of data through that data pipeline results in too much latency, does not allow for fresh data to reach the application fast enough to be considered in an analysis, and further SQL in itself places restrictions on the type of analysis that can be done. The larger the data volumes, the larger the time and effort needed to move it from one tier to another. As a result, the traditional data architecture experience, and will continue to experience, performance problems even as we invest more in hardware and services.

How do applications perform analysis on 'Big Data' on traditional data architectures?

The performance and latency problems of legacy data architectures or 'data pipelines' are so severe that application developers and analysts unconsciously compromise the quality of their analysis by avoiding "big data computation" - instead they first reduce the "big data" to "small data" - via SQL-based aggregations/windowing/sampling - and then perform computations on the "small data" set versus the entire 'big data' set. As a result, data analysts have had to compromise performance for lower-quality analytics resulting in sub-optimal decisions and lost revenue opportunities.

What principle of the traditional data architecture needs to change to handle 'Big Data'?

The traditional data architecture is built on the principle of moving data to applications for processing. Since data movement is hard and expensive, we must move application processing to data.

What is Aster Data nCluster 4.0?

Aster Data nCluster 4.0 is the industry's first massively-parallel data-application server. It provides a very strong data management layer - ANSI SQL interface, ACID transactions, Information Lifecycle Management, indexes, cost-based query optimizer, compression, security and other database features. It provides a very strong application processing framework - multiple language support, workload management, security, statistics collection, error logging and other application server features. It is architected to co-locate both data management and application processing as first-class citizens on the same infrastructure.

How do applications perform analysis on 'Big Data' on Aster Data nCluster 4.0?

Aster Data nCluster 4.0 enables applications (written in a variety of languages - Java, C, C++, Perl, Python, etc.) to be embedded within the database. Existing applications can be embedded without changing the application logic but by merely adding a few Map-Reduce functions. The applications that are run within Aster Data nCluster 4.0 can iteratively read/write/update all the data very efficiently without having to summarize/window/sample the data. The application is transparently parallelized across the servers that store the relevant data.

Do applications need to be re-written to embed them within Aster Data nCluster 4.0?

No, the application logic does not need to be re-written. The application needs to be augmented with a few Map-Reduce functions to ensure communication on the nCluster's MPP architecture. Our customers and partners have ported their existing applications (written in Java, C, C++, Perl) without any change in their application logic.

What are the advantages of Data-Application Servers built for running applications-within as embodied by Aster Data nCluster 4.0?

Since data accesses are local, and application processing is done on MPP infrastructure, applications can analyze terabytes to petabytes of data in minutes to seconds. It enables analysis to be done on fresh data (as soon as data enters the database) and it enables analysis to be done more frequently (as performance of analysis is vastly improved by as much as 10x-100x over traditional architectures). With this, you can empower your applications and users to rapidly generate insights and rule updates and gain competitive advantage.

Do you have any other questions? Please ask.

Top Picks
Data Sheet: Aster Data nCluster 4.0 Data-Application Server
Whitepaper: New MapReduce Whitepaper
Webcast: Bringing Big Data Analytics to the Enterprise - 11/12, with Merv Adrian