Aster Data与Hadoop / Hive之间的区别

Question

All the components in Aster Data seems to have a similar component in Hadoop stack.

AFS => HDFS

SQL-MR => Hive

AMC => Ambari

ACT => beeline/hive terminal

Postgres for storing metadata => Hive can be configured to store metadata in any RDBMS

Queen/Worker => NameNode/Datanode

SQL-GR => Giraph

Apart from providing a package of pre-built functions, is there anything that is strikingly different and not available in Hadoop?

Answer 1

Your question is not bad, it's wrong. Likely coming from the Hadoop side you made the question using Hadoop architecture which is assembly of layered and/or integrated somewhat independent components, each with its own functional spec, configuration and execution environment, etc.

Why is that wrong? Because you wouldn't ask the same question if comparing Hadoop and Oracle or SAP databases, right?

Yes, each counterpart on Aster side matches up with Hadoop stack offering - the difference is with Aster there is no such stack (at least as of 6.x yet). Aster is database and analytical engine matching Hadoop stack functions but not components.

For example, ACT is a command-line utility similar to SQL*Plus for Oracle operating over client/server interface. It's nothing like Hive infrastructure on top of Hadoop. Aster File System comes as complex plug-able functional layer integrated into Aster software - not as independent framework and software that are HDFS and Hadoop.

The most striking difference goes for Aster analytical engine consisting of SQL, SQL/MR and SQL/GR. What it means is that there is no functional or operational gaps between data storage and its operators (SQL statements an SQL/MR or SQL/GR functions) operating on the data store - they live inside the same environment (configuration, execution, maintenance, support). For example, columnar and row-based tables are completely transparent for any operation performed on the them (barring constraints defined a priori and by design).

So, your analogy does explain Hadoop side of the equation without really giving proper due to Aster.

Answer 2

I think u unnecessarily went for nitty grity. The question was in conceptual terms. Say Netezza and Teradata MPP. In Basic concept they are same and they do the same kind of work but they are two different physical implementation with their own algo,storage,indexes etc.

On a very high level Hive and Aster has similarity as they both run map reduce on a distributed storage .

The Only difference on very high level is that at very high level Aster can run typical RDMS query as well as implicit map reduce where as Hive is only map reduce.