If the rows in Teradata are not evenly distributed across all AMPs, it's
a big performance problem.
In Snowflake skewing is prevented by automatically redistributing data
scans between the compute nodes in case of bottlenecks.
What about statistics in Teradata and Snowflake? What is the difference?
In Teradata, statistics must be defined by the user and are usually
collected.
Snowflake collects statistics automatically when loading or updating
tables.
Teradata requires its own hardware. Is
this also the case with Snowflake?
Teradata always comes as a combination of hardware and software. Snowflake
is just software. The hardware is provided in the cloud. Amazon S3 is used as
data storage. Instances of Amazon EC2 provide the processing power.
Will Snowflake push Teradata out of the
market?
No, because Teradata is a mature system and is used where outsourcing to
the cloud is not possible. But Snowflake will take its market share.
Teradata is old but widespread and has a mature architecture. I see
Snowflake as more of a competitor to Hadoop, because it has excellent support
for semi-structured data and is much simpler than Hadoop.
Click here to know more about #TeradataSnowflake
No comments:
Post a Comment