Join Data Science Central
I echo Andrew's question about whether or not to expect more from built from scratch columnar databases. Anyone have insight? We are using Tableau against Oracle and the performance isn't great when hitting a billion row table.
Just a suggestion. It might be more appropriate to bracket Amazon Dynamo as the key-value pair NoSQL DB, rather than Amazon S3. S3 use cases are different - possibly for content delivery, analysis, archival, disaster recovery etc.
"SQL Server is a general-purpose database system that traditionally stores data in row format. To improve
performance on data warehousing queries, SQL Server 2012 adds columnar storage and efﬁcient batch-at-atime processing to the system. Columnar storage is exposed as a new index type: a column store index. In otherwords, in SQL Server 2012 an index can be stored either row-wise in a B-tree or column-wise in a column storeindex. SQL Server column store indexes are “pure” column stores, not a hybrid, because different columns arestored on entirely separate pages. This improves I/O performance and makes more efﬁcient use of memory."
Nice summary - thanks for posting! I wonder if you can provide any commentary on the new ColumnStore Indexes in SQL Server 2012. Is this providing true columnar storage? My own testing suggests approximately 10x speed increases summarizing the same 50 million record fact table with the same SQL query. That's a fantastic improvement to have but should I be looking for more from a database designed to handle columnar storage from scratch?
Welcome toData Science Central
Sign Upor Sign In
Added by Tim Matteson
© 2019 Data Science Central ®
Report an Issue |
Terms of Service
Please check your browser settings or contact your system administrator.