7

This question is inspired by the discussion here and in the chat. Different vendors use different but similarly sounding terms for their respective implementation of columnar databases and tables. It is "columnstore" for SQL Server and MariaDB; "column store" (with a space) for Oracle; "column-organized tables" for Db2; etc.

On [dba.se] we have one tag, "columnstore" (and its synonym, "columnstore-index") that is being used mostly for SQL Server-related questions, however, sometimes questions about other vendor's products and vendor-neutral questions are tagged with it.

Do we want/need to have a vendor-neutral tag for such questions, e.g. "columnar-storage", or should we update guidance for the existing "columnstore" tag to explicitly allow/recommend tagging also non-SQL Server questions?

8

No

We don't need another tag for questions about column-oriented storage mechanisms in databases. Questions not related to SQL Server columnstore are rare and don't warrant a separate tag. Terminology itself is vague and imprecise, and adding another similar tag will create more confusion. We will explicitly mention in the "columnstore" tag wiki that it is OK to use it for questions related to vendors other than Microsoft and for non-vendor-specific questions about the technology in general, something along these lines:

For questions about column-oriented storage and access mechanisms in databases. Although the term "columnstore" itself is used by Microsoft and MariaDB for their respective implementations, you should use this tag for questions about other vendors' technologies (e.g. IBM Db2 column-organized tables) and for questions about the technology in general, not specific to a particular vendor.

1
  • I'd add that even vendors such as Oracle and Microsoft have multiple column-oriented features with wildly varying implementations/proper use cases. We'd have to create a tag for each vendor's technology (and meticulously tag questions where users forgot to add the specific technology they're using), and I just don't see that being practical. – bbaird Jan 7 at 14:57
0

Yes

To avoid further confusion we should limit the "columnstore" tag purpose to questions related to the DBMSes that use the term specifically, i.e. Microsoft SQL Server and MariaDB. The tag wiki should be updated accordingly.

For questions related to other vendors' implementations, as well as non-vendor-specific questions, we should create a new tag "columnar-storage" and indicate in its wiki that it should not be used for Microsoft SQL Server and MariaDB.

-1

There should be two tags:

  • columnar-store for MariaDB and friends - see a list here. This tag should be for systems which specifically store classic RDBMS tables but with the values stored column-by-column and not row-by-row. Otherwise, the syntax would be unchanged and classic SQL should be [ an | the ] access method (of choice). This is my understanding of how the MariaDB columnar store works.

    SQL is the most proven way of processing data. So, MariaDB ColumnStore is compatible with standard SQL using the MariaDB interface. Full SQL compliance means MariaDB ColumnStore works out-of-the-box with your existing business intelligence tools and SQL queries.

  • column-family (or possibly wide-column) for Cassandra and friends - see a list here. From here:

    Unlike a table in an RDBMS, different rows in the same column family do not have to share the same set of columns, and a column may be added to one or multiple rows at any time.

    The column-family databases are more typically distributed and have alternatives to SQL (which may or may not look similar), but are more NoSQL systems than classic RDBMS's.

Unfortunately, there is an element of overlap (and/or confusion) over the exact demarcation line between these two types of system - this site (formerly NoSQL-Database.org) lists some systems as "column family" which are shown as "columnar" stores on the wiki page referenced above.

My personal opinion is that we should create the two tags above (with usage guidance) and the extant columnstore tag (no usage guidance and only 126 questions) should be retired - with possibly wide-family as a synonym.

With respect to the SQL Server aspect, it's not my gig so I'll leave that to others to suggest suitable tags for its functionality. However, having quickly searched, I think that the columnstore-index tag should be reserved for the SQL Server functionality - but SQL Server guys may want to chip in on this?

3
  • I just wish whoever downvoted what I wrote had put in a reason - I don't mind and am not the vindictive sort - just want to know what the problem is? – Vérace Jan 7 at 20:32
  • i didn't vote in this question, but the purpose is, is there a need for a vendor indepedent column store tag, which will be used for general concepts abput column store, but yu plead for two more tags, as such you didn't answer the question, but made a new question fr more tags. – nbk Jan 9 at 11:28
  • 1
    @Vérace voting in meta means usually agreement or not with the answer. I suppose whoever downvoted does not like your idea of multiple tags for column oriented dbs – ypercubeᵀᴹ Jan 11 at 1:46

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .