Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Sparse data: efficient storage and retrieval in an RDBMS

Tags:

I have a table representing values of source file metrics across project revisions, like the following:

Revision FileA FileB FileC FileD FileE ...
1           45     3    12   123   124
2           45     3    12   123   124
3           45     3    12   123   124
4           48     3    12   123   124
5           48     3    12   123   124
6           48     3    12   123   124
7           48    15    12   123   124

(The relational view of the above data is different. Each row contains the following columns: Revision, FileId, Value. The files and their revisions from which the data is calculated are stored in Subversion repositories, so we're trying to represent the repository's structure in a relational schema.)

There can be up to 23750 files in 10000 revisions (this is the case for the ImageMagick drawing program). As you can see, most values are the same between successive revisions, so the table's useful data is quite sparse. I am looking for a way to store the data that

  • avoids replication and uses space efficiently (currently the non-sparse representation requires 260 GB (data+index) for less than 10% of the data I want to store)
  • allows me to retrieve efficiently the values for a specific revision using an SQL query (without explicitly looping through revisions or files)
  • allows me to retrieve efficiently the revision for a specific metric value.

Ideally, the solution should not depend on a particular RDBMS and should be compatible with Hibernate. If this is not possible, I can live with using Hibernate, MySQL or PostgreSQL-specific features.

like image 272
Diomidis Spinellis Avatar asked Jan 05 '09 15:01

Diomidis Spinellis


People also ask

What is sparse data in DBMS?

"Sparse" data, in which relations have many attributes that are null for most tuples, presents a challenge for an RDBMS. The traditional RDBMS approach to physical storage assumes that rows of tables have few null values, and that most tables will have tens of columns, and never more than a few hundred columns.

What is sparse data type?

Sparse data is data that has mostly unused elements (elements that don't carry any information ). Sparse Data: is a data set where most of the item values are zero. Dense Array: is the opposite of a sparse array: most of the values are not zero.

What is sparse column?

Sparse columns are ordinary columns that have an optimized storage for null values. Sparse columns reduce the space requirements for null values at the cost of more overhead to retrieve non-NULL values. Consider using sparse columns when the space saved is at least 20 percent to 40 percent.


1 Answers

This is how I might model it. I've left out the Revisions table and Files table as those should be pretty self-explanatory.

CREATE TABLE Revision_Files
(
    start_revision_number   INT NOT NULL,
    end_revision_number     INT NOT NULL,
    file_number             INT NOT NULL,
    value                   INT NOT NULL,
    CONSTRAINT PK_Revision_Files PRIMARY KEY CLUSTERED (start_revision_number, file_number),
    CONSTRAINT CHK_Revision_Files_start_before_end CHECK (start_revision_number <= end_revision_number)
)
GO

To get all of the values for files of a particular revision you could use the following query. Joining to the files table with an outer join would let you get those that have no defined value for that revision.

SELECT
    REV.revision_number,
    RF.file_number,
    RF.value
FROM
    Revisions REV
INNER JOIN Revision_Files RF ON
    RF.start_revision_number <= REV.revision_number AND
    RF.end_revision_number >= REV.revision_number
GO

Assuming that I understand correctly what you want in your third point, this will let you get all of the revisions for which a particular file has a certain value:

SELECT
    REV.revision_number
FROM
    Revision_Files RF
INNER JOIN Revisions REV ON
    REV.revision_number BETWEEN RF.start_revision_number AND RF.end_revision_number
WHERE
    RF.file_number = @file_number AND
    RF.value = @value
GO
like image 111
Tom H Avatar answered Sep 30 '22 18:09

Tom H