SD-SQL Server: a Scalable Distributed Database

1
SD-SQL Server a Scalable Distributed Database

Soror SAHRI http//ceria.dauphine.fr/soror/soror.h
tml soror.sahri_at_dauphine.fr
2
Objective

• Most of DBSs are parallel
• SQL Server, Oracle,..etc.
• DBSs do not provide scalable data partitioning.
• Application of the SDDS technology to DBSs.

3
SD-SQL Server Architecture
User/Application
User/Application
SD-SQL Server Managers
SD-SQL client
SD-SQL server
SD-SQL peer
Di
D1
D2
Split
SQL Server
SQL Server
SQL Server
4
Scripts

• Implementing the Server Side
• Management of a scalable table
• the Split
• Implementing the Client Side
• View adjustment
• Query Processing
• Performance Measurements

5
The Server side
DB_1
DB_2
Segment
Segment

Split
Split
SD_C
SD_RP
SD_C
SD_RP
Meta-tables
Meta-tables
SD_SITE
SD_SIZE
SD_SITE
SD_SIZE
SQL
SQL Server 1
SQL Server 2
6
The Client side
Distributed Partitioned Union all View

Db_1.Segment1
Db_2. Segment1
7
The achieved work The Server side

• The Split mechanism
• Split between 2 servers
• Split between several servers (5 servers)
• Management of a scalable table
• Create a scalable table
• Alter a scalable table
• Drop a scalable table
• Create a scalable view
• Management of Indexes on a scalable table

8
The achieved work The Client side

• The View Adjustment
• Query Processing, we allow a large number of
  queries
• Most of the SQL commands (aggregations) in the
• SELECT clause.
• Most of the SQL commands (joins) in the WHERE
• Several objects in the FROM clause tables,
  views
• Tables with aliases
• Scalable view at several levels (3 levels)

9
Performance Measurements

• Use of the SkyServer Database from the website
  http//research.microsoft.com/gray/SDSS
• Make of PhotoObj table of SkyServer DB as a
  scalable table. PhotoObj is a table of 158,426
  tuples (about 260 MB)

10
Performance Measurements
SD-SQL Server Segment split time
11
Performance Measurements
SD-SQL Server query (Q1) execution time
(Q1) SELECT TOP 10 objid FROM PhotoObj
WHERE objid NOT IN (SELECT objid
FROM _Photoobj)
12
Performance Measurements
SD-SQL Server query (Q2) execution time
(Q2) SELECT COUNT () FROM T1
13
Performance Measurements
Comparison between SQL Server and SD-SQL Server
execution time of query (Q2)
(Q3) SELECT COUNT () FROM PhotoObj
PhotoObj contains 1 segment of _PhotoObj
14
Future Works

• More Performance Measurements
• more comparisons between SQL Server and SD-SQL
  Server
• Measurements with more than 2 partitioned
  segments