Before we delve into Data Partitioning, here’s a review of some fundamental database concepts. In terms of this article the focus will be on Oracle as the database.

Fundamental Database (ORACLE) concepts:
A Table:
(i) Stores structured data
(ii) is a database object housed in a segment.
A normal table (i.e. non-partitioned) is exactly one segment. A partitioned table will be made up of as many segments as it has partitions.

A Tablespace:
(i) is a logical container for segments
(ii) may be empty but it will most likely hold one or more segments
(iii) may be made up of one or more data files.

A Data file:
(i) is the physical operating system (OS) disk file that stores data. All data in an Oracle database ends up in a data file that is part of a file system configured by the OS or as a raw device managed by Oracle

A Data Block: is Oracle manages the storage space in the data files of a database in units called data blocks.

A Segment: is made up of extents and is the logical container for an object in an Oracle database

An Extent: is a set of contiguous data blocks

Data Partitioning (as it relates to the HVS)
Data Partitioning, in a nutshell, decomposes a database object (indexes, tables) into smaller more manageable pieces called partitions. The goal of data partitioning is to reduce the amount of data read for a particular SQL operation so that the overall response time is reduced. The data is organized using a partition key which is essentially a set of columns that determines in which partition a given row of objects will reside. It is important to note that the underlying database must also be partition enabled (which Oracle has been for quite a while). HVS uses i_partition as the partition key and HVS does Range partitioning, i.e., each partition is specified by the value of its partition key (i_partition). When using HVS, a SysObject, all its associated content objects, any local ACLs being referenced by the SysObject, and so on, will be assigned to the partition designation of the SysObject. For a LWSO, if the i_partition is NOT explicitly set, it will by default, be in the same partition as the parent, thus sharing the same i_partition id. Partition pruning is essentially directing a query to a subset of partitions rather than the entire table.

Data Partitioning leads to improved manageability (storage of data files on different physical drives), improved availability (partition independence) and optimized queries (using partition pruning).

What is HVS? Made broadly available in version 6.5 (you need to purchase the key) it is a generic term that refers to 3 broad areas of enhancements namely 1) Lightweight SysObjects (LWSO) 2) data partitioning and 3) scoping & batching.

This article will serve as an introduction to LWSO and the next article will deal with data partitioning and then batching & scoping. Stay tuned as the final post will deal with findings from base testing these enhancements.

LWSO share common attributes between parent objects providing the following benefits:

• normalizes the object model,
• reduces metadata storage space, and
• enhances repository scalability.

It is important to note that LWSO are subtypes of shareable types which are subtypes of dm_sysobject. Many LWSO can share the same parent type.

When LWSO have a parent object the object is called unmaterialized (or dematerialized) LWSO.
Unmaterialized LWSO will have identical values for all the attributes of its shared parent.
LWSO get converted from unmaterialized to materialized when certain actions such as checkout, link, unlink, or etc are applied to the LWSO.

A materialized object does not have a parent and can become a parent for another LWSO object. Materialized objects are full SysObjects and are called private parent if they do not have any child LWSO. Materialized objects can be converted to unmaterialized objects by re-parenting them. Re-parenting maps LWSO i_sharing_parent attribute back to a sharable parent converting the object to a unmaterialized object.

Documentum provides the following for the manipulating LWSO:
DQL:

• MIGRATE TO LITE (converts a dm_sysobject to LWSO),
• CREATE SHAREABLE TYPE,
• CREATE LIGHTWEIGHT TYPE.

DFC: IDfLightObject, IDfSession, IDfPersistentObject.

In the next article we will explore data partitioning and its implementation in the HVS.

It’s simple.  The new Content I/O Suite.  Armedia’s Content I/O Suite allows users to collect information from all of those billions of sources, migrate it to the centralized repository in the company, and then, if needed, deliver that information to the different web portals that may need to view the information later.  In fact, it can even send the information to multiple repositories all from the same interface.  So what if Human Resources uses Documentum, and Accounting uses Laserfiche, and some departments have their own collaborative eRoom set up, but you just merged with another company who uses SharePoint and FileNet.  What if one user could capture the information coming in from the website, or through email, or through regular snail mail, and deliver it into the appropriate system, without that one user ever having to be trained in each of those systems?  Sounds too good to be true, doesn’t it?  Check it out at http://armedia.com/products/suite/.

-Kristin Evans