Oracle Showstoppers

1
Oracle Showstoppers

• CMS Response

Zhen Xie CERN-EP/CMC
2
Implementation Issues

• Uniform software installation
• Oracle Enterprise Edition
• Database portability
• Server Solaris, Linux SuSE
• OCCI CC, gcc
• Data transportation
• Use Import/Export utilities
• transport data in binary format
• Oracle supported CPU and disk hardware
• Need more info

3
CMS Object Model Support

• Support Object with Identity.
• Support templates, STL, inheritance.
• Object navigation through REF data type. Scoped
  REF is the OID of the target object
• Objectivity Shallow copy and deep copy-like
  mechanism. Unclear.
• A view is a virtual table presenting data
  contained in one or more other tables
• A materialized view is a replica of a target
  table
• User can create new schema refering to public
  shared data

4
OCCI/Application Interface

• Proposed interface
• OCCI/OTT/application
• OTT data model drives object model, requires
  modification on existing classes. Dropped.
• Our prototype
• OCCI/class library/application
• Data type information extracted from header files
  using a C compiler front-end.
• No change on application level classes. Little
  manual work by user.

5
WAN/GRID Support

• Distributed Database environment
• One single logical database with different
  physical servers.
• Replicated Database environment
• Master site.
• Multitier materialized view sites.
• Sites are syncronized by refresh processes.
• An integrated Oracle Server techology

6
Security

• System security
• Default password-based authentication
• Support SSL protocol
• Data security
• DBA, user and other roles have different
  privileges
• Data encryption using SSL

7
OID Issues

• Logical OID
• System generated globally unique
• Primary-key based OID
• Logical to physical mapping
• Btree index
• OID Preservation in replication
• OID is preserved
• REF is preserved

8
Performance

• Storage
• DataIndex overhead wrt transient object sizelt10
  for object size gt 1kB
• Speed
• Started setting up some tests. A dedicated
  controlled environment needed.

9
Space Overhead () vs Object Size
80
800
10
Conclusions

• No obvious showstoppers found
• Required features exist
• No exact matcheses found at some points
• Some questions we are not able to answer.
• Obviously missing a middle layer between OCCI and
  existing software. We are on our own at this
  level.

11
If There Would Be a Second Phase

• An object oriented interface to the
  object-relational database remains our interest.
• More resource needed to evaluate Oracle as a
  database database deployment, performance,
  replication environment setup, etc.