Hardware Requirements in Unicode Systems

Summary

Symptom

You want to know if the database size of your SAP system will increase during a Unicode conversion or a Combined Upgrade & Unicode conversion.
You are looking for an overview of CPU/Main Memory hardware resource requirements for Unicode SAP systems.

Other terms

Database encoding, UTF-8, UTF-16, hardware compression, database reorganization

Reason and Prerequisites

What you must take into account when planning a system conversion to Unicode.

Solution

The figures in this Note are average numbers based upon parallel benchmarking of non-Unicode and Unicode customers systems. Depending on the transactions in use, the figures might differ.


CPU: Average increase +10...30%
Depends on transaction mix, MDMP or Single Code Page, single byte or double byte languages. The figures also depend on the processor type. Modern processors showed 10% increase, older processors up to 30%.


Main Memory: Average increase +40...50%
Reason: Application servers are based on UTF-16

Network load: ~0%
Almost no change due to efficient compression.

Database size: Average increase

• UTF-8: +10%

           +10% is the observed maximum for larger systems (DB size > 200), smaller systems might grow more

• UTF-16: +20...60%

Note: Additional CPU/Main Memory hardware resource requirements on standalone DB servers must be provided by DB vendors.

The database size increase depends on various factors. The most important one is the database encoding scheme of the platform, e.g. UTF-16 requires remarkably more space than CESU-8. Other factors are:

• Database settings (page size, extent size)

• Applications/transactions running on this database (ratios: tables/indices, text data/binary data)

• Reorganization frequency. A Unicode conversion includes a DB reorganization, so - in UTF-8 case - DB growth is often compensated by the shrinking due to reorganization (especially the indexes)

• z/OS: Hardware compression (reduces DB size by approx. 40%)

• Languages in use, e.g. English (single byte) vs. Japanese (double byte)

Database size requirements for different platforms:

Database (Platform)   Encoding   Additional Storage Requirements
DB2 for AS/400	UTF-16     +10...20% *
DB2 for z/OS	UTF-16     -20...10% **
IBM DB2 for Linux,    UTF-8      -10%
Unix and Windows
MaxDB                UCS-2      +40...60%
MS SQL                UCS-2      +40...60%
Oracle                CESU-8    -10%
________________________________________________________________
*Small growth as biggest part of the ASCII based database is already Unicode
**SAP Unicode installations on z/OS always use hardware compression which overcompensates the DB growth due to Unicode

Average database growth measured in customer systems (sum of all sizes):
UTF-8 and CESU-8: -13% (more than 90% of the databases have shrunk)
UTF-16 and UCS-2: +30...40%