Introduction
To get some experience in migrating the system i will do a test migration. The source system runs on hpux ia64 with Oracle 10g. The target will be...
... my office PC, running XP :smile: ))
To increase the fun level I will use Oracle 11g (well there is another reason: OLTP compression, but i cover this somewhere else). Luckily 11g it is still available on 32-bit. Careful this is only experimental and...
... not supported!!https://weblogs.sdn.sap.com/weblogs/images/251794761/not_supported.png|height=29|alt=not supported by SAP|width=52|src=https://weblogs.sdn.sap.com/weblogs/images/251794761/not_supported.png|border=0!
I will only run the database on my PC, and connect a SAP instance to it, to test if it is running properly. I am planning to run the final installation on Win 2008 Server 64-bit.
h3. Distmon
The distmon package can be used to do distributed system copies, it is mostly used for unicode conversions on large databases. As I did some unicode testing lately, i have all the procedures ready. The distmon package can be obtained from note 855772 - Distribution Monitor
- the system is non-unicode, and i am not changing this, so i export/import with codepage 1100
- as the source db is only approximately 120gb in size, i don't do any performance optimizations, no package/table splitting
- coincidentally the system were i am running the distribution monitor is a linux system (SLES 10), so i am exporting from hpux on to linux, importing to win xp.
h3. Results
- The export took 4.5 hrs, which is pretty poor for 120gb. My best time was 40min, but as said earlier i did not do any optimization this time.
- Export size was 12gb
- The import took 8-10 hrs, but this isn't very accurate, as i needed to restart it a few times and i was also working on other things in parallel. Obviously the one single disk was completely busy the whole time, i had a feeling things improved notably when i added my external USB drive and configured a sapdata2 :smile: ))
- I was doing table and index compression during the import. I modified the DDLORA.TPL for the tables. To get the indexes compressed i had to create SMIGR_CREATE_DDL-style files. I made scripts for that, but the indexes were not created during the import. I had to run the script manually. First glance on the compression:
Segment | MB original | MB compressed |
EDI40 | 2571 | 1606 |
MARC | 2432 | 832 |
E071K | 2300 | 1216 |
WBCROSSGT0</td><td align="right">2948</td><td align="right"> 1216</td></tr><tr><td>E071K0
1753 | 336 |
GVD_SEGSTAT0</td><td align="right">1558</td><td align="right"><p> 104 </p></td></tr></tbody></table><p> </p><p>I am fully aware that these numbers do not say a lot. Due to the export/import and the PK ordering of the data will significantly influence the segment sizes. But i made a comparison in the new system with E071K0, uncompressed 953mb, with compress 4 only 336mb. "Codename SOL" - Intro / TOC
