| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Usenet -> c.d.o.server -> Re: fragmentation
On Sun, 25 Jan 1998 14:14:37 -0000, "Adrian Shepherd" <Adrian.Shepherd_at_BTINTERNET.COM> wrote:
>I disagree with some of this,
>
>Extents allocated at the beginning of a data file, with the next extents
>allocated at the end of the datafile imply poor physical location and will
>cause seek times to increase. Poor seek times wont effect small databases or
>small insert/select operations, they will impact on full table scans and
>defeat logical i/o tuning such as clustering tables etc...
There is an excelent technical article about various aspects of Oracle
db fragmentation available on
http://www.europa.com/~orapub/abs101.html. The title is "Avoiding a
Database Reorganization", author is Craig A. Schallahamer, an Oracle
insider.
The main conclusion regarding segment fragmentation is similar to what Wim Coekaerts has writen: in a production system a few extra I/Os per full table scan, caused by the multiple extents segment, have preacticaly no effect on performace! Remember, if your table has N extents, then no more then N adition I/O requests could be made during
its full scan because of its N extents.
And as far as tablespace fragmentation is concerned, again Wim's suggestion about making all extents in a given tablespace the same size to avoid TS fragmentation is correct.
>Wim Coekaerts wrote in message <34C917CE.A3E_at_pacbell.net>...
>>One thing to note on fragmentation, it is always good practise to have
>>objects with the same extent size in the same tablespace, this way you
>>basically cannot have fragmentation... Unless you consider the fact that
>>you have multiple extents belonging to one object, fragmentation.
>>
>>now, tests have shown that the fact that a table has multiple extents
>>does not necessarily degrade performance...(hardly ever does) also, it
>>does not really matter whether all those extents are lined up or not, so
>>I would not worry too much about having more than 1 extent per object,
>>as long as you are not having problems allocating free extents because
>>of different size extents...
>>
>>hope this was clear, heh...
>>
>>Wim
Regards,
Jurij Modic Republic of Slovenia jurij.modic_at_mf.sigov.mail.si Ministry of Finance ============================================================The above opinions are mine and do not represent any official standpoints of my employer Received on Mon Jan 26 1998 - 00:00:00 CST
 |
 |