vBulletin Search Engine Optimization
| |||||||
| Register | FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
| ||||
| I'm going to use cursor load to load 200GB data in my production database. My database has 2 partitions. but I cannot find more info in the manual about this cursor load. I'm concern about loading 200GB , this large amount of data, is there any limitation? I'm not sure from perfomance point of view, is there any difference between loading from cursor and loading from flat files on the same box? is there any memeory limitation for cursor load ? If anybody could share some experience on it, I would be very appreciated. Thanks. |
| |||
| My experience is that it's faster to load from a file than a cursor. If you are loading into an MDC add an order by clause to the select to match the orgainize by in the MDC. Make sure you use the +c option when running the script, or somehow else make sure that you don't commit between declaring and opening the cursor and starting the load or the load will fail to start. |
| |||
| jane wrote: > I'm going to use cursor load to load 200GB data in my production > database. > My database has 2 partitions. > > > but I cannot find more info in the manual about this cursor load. > > I'm concern about loading 200GB , this large amount of data, is there > any limitation? > > I'm not sure from perfomance point of view, is there any difference > between loading from cursor and loading from flat files on the same > box? > > is there any memeory limitation for cursor load ? Maybe you're missing something. Loading from a cursor allows you to declare a cursor for a select statement (i.e. data is already in the database) and then load from the cursor directly into another table. This saves you the step of having to export the data to a flat file, and is MUCH more efficient than 'insert into ... select ...' when loading large amounts of data. If your data is in a flat file, you do not have the option of loading from a cursor. There aren't memory restrictions to this -- load is simply reading the cursor as it traverses the result set. |
| |||
| Thanks for the reply... about '+c' option, I'm wondering that's that for? I did not find it in manural... I tried to do the declare and load without '+c' , it run successfully.. but I used small amount of data... could you give me some more explain on that.. Thanks. |
| |||
| HI, Pyle, also for compare loading from file or load from cursor.. if load from file, there must have physical read I/O happened, but load from cursor , there is not physical read I/O happened for read from cursor.. I suppose load from file should slower than load from cursor from I/O point of view.. you said load from file is faster than load from cursor, do you have any idea why? |
| |||
| HI, Pyle, also for compare loading from file or load from cursor.. if load from file, there must have physical read I/O happened, but load from cursor , there is not physical read I/O happened for read from cursor.. I suppose load from file should slower than load from cursor from I/O point of view.. you said load from file is faster than load from cursor, do you have any idea why? |
| ||||
| "jane" <esthershensh@yahoo.com> a écrit dans le message de news:1128458071.601677.59900@g44g2000cwa.googlegro ups.com... > I'm going to use cursor load to load 200GB data in my production > database. > My database has 2 partitions. > > > but I cannot find more info in the manual about this cursor load. > > I'm concern about loading 200GB , this large amount of data, is there > any limitation? > > I'm not sure from perfomance point of view, is there any difference > between loading from cursor and loading from flat files on the same > box? > > is there any memeory limitation for cursor load ? > > > If anybody could share some experience on it, I would be very > appreciated. > > Thanks. Hi, we use quite often load from cursor, in dpf context, because it is easier to alter big fact tables this way: we first load to a table in a temp schema, recreate the fact table fitting our needs (like transformation to MDC), and load from the temp schema. This way, the DB supports the duplication of the data - we are not sure that export will fit in FS, and doing twice load from cursor is faster than exporting + loading from file. You might consider increase UTIL_HEAP_SZ, especially if you load MDC. HTH, Jean-Marc |