Webservices and huge datavolume

Hi,

Some questions about webservices and huge datavolumes. When sending, for 
example more than 20.000 java beans with a deep of two levels, an axis 
framework needs more than 3 minutes to parse them into a xml format (With 
a pentium 4 and two GB ram). The result is a document which is greather 
than 40 MB (the raw data needs 4 MB in the database). Also some frameworks 
stop parsing the data with a deep more than 100 elements (IBM).
The question is:
- Which techniques are provided to send so many elements?
- Can webservices provide a cache for the xml documents which are already 
parsed?
- Is there a solution to compress the size of such an document?
- How can a tree send by webservices and xml with a deep more than 100?


Thanks,
Gneisenau