Hi, there

I am starting to use berkeley db java edition for testing datamining
algorithm on a large dataset. I found it very slow to check the number
of records in my database using the cursor:

DatabaseEntry key = new DatabaseEntry();
DatabaseEntry data = new DatabaseEntry();
Cursor cursor = database.openCursor(null, null);
while(cursor.getNext(key, data, LockMode.DEFAULT) ==
OperationStatus.SUCCESS)
count++;
cursor.close();
return count;


Could anybody tell me an efficient method to get it? BTW, there are
about 8 million records in the database and take about 60G database. I
used linux with 4G memory. I use the java -server -Xms2000m -Xmx200m
when running.

Thanks a lot

Tong

Here is a technical tip which should help:

http://www.sleepycat.com/newsletters...10_count.shtml

Ron

Ron,

Thanks for your timely response. I tried the last approach by geting
the BTreeStats. The program has been running for 15 hours waiting for
the return of Database.getStats() and it is still running there.

Also, iterating through all the records using the cursor is even slower
than MySQL. I guess the reason could be the database (65G, 8miilion
records) was built by many add and delete operations. And the log files
didn't physically delete the records and it affects the iterating
speed.

Tong