I'm building a hash-based databased where key and data are both 8-byte
integers. There is roughly 50M entries, so the actual data size is
about 800M, but the produced db file is 2.1G. Why does it use so much
space for a hash table which is supposed to be a compact data
structure?

Here's my configuration:

db.set_cachesize(10, 0 * 1024 * 1024, 1); //If I set it to 2G,
then it becomes terribly slow after 20M entries. I don't understand
why.
u_int32_t pageSize = 4 * 1024; // block size on my machine
db.set_pagesize(pageSize);
db.set_h_ffactor((pageSize - 32) / (keySize + 8 + 8)); //
according to recommendation
db.set_h_nelem(50000000);
db.open(NULL, "test.db", NULL, DB_HASH, DB_CREATE |
DB_TRUNCATE, 0);

Thanks!

btw, the keys are first 8 bytes of MD5 digest of a word

On 6 22 , 9 13 , Smalltalk80 <timtan... (AT) gmail (DOT) com> wrote:
The smaller the databseEntry,the lager space will be used in bdb.
for example:
if you set the key as 8-byte integer,the data as 20 8-byte
integers,then 2.5M entries will use much less space than before.