Can I use Berkeley DB sqlite3 SQL API in tmpfs?

Question

Hi
To improve the efficiency of data access, I want to use Berkeley DB as in-memory SQL database. At the beginning, I assigned the argument of filename as “:memory:” when I called sqlite3_open(). Then I found in this way, it just built a private temporary in-memory database which could not be accessed between processes or between threads.
So I tried in the following way:
1.     Firstly, built a tmpfs on my OS:
mkdir /dev/shm/tmp
chmod 1777 /dev/shm/tmp
mount --bind /dev/shm/tmp /tmp/bdb_test
2.     Then I called the function sqlite3_open() with the argument /tmp/bdb_test/test.db, thus Berkeley DB would create the database test.db in memory, but it is transparent to BerkeleyDB.
Can I do this in the above way? Is there any potential risk?
Is there any other method to build a in-memory SQL database which could be shared between processes and threads?
Thanks!
Ps: My testbed information:
CentOS6
$ uname -a
Linux localhost.localdomain 2.6.32-71.el6.x86_64 #1 SMP Fri May 20 03:51:51 BST 2011 x86_64 x86_64 x86_64 GNU/Linux
Berkeley DB version: db-5.2.28
When I build BDB, I using the following options:
../dist/configure enable-sql_compat enable-test --with-tcl=/usr/lib64
Edited by: 882071 on 2011-9-6 下午7:27

Answer 1

Hi,
Then I called the function sqlite3_open() with the argument /tmp/bdb_test/test.db, thus Berkeley DB would create the database test.db in memory, but it is transparent to BerkeleyDB.
Can I do this in the above way? Is there any potential risk? I do not see any problem with using a tmpfs (temporary filesystem) to host the database. Of course, as you probably know, there are no durability guarantees with tmpfs, no recoverability for the database in this case.
Is there any other method to build a in-memory SQL database which could be shared between processes and threads?I do not have the resources to look into this now, but I would suggest trying the sqlite3_open_v2() function for opening the connections, with the shared cache mode (the SQLITE_OPEN_SHAREDCACHE flag), like this (making these calls in each process that opens a connection to the database):
#define TRUE 1
sqlite3_enable_shared_cache(TRUE);
sqlite3_open_v2(":memory:", &pDbHandle, SQLITE_OPEN_SHAREDCACHE | SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE, "");(look into the SQLITE_OPEN_FULLMUTEX and SQLITE_OPEN_NOMUTEX flags as well, for specifying the threading model for opening the database connection)
Regards,
Andrei