- Why Transactions?
- Application Structure
- Opening the Environment
- Opening the Databases
- Recoverability and Deadlock Avoidance
- Repeatable Reads
- Transactional Cursors
- Nested Transactions
- Environment Infrastructure
- Deadlock Detection
- Performing Checkpoints
- Database and Log File Archival Procedures
- Log File Removal
- Recovery Procedures
- Recovery and Filesystem Operations
- Berkeley DB Recoverability
- Transaction Throughput
Generally, the speed of a database system is measured by the transaction throughput, expressed as the number of transactions per second. The two gating factors for Berkeley DB performance in a transactional system are usually the underlying database files and the log file. Both are factors because they require disk I/O, which is slow relative to other system resources such as CPU.
In the worst-case scenario:
Database access is truly random and the database is too large to fit into the cache, resulting in a single I/O per requested key/data pair.
Both the database and the log are on a single disk.
This means that for each transaction, Berkeley DB is potentially performing several filesystem operations:
Disk seek to database file
Database file read
Disk seek to log file
Log file write
Flush log file information to disk
Disk seek to update log file metadata (for example, inode information)
Log metadata write
Flush log file metadata to disk
There are a number of ways to increase transactional throughput, all of which attempt to decrease the number of filesystem operations per transaction:
Tune the size of the database cache. If the Berkeley DB key/data pairs used during the transaction are found in the database cache, the seek and read from the database are no longer necessary, resulting in two fewer filesystem operations per transaction. To determine whether your cache size is too small, see "Selecting a Cache Size."
Put the database and the log files on different disks. This allows reads and writes to the log files and the database files to be performed concurrently.
Set the filesystem configuration so that file access and modification times are not updated. Note that although the file access and modification times are not used by Berkeley DB, they may affect other programs--so be careful.
Upgrade your hardware. When considering the hardware on which to run your application, however, it is important to consider the entire system. The controller and bus can have as much to do with the disk performance as the disk itself. It is also important to remember that throughput is rarely the limiting factor, and that disk seek times are normally the true performance issue for Berkeley DB.
Turn on the DB_TXN_NOSYNC flag. This changes the Berkeley DB behavior so that the log files are not flushed when transactions are committed. Although this change will greatly increase your transaction throughput, it means that transactions will exhibit the ACI (atomicity, consistency, and isolation) properties, but not D (durability). Database integrity will be maintained, but it is possible that some of the most recently committed transactions may be undone during recovery instead of being redone.
If you are bottlenecked on logging, the following test will help you confirm that the number of transactions per second that your application does is reasonable for the hardware on which you're running. Your test program should repeatedly perform the following operations:
Seek to the beginning of a file
Write to the file
Flush the file write to disk
The number of times that you can perform these three operations per second is a rough measure of the number of transactions per second of which the hardware is capable. This test simulates the operations applied to the log file. (As a simplifying assumption in this experiment, we assume that the database files are either on a separate disk; or that they fit, with some few exceptions, into the database cache.) We do not have to directly simulate updating the log file directory information because it will normally be updated and flushed to disk as a result of flushing the log file write to disk.
Running this test program, in which we write 256 bytes for 1000 operations on reasonably standard commodity hardware (Pentium II CPU, SCSI disk), returned the following results:
% testfile -b256 -o1000 running: 1000 ops Elapsed time: 16.641934 seconds 1000 ops: 60.09 ops per second
Note that the number of bytes being written to the log as part of each transaction can dramatically affect the transaction throughput. The test run used 256, which is a reasonable size log write. Your log writes may be different. To determine your average log write size, use the db_stat utility to display your log statistics.
As a quick sanity check, the average seek time is 9.4 msec for this particular disk, and the average latency is 4.17 msec. That results in a minimum requirement for a data transfer to the disk of 13.57 msec, or a maximum of 74 transfers per second. This is close enough to the previous 60 operations per second (which wasn't done on a quiescent disk) that the number is believable.
An implementation of the previous example test program for IEEE/ANSI Std 1003.1 (POSIX) standard systems is included in the Berkeley DB distribution.