Creating multiple MAIL.BOX databases

Server processes -- including server threads and the Router -- that write to MAIL.BOX require exclusive access to it. To ensure exclusive access, processes that write to or read from MAIL.BOX lock the database to prevent simultaneous access by other processes. Other processes trying to access the database must wait until the currently active process completes and unlocks the database before they can complete.

In most cases, a mail process locks MAIL.BOX for only an instant. However, longer wait times occur when the Router or another process reads or writes a large message. When there is a large amount of new mail -- for example, on a busy system with heavy mail traffic -- several server threads may try to deposit mail into MAIL.BOX while the Router attempts to read and update mail. Under heavy loads, such contention for a single MAIL.BOX database degrades performance.

You can improve performance significantly by creating multiple MAIL.BOX databases on a server. Using multiple MAIL.BOX databases removes contention for MAIL.BOX, allows multiple concurrent processes to act on messages, and increases server throughput. While reading one MAIL.BOX, the Router marks the database "in use" so other server threads trying to deposit mail move to the next MAIL.BOX. As a further benefit, having multiple MAIL.BOX databases provides failover in the event that one MAIL.BOX becomes corrupted.

When creating additional MAIL.BOX databases, consider placing each one on a separate disk. Because disk contention is rarely an issue for MAIL.BOX, placing each additional MAIL.BOX database on a different disk will not improve performance per se. However, distributing the databases across multiple disks does ensure greater availability in the event of a disk failure.

Creating a second MAIL.BOX database offers a large performance improvement over using a single MAIL.BOX. Depending on server mail traffic, adding a third and fourth MAIL.BOX database may further improve performance. However, the improvement gained with each additional MAIL.BOX is increasingly smaller.

You specify the number of MAIL.BOX databases on the Router/SMTP - Basics tab of the Configuration Settings document. Changes to the mailbox count take effect only after the next server restart.

After you configure a second MAIL.BOX database, you can use mail statistics to determine whether additional MAIL.BOX databases are needed.

After the new mailboxes are created, the old MAIL.BOX file is no longer used by the router. Any messages in MAIL.BOX need to be copied into one of the newly created mailboxes on the server. The file naming convention used when multiple mailboxes are created is MAIL1.BOX, MAIL2.BOX, ... MAILN.BOX, where N is the number of mail boxes specified in the Configuration Settings document.