++ed by:

10 PAUSE users
10 non-PAUSE users.

Rob Kinyon




This document is out-of-date. It describes an intermediate file format used during the development from 0.983 to 1.0000. It will be rewritten soon.


This is a document describing the internal workings of DBM::Deep. It is not necessary to read this document if you only intend to be a user. This document is intended for people who either want a deeper understanding of specifics of how DBM::Deep works or who wish to help program DBM::Deep.


DBM::Deep is broken up into five classes in three inheritance hierarchies.

  • DBM::Deep is the parent of DBM::Deep::Array and DBM::Deep::Hash. These classes form the immediate interface to the outside world. They are the classes that provide the TIE mechanisms as well as the OO methods.

  • DBM::Deep::Engine is the layer that deals with the mechanics of reading and writing to the file. This is where the logic of the file layout is handled.

  • DBM::Deep::File is the layer that deals with the physical file. As a singleton that every other object has a reference to, it also provides a place to handle datastructure-wide items, such as transactions.


DBM::Deep uses a tagged file layout. Every section has a tag, a size, and then the data.

File header

  • File Signature

    The first four bytes are 'DPDB' in network byte order, signifying that this is a DBM::Deep file.

  • File tag/size

    This is the tagging of the file header. The file used by versions prior to 1.00 had a different fifth byte, allowing the difference to the determined.

  • Version

    This is four bytes containing the file version. This lets the file format change over time.

  • Constants

    These are the file-wide constants that determine how the file is laid out. They can only be set upon file creation.

  • Transaction information

    The current running transactions are stored here, as is the next transaction ID.

  • Freespace information

    Pointers into the next free sectors of the various sector sizes (Index, Bucketlist, and Data) are stored here.


The Index parts can be tagged either as Hash, Array, or Index. The latter is if there was a reindexing due to a bucketlist growing too large. The others are the root index for their respective datatypes. The index consists of a tag, a size, and then 256 sections containing file locations. Each section corresponds to each value representable in a byte.

The index is used as follows - whenever a hashed key is being looked up, the first byte is used to determine which location to go to from the root index. Then, if that's also an index, the second byte is used, and so forth until a bucketlist is found.


This is the part that contains the link to the data section. A bucketlist defaults to being 16 buckets long (modifiable by the max_buckets parameter used when creating a new file). Each bucket contains an MD5 and a location of the appropriate key section.

Key area

This is the part that handles transactional awareness. There are max_buckets sections. Each section contains the location to the data section, a transaction ID, and whether that transaction considers this key to be deleted or not.

Data area

This is the part that actual stores the key, value, and class (if appropriate). The layout is:

  • tag

  • length of the value

  • the actual value

  • keylength

  • the actual key

  • a byte indicating if this value has a classname

  • the classname (if one is there)

The key is stored after the value because the value is requested more often than the key.


DBM::Deep is written completely in Perl. It also is a multi-process DBM that uses the datafile as a method of synchronizing between multiple processes. This is unlike most RDBMSes like MySQL and Oracle. Furthermore, unlike all RDBMSes, DBM::Deep stores both the data and the structure of that data as it would appear in a Perl program.


DBM::Deep attempts to be CPU-light. As it stores all the data on disk, DBM::Deep is I/O-bound, not CPU-bound.


DBM::Deep uses extremely little RAM relative to the amount of data you can access. You can iterate through a million keys (using each()) without increasing your memeory usage at all.


DBM::Deep is I/O-bound, pure and simple. The faster your disk, the faster DBM::Deep will be. Currently, when performing my $x = $db->{foo}, there are a minimum of 4 seeks and 1332 + N bytes read (where N is the length of your data). (All values assume a medium filesize.) The actions taken are:

1 Lock the file
1 Perform a stat() to determine if the inode has changed
1 Go to the primary index for the $db (1 seek)
1 Read the tag/size of the primary index (5 bytes)
1 Read the body of the primary index (1024 bytes)
1 Go to the bucketlist for this MD5 (1 seek)
1 Read the tag/size of the bucketlist (5 bytes)
1 Read the body of the bucketlist (144 bytes)
1 Go to the keys location for this MD5 (1 seek)
1 Read the tag/size of the keys section (5 bytes)
1 Read the body of the keys location (144 bytes)
1 Go to the data section that corresponds to this transaction ID. (1 seek)
1 Read the tag/size of the data section (5 bytes)
1 Read the value for this data (N bytes)
1 Unlock the file

Every additional level of indexing (if there are enough keys) requires an additional seek and the reading of 1029 additional bytes. If the value is blessed, an additional 1 seek and 9 + M bytes are read (where M is the length of the classname).

Arrays are (currently) even worse because they're considered "funny hashes" with the length stored as just another key. This means that if you do any sort of lookup with a negative index, this entire process is performed twice - once for the length and once for the value.



Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as the almighty BerkeleyDB. But it makes up for it in features like true multi-level hash/array support, and cross-platform FTPable files. Even so, DBM::Deep is still pretty fast, and the speed stays fairly consistent, even with huge databases. Here is some test data:

    Adding 1,000,000 keys to new DB file...

    At 100 keys, avg. speed is 2,703 keys/sec
    At 200 keys, avg. speed is 2,642 keys/sec
    At 300 keys, avg. speed is 2,598 keys/sec
    At 400 keys, avg. speed is 2,578 keys/sec
    At 500 keys, avg. speed is 2,722 keys/sec
    At 600 keys, avg. speed is 2,628 keys/sec
    At 700 keys, avg. speed is 2,700 keys/sec
    At 800 keys, avg. speed is 2,607 keys/sec
    At 900 keys, avg. speed is 2,190 keys/sec
    At 1,000 keys, avg. speed is 2,570 keys/sec
    At 2,000 keys, avg. speed is 2,417 keys/sec
    At 3,000 keys, avg. speed is 1,982 keys/sec
    At 4,000 keys, avg. speed is 1,568 keys/sec
    At 5,000 keys, avg. speed is 1,533 keys/sec
    At 6,000 keys, avg. speed is 1,787 keys/sec
    At 7,000 keys, avg. speed is 1,977 keys/sec
    At 8,000 keys, avg. speed is 2,028 keys/sec
    At 9,000 keys, avg. speed is 2,077 keys/sec
    At 10,000 keys, avg. speed is 2,031 keys/sec
    At 20,000 keys, avg. speed is 1,970 keys/sec
    At 30,000 keys, avg. speed is 2,050 keys/sec
    At 40,000 keys, avg. speed is 2,073 keys/sec
    At 50,000 keys, avg. speed is 1,973 keys/sec
    At 60,000 keys, avg. speed is 1,914 keys/sec
    At 70,000 keys, avg. speed is 2,091 keys/sec
    At 80,000 keys, avg. speed is 2,103 keys/sec
    At 90,000 keys, avg. speed is 1,886 keys/sec
    At 100,000 keys, avg. speed is 1,970 keys/sec
    At 200,000 keys, avg. speed is 2,053 keys/sec
    At 300,000 keys, avg. speed is 1,697 keys/sec
    At 400,000 keys, avg. speed is 1,838 keys/sec
    At 500,000 keys, avg. speed is 1,941 keys/sec
    At 600,000 keys, avg. speed is 1,930 keys/sec
    At 700,000 keys, avg. speed is 1,735 keys/sec
    At 800,000 keys, avg. speed is 1,795 keys/sec
    At 900,000 keys, avg. speed is 1,221 keys/sec
    At 1,000,000 keys, avg. speed is 1,077 keys/sec

This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and values were between 6 - 12 chars in length. The DB file ended up at 210MB. Run time was 12 min 3 sec.


One of the great things about DBM::Deep is that it uses very little memory. Even with huge databases (1,000,000+ keys) you will not see much increased memory on your process. DBM::Deep relies solely on the filesystem for storing and fetching data. Here is output from top before even opening a database handle:

  22831 root      11   0  2716 2716  1296 R     0.0  0.2   0:07 perl

Basically the process is taking 2,716K of memory. And here is the same process after storing and fetching 1,000,000 keys:

  22831 root      14   0  2772 2772  1328 R     0.0  0.2  13:32 perl

Notice the memory usage increased by only 56K. Test was performed on a 700mHz x86 box running Linux RedHat 7.2 & Perl 5.6.1.