DLT4000 compression and encrypted database

[Doug Monroe]

Hello,

This weekend we upgraded our database application to one that encrypts the data files in the file systems. Now, backups no longer fit on the media. In fact, a 19GB file system no longer fits on a 20/40 tape.

Details:

LVM, vxfs file system, one very large file and a few small ones.
Quiescent file system.
vxdump 0bf 32 /dev/rmt/0mn /dev/vg03/datafs
DLT4000 HVD tape drive.
rp7410 system.
The database is Cache 5.1
The only change has been the database upgrade.

My guess is that the encryption of the database file is foiling the compression algorithm in the tape drive. Anybody encountered this? Do I need to get a different tape system so I don't have to swap tapes for my backup, or is there another solution?

Thanks,

Doug

[A. Clay Stephenson]

Bear in mind, that a "19GiB" filesystem could actually be bigger than 19GiB if sparse files are in play --- and sparse files are very common in databases. It's a bit surprising that the native 20GiB capacity isn't big enough but depending upon the backup used, the metatdata overhead could account for the extra 1GiB and that doesn't take sparse files into account. In any event, never, ever assume that you have anything more than native capacity. The compressed capacity assumes a marketing 2:1 compression ration. The actual compression ratio is extremely data dependent and encrypted data is generally very difficult to compress.

[Steven E. Protter]

Shalom,

Ultrium tape drives are the solution. Their starting capacity is 200 GB for Ultrium 1.

Compression ratios on encrypted data are not likely to exceed 1.5 to one.

SEP

[Steven Schweda]

> My guess is [...]

Most likely.

Many encryption schemes include some form of
compression, as it helps to scramble the
bits. For example:

http://www.gzip.org/

: [...] PGP already incorporates the gzip
: compression code, so use PGP if you need
: compression plus strong encryption.

If your new encryption system is not doing
this, perhaps it should (depending, of
course, on performance effects and available
CPU cycles).

In any case, encrypted data tend to appear
pretty random, which tends to defeat typical
compression schemes.