Introduction to Storage Deduplication for the SQL Server DBA

1
Introduction to Storage Deduplication for the SQL
Server DBASQLDBApros
SQL Server DBA Professionals
2
Introduction to deduplication

• SQL Server DBAs across the industry are
  increasingly facing requests to place database
  backups on deduplication storage while also
  considering whether or not compressing backups
  while using deduplication storage is a good idea.

3
Let us explain

• Deduplication is not a new term it has been
  circulating widely the past few years as major
  companies began releasing deduplicating storage
  devices. Deduplication simply means not having to
  save the same data repeatedly.

4
Imagine this

• Imagine the original data as separate files the
  same files multiply as multiple users save them
  to their home directories, creating an excess of
  duplicate files that contain the same
  information. The object of deduplication is to
  store unique information once, rather than
  repeatedly.
• In this example, when a copy of a file is already
  saved, subsequent copies simply point to the
  original data, rather than saving the same
  information again and again.

Download Your Free Trial of our Backup
Compression Tool
5
Chunking

• Files enter the process intact, as they would in
  any storage. They are then deduplicated, and
  compressed.
• On many appliances, data is processed in real
  time. Unlike the file example above,
  deduplication appliances are more sophisticated.
• Most deduplication appliances offer an
  architecture wherein incoming data is
  deduplicated inline (before it hits the disk) and
  then the unique data is compressed and stored.
• Sophisticated algorithms break up files into
  smaller bits. This is called chunking. Most
  chunking algorithms offer variable block
  processing and sliding windows that allow for
  changes within files without much loss in
  deduplication.

Download Your Free Trial of our Backup
Compression Tool
6
Fingerprints

• For instance, if a one-line change is made in one
  of several nearly identical files, sliding
  windows and variable block sizes break up the
  larger file into smaller ones and effectively
  store the small pieces of changed information
  without thinking the one line of change means the
  entire file is new.
• Each chunk of data gets hashed, which you can
  think of as a fingerprint. If the system
  encounters a piece of data bearing a fingerprint
  it recognizes, it merely updates the file map and
  reference count without having to save that data.
  
• Unique data is saved and compressed.

Download Your Free Trial of our Backup
Compression Tool
7
Reduce storage needs

• Both target and source-side (at the server
  itself) deduplication help to reduce demand for
  storage by eliminating redundancies in data and
  reducing the amount of data being sent across the
  network.
• In source-side deduplication, vendor APIs can
  quickly query a storage device to see if the
  chunks of data already reside on the storage
  rather than sending all of the bits across the
  network for the storage device to process.

Download Your Free Trial of our Backup
Compression Tool
8
Replication

• Replication is another key feature.
• Replication features found in todays
  deduplication appliances are a boon for DBAs.
• It allows them to ensure that backup data from
  one data center can be easily replicated to
  another by moving only the needed deduplicated
  and compressed chunks.

Download Your Free Trial of our Backup
Compression Tool
9
Rehydration

• Deduplication does mean that the file will
  eventually have to be rehydrated (put back
  together) if its needed.
• Bits are read, decompressed, and reassembled.
• Read speed may slow down compared to
  non-deduplicated storage because of needed
  decompression, reassembly, and transmission over
  the network.
• Additionally, fragmentation on the storage device
  can cause additional rehydration overhead.

Download Your Free Trial of our Backup
Compression Tool
10
Learn MoreBackup Compression Tool Free
TrialDownload Compression WhitepaperFollow us
on Twitter _at_SQLDBApros