Bucket Data Placement Policy – Spread VS Mirror

Last Updated: Aug 30, 2017 02:52PM PDT

General Data Placement Process Background
Every object that is written to NooBaa goes through a digestion process which includes the following steps:

  • Fragmentation
  • Deduplication
  • Compression
  • Encryption


The output of this process is a set of encrypted chunks.

Each chunk is stored in a manner deemed safe by the system. Chunks will therefore be stored in one of the following ways:

A minimum of three copies of the same chunk on storage nodes, where data redundancy is achieved by storing the same chunk in multiple locations.

A single copy of the chunk in a cloud pool, where data redundancy is achieved based on the cloud vendor's internal redundancy.

 

Spread Data Placement
When to use it?

  • When the system contains multiple pools.
  • For buckets that contain data not requiring full site failure redundancy.


Under the hood

  • The system will create three copies across the underlying pools for each of the chunks.
  • There is no guaranteed chunk distribution between the pools.
  • The client will receive an acknowledge once redundant copies have been created.

[Spread Policy Diagram]


 
Mirror Data Placement
When to use it?

  • When the system contains multiple pools.
  • For bucket, that retains data requiring full site failure redundancy.


Under the hood

  • The system will create three copies on each of the underlying pools for each of the chunks.
  • Data will be asynchronously replicated between pools.
  • It is guaranteed that each of the chunks will be stored redundantly on each of the underlying pools.



[Mirror Policy Diagram]
 

Did you find this article helpful?
195e69fe2c7a9be6b2b2aa6b87f970548@noobaa.desk-mail.com
https://cdn.desk.com/
false
desk
Loading
seconds ago
a minute ago
minutes ago
an hour ago
hours ago
a day ago
days ago
about
false
Invalid characters found
/customer/en/portal/articles/autocomplete