A “backup” is a static copy of a serverless index.

A “collection” is a static copy of a pod-based index.

A backup or collection is a static copy of your index that only consumes storage. It is a non-queryable representation of a set of records. You can create a backup or collection of an index, and you can create a new index from a backup or collection. This allows you to restore the index with the same or different configurations.

Once a backup or collection is created, it cannot be moved to a different project.

Serverless index backups are in public preview and available only on Standard and Enterprise plans.

Use cases

Creating a backup or collection is useful when performing tasks like the following:

  • Protecting an index from manual or system failures.
  • Temporarily shutting down an index.
  • Copying the data from one index into a different index.
  • Making a backup of your index.
  • Experimenting with different index configurations.

Performance

Backup and restore times depend upon the size of the index and number of namespaces:

  • For less than 1M vectors in a namespace, backups and restores take approximately 10 minutes.
  • For 100,000,000 vectors, backups and restores can take up to 5 hours.

Collections operations perform differently, depending on the pod type of the index:

  • Creating a p1 or s1 index from a collection takes approximately 10 minutes.
  • Creating a p2 index from a collection can take several hours when the number of vectors is on the order of 1,000,000.

Limitations

Backup limitations are as follows:

  • Backups can be created only for serverless indexes with up to 100,000,000 vectors or 2,000 namespaces.
  • There is a quota of 50 serverless backups per project.
  • Backups are stored in the same region as the source index.
  • You can only restore an index to the same region as the source index.
  • Backups only include vectors that were in the index at least 15 minutes prior to the backup time. This means that if a vector was inserted into an index and a backup was immediately taken after, the recently inserted vector may not be backed up. More specifically, if a backup is created only a few minutes after the source index was created, the backup may have 0 vectors.
  • You can only perform operations on backups in the current Pinecone project.

Collection limitations are as follows:

  • You can only perform operations on collections in the current Pinecone project.

Pricing

See Pricing for up-to-date pricing information.