back to article No objections to object stores: Everyone's going smaller and faster

A couple of weeks ago I published an article about high performance object storage. Reactions have been quite diverse. Some think that object stores can only be huge and slow and then others who think quite the opposite. In fact, they can also be fast and small. In the last year I’ve had a lot of interesting conversations with …

  1. CaitlinBestler

    Object Storage is not NECESSARY for smaller clusters.

    Small clusters can use centralized metadata to manage storage very quickly.

    Centralized metadata can support both POSIX and object semantics.

    But when you scale up the metadata must be distributed, making efficient support of POSIX untenable.

    That said, there is no inherent reason for "file servers" to be faster than "object servers".

    The metadata operations required to put a new version of a 10 MB document are actually

    less demanding with object storage than with file storage. The only advantage file APIs have

    is their familiarity to developers and the maturity of the code base implementing POSIX APIs.

    Neither the existing code or the APIs will scale to the degree that object APIs allow, however.

    1. Anonymous Coward
      Anonymous Coward

      Re: Object Storage is not NECESSARY for smaller clusters.

      Managing billion of small files is not actually efficient with traditional Posix Filesystem. Filesystem built on top of object storage allow you to store and manage billions of small files. Distributed object storage cluster enable you to distribute file storage metadata across object storage nodes. As for instance, you can have one distributed object storage cluster with SSD for storing metadata and one bigger distributed object storage cluster with traditional drives for storing data. Small files should not be a concern if your underlying object storage is fast for small files, but not all object storage cluster can fit this purpose.

  2. Mark Hahn

    but why?

    I was hoping you might discuss object storage for smaller *objects* - that would be interesting. An article about timid, half-hearted implementations of only a hundred disks or less, who cares?

    It's easy to see how some workloads fit well for object storage. It's much harder to see how it'll challenge the prevalence of normal filesystems, where files are often tiny. After all, object storage is just a filesystem that can't efficiently handle large files, and refuses to manage your metadata/namespace for you!

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon