Hammerspace differentiates file copies from file instantiations – Blocks and Files

Hammerspace, the global file data environment company, claims that copies are different instances of the same file, and therefore forks, while instantiations are not and therefore not forks.

What does this mean, and why is it important?

The update was provided to attendees during an IT press tour in Palo Alto. Let’s start with a file: Testfile.dat. It consists of data and metadata and suppose it takes 100 blocks on a storage disk for data and one block for metadata – 101 blocks in total. Now let’s make a copy.

David Flynn.

The copy consists of file data, 100 blocks, and metadata, one block, totaling up to 101 blocks. Hammerspace Founder and CEO David Flynn says it’s a separate file – in fact, a fork of the original file. There may be changes that are not propagated to the original file. Its content may therefore differ from the original file. And vice versa if the original file changes.

Now imagine Hammerspace’s GDE (Global Data Environment) instantiating – as it calls it – the original file. This can happen when a user in a data center in San Francisco accesses a file that is present in their local GDE configuration but whose data is stored in New York. This is his base location. In the Hammerspace scheme, file metadata is shared between sites on a peer-to-peer basis so that all users everywhere in the GDE see the same files in their file:folder structure.

When the San Francisco user decides to access Testfile.dat, the constituent data blocks are passed from New York to San Francisco and instantiated there. This file instantiation has the exact same metadata as the master New York copy. All changes are propagated to New York so that there is one version of the truth of the file contents and no file forking takes place.

Thus, Hammerspace GDE instantiations use exactly the same metadata as the master copy. In the GDE file system, metadata is a control plane.

Participants also learned that GDE is structured to operate synchronously in local domains (data centers) and asynchronously in geo-distributed data centers – where it is ultimately consistent.

We were told that data is digital but desperately needs a digital transformation – to make it less manual in its management. Hammerspace presentations have a way to reset its preconceptions