Naming is hard: ArcticDB is now FrostDB (+updates!)

We made a mistake in naming our open-source in-memory columnar database.

June 16, 2022
golang
apache arrow
apache parquet
parca
column database

As the old saying goes, naming is hard, and we just experienced this firsthand. In May we announced arcticDB, and recently MAN Group gave us a heads up that they hold a trademark for that name (thank you for the friendly and productive collaboration). We would like to apologize for this, at the time we thought it was simply another open-source project which often have overlapping names. Therefore we have changed the name, and (our) arcticDB is now: FrostDB.

ArcticDB to FrostDB logo rebrand

Project update

Since we’re already talking about FrostDB, we felt it was a great opportunity to share what has happened since we introduced it. Also a number we’re excited about: 7 non-Polar Signals employees have contributed to FrostDB since the announcement. We care a great deal about building a community around this project so this is exciting.

Persistence

One of the most talked-about topics when we originally introduced it was that FrostDB didn’t feature any sort of persistence, data was just eventually rotated out of the storage once a memory limit was reached and the data was gone - this is changing. Thanks to a very motivated member of the community the experimental feature of persisting data to object storage has landed in PR#63. While this is only the beginning and we have yet to work on compaction, page caching, and improved query planning this is a significant step.

Transaction ergonomics

There are various improvements in the ergonomics of transactions:

  • Transactions can now be properly aborted using context propagation (PR#65).
  • Writers can now wait for their transaction to become visible (PR#66).
  • Readers can now have better control over the scope of a transaction allowing them to be more explicit about their consistency requirements (as part of PR#92).

Correctness

So many correctness bugs fixed.

What’s next?

As already alluded to in the persistence section, there is still much work to be done on that front. Another area of active development is improving the performance of the query engine, a particularly interesting area we’re working on is a “shared-nothing” query execution model to make the best possible use of parallelized execution of queries (tracked in #88).

Once again, we apologize to MAN Group for our mistake. Long live FrostDB!

If you want to chat about FrostDB or Parca, join the Discord server.

Discuss:
Sign up for the latest Polar Signals news