1. 30
  1. 3

    Really excited for this! Just submitted a talk around a new deterministic LSM-tree storage engine for TigerBeetle [1] written in Zig and using io_uring with some new ideas for pipelined/incremental/streaming compaction and how the storage engine integrates with the distributed consensus protocol. @ifreund and myself did a live stream [2] giving a high-level overview of some of these ideas today. Would love to talk more at hytradboi. Also awesome to see Frank McSherry on the list!

    [1] https://github.com/coilhq/tigerbeetle

    [2] https://www.twitch.tv/videos/1229342192

    1. 2

      That sounds really interesting. I’ll check out the video.

    2. 2

      This looks great! I’m all for database-centric architectures. am a big fan of Materialize (right in the process of introducing it in a project I’m working in) and followed Eve closely when it was still alive. (I created the little bouncing ball example that became a benchmark of sorts towards the end.) Will definitely try to participate in this!

      1. 2

        What are Materialize and Eve? I’m puzzled by “little bouncing ball” and relationship with databases — that sounds pretty interesting!

        1. 4

          Materialize is a streaming database based on differential dataflow, from Frank McSherry and team. McSherry is well-known for his “Cost that outperforms single-threaded” (COST) paper [PDF].

          1. 3

            This is one of my favorite papers from Frank McSherry. The other being “A COOL AND PRACTICAL ALTERNATIVE TO TRADITIONAL HASH TABLES” which is a great intro to Cuckoo hashing, and a great analysis: https://www.ru.is/faculty/ulfar/CuckooHash.pdf