From b83bd999997a3c2f32e7868ada5b7055200fdf05 Mon Sep 17 00:00:00 2001 From: Alfonso Subiotto Marques Date: Wed, 26 Nov 2025 13:24:53 +0100 Subject: [PATCH] writing: add parquet to vortex blog post Signed-off-by: Alfonso Subiotto Marques --- content/writing/interface-parquet-vortex.md | 13 +++++++++++++ 1 file changed, 13 insertions(+) create mode 100644 content/writing/interface-parquet-vortex.md diff --git a/content/writing/interface-parquet-vortex.md b/content/writing/interface-parquet-vortex.md new file mode 100644 index 0000000..c31a3a6 --- /dev/null +++ b/content/writing/interface-parquet-vortex.md @@ -0,0 +1,13 @@ +--- +title: "Questioning an Interface: From Parquet to Vortex" +date: 2025-11-25T00:00:00Z +draft: false +tags: ["databases", "datafusion", "vortex", "apache parquet"] +originalPost: + site: "Polar Signals" + url: "https://www.polarsignals.com/blog/posts/2025/11/25/interface-parquet-vortex" +--- + +An exploration of how Polar Signals migrated their profiling database from Apache Parquet to Vortex, achieving a 70% average performance improvement across all queries. The shift demonstrates how interface design choices can impose unexpected limitations on system performance. + +Parquet's "retrieval" differs fundamentally from "querying"—converting Parquet files to Arrow format for actual computation consumed significant CPU resources. Vortex, with its zero-copy conversion to Arrow and SIMD-friendly encodings, proved to be a better fit for our use case.