{
 "name": "Spark Accelerator Ledger",
 "edition": "2026-11-10",
 "read_date": "2026-11-10",
 "source": "https://sparkaccelerationcompare.com/datasets/",
 "license": "CC BY 4.0",
 "cite": "Spark Acceleration Compare, \"Spark Accelerator Ledger\" and \"Change Budget\", edition 10 Nov 2026, https://sparkaccelerationcompare.com/datasets/ (CC BY 4.0).",
 "description": "Every published benchmark found for the 12 options on this site, with who ran it, on what and at what cost, as published. Ordered by publication date, newest first.",
 "rows": [
  {
   "published_date": "2026-11-03",
   "option": "Apache DataFusion Comet",
   "title": "Comet 1.1.0 and 1.2.0 against Spark 4.1.3, TPC-DS SF1000",
   "workload": "TPC-DS derived queries, 103 queries, Parquet (zstd) in S3",
   "data_size": "1 TB (scale factor 1000)",
   "hardware": "Same EKS cluster on r6i instances as the Spark 4.2.0 run",
   "result_as_published": "Spark 4.1.3 1,336 s; Comet 1.1.0 613 s (2.18x); Comet 1.2.0 532 s (2.51x)",
   "cost_result_as_published": "None published",
   "ran_by": "Apache DataFusion Comet project",
   "ran_by_is_vendor": false,
   "run_by_maker": true,
   "method_published": "Yes, scripts in the Comet repository and the apache/datafusion-benchmarks repository",
   "source_url": "https://raw.githubusercontent.com/apache/datafusion-comet/main/docs/source/contributor-guide/benchmark-results/tpc-ds.md",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2026-11-03",
   "option": "Apache DataFusion Comet",
   "title": "Comet 1.1.0 and 1.2.0 against Spark 4.2.0, TPC-DS SF1000",
   "workload": "TPC-DS derived queries, 103 queries, Parquet (zstd) in S3, each query run twice, mean reported",
   "data_size": "1 TB (scale factor 1000)",
   "hardware": "EKS cluster on r6i instances with EBS; 32 executors x 16 cores, 64G executor memory (Comet adds 32G off-heap), driver 128G",
   "result_as_published": "Spark 4.2.0 994 s; Comet 1.1.0 601 s; Comet 1.2.0 488 s (published as 2.0x faster than Spark and 1.2x faster than Comet 1.1.0)",
   "cost_result_as_published": "None published",
   "ran_by": "Apache DataFusion Comet project",
   "ran_by_is_vendor": false,
   "run_by_maker": true,
   "method_published": "Yes, scripts in the Comet repository and the apache/datafusion-benchmarks repository",
   "source_url": "https://raw.githubusercontent.com/apache/datafusion-comet/main/docs/source/contributor-guide/benchmark-results/tpc-ds.md",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2026-09",
   "option": "Flarion",
   "title": "Flarion TPC-DS SF100, six engines on one machine",
   "workload": "TPC-DS, 103 queries, sum of the lowest per-query times, two runs",
   "data_size": "100 GB (SF100)",
   "hardware": "One i4i.8xlarge (32 vCPU, 256 GB), us-west-2a, Parquet read from S3; Databricks rows on the same instance type in us-east-1",
   "result_as_published": "Total query time: Databricks Standard 1,180 s; Spark 3.5.5 1,135 s; Comet 1.0.0 652 s; Gluten with Velox 1.7.0 557 s; Flarion (build 2026-09-12) 386 s; Databricks Photon 362 s",
   "cost_result_as_published": "None published",
   "ran_by": "Flarion",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "Partly: machine, builds and settings published; Flarion is not downloadable and no scripts were found",
   "source_url": "https://www.flarion.io/benchmarks",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2026-09",
   "option": "Flarion",
   "title": "Flarion TPC-DS SF1000, six engines on one machine",
   "workload": "TPC-DS, 103 queries, sum of per-query times, one run, 1,800 s cap per query",
   "data_size": "1 TB (SF1000)",
   "hardware": "One i4i.8xlarge (32 vCPU, 256 GB), us-west-2a, Parquet read from S3; Databricks rows on the same instance type in us-east-1",
   "result_as_published": "Total query time: Databricks Standard 103.1 min; Spark 3.5.5 135.6 min; Comet 84.4 min; Gluten with Velox 46.9 min; Flarion 38.3 min; Databricks Photon 30.4 min",
   "cost_result_as_published": "None published",
   "ran_by": "Flarion",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "Partly: machine, builds and settings published; Flarion is not downloadable and no scripts were found",
   "source_url": "https://www.flarion.io/benchmarks",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2026-08-25",
   "option": "NVIDIA cuDF plugin for Apache Spark",
   "title": "GPU Spark on Amazon EMR on EKS with G7 instances, TPC-DS 3 TB",
   "workload": "TPC-DS, all 103 queries, power run, Parquet on S3, median of 5 iterations",
   "data_size": "3 TB",
   "hardware": "8 x g7.4xlarge (1 x NVIDIA RTX PRO 4500 Blackwell, 16 vCPU, 64 GB) against 8 x m9gd.4xlarge, m8id.4xlarge and m8a.4xlarge; EMR on EKS 7.12.0, Spark 3.5.6, cuDF plugin 26.04.2, us-east-1",
   "result_as_published": "GPU 281 s (4.7 min); CPU 1,010 to 1,043 s (16.8 to 17.4 min); published speedup 3.7x",
   "cost_result_as_published": "$2.06 per run on GPU against $2.93 to $3.18 per run on CPU; 'up to 31% less per run than Graviton'",
   "ran_by": "AWS with NVIDIA",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "Yes, cluster, Spark config, EMR release and pricing method in the post (no scripts linked)",
   "source_url": "https://aws.amazon.com/blogs/big-data/gpu-accelerated-apache-spark-with-amazon-emr-and-nvidia-rtx-pro-4500-on-amazon-ec2-g7-instances-runs-up-to-3-7x-faster/",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2026-06-11",
   "option": "DualBird",
   "title": "DualBird sort-based Iceberg compaction, 100 GB",
   "workload": "Sort-based Apache Iceberg compaction on a synthetic denormalized event table (about 370 million rows, 64 columns, ZSTD level 3)",
   "data_size": "100 GB",
   "hardware": "AWS F2 instances for DualBird; baselines are vanilla Spark and an unnamed C++ accelerated Spark; instance types and counts not stated in the public post",
   "result_as_published": "About 12x to 20x faster task execution against vanilla Spark and an unnamed C++ accelerated Spark (DualBird's wording; no absolute times published)",
   "cost_result_as_published": "55% to 85% lower EC2 cost; about $2 per TB for DualBird against about $6 per TB rising past $13 per TB for CPU-based clusters",
   "ran_by": "DualBird",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "Behind a sign-up form (white paper dated 2026-07-01); no public repository found",
   "source_url": "https://dualbird.io/resources/blog/iceberg-compaction-benchmark-faster-spark-lower-cost",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2026-06-11",
   "option": "Speedata",
   "title": "Speedata APU on TPC-H SF100, Microsoft CoddSpeed paper",
   "workload": "TPC-H queries Q07, Q10, Q12, Q13 and Q16 through Microsoft's coprocessor abstraction layer, standalone APU against single-core SQL Server",
   "data_size": "TPC-H scale factor 100, Parquet, warm in CPU memory",
   "hardware": "Speedata C-200 APU; baseline SQL Server on Intel Xeon Platinum 8473C at 2.80GHz, single core",
   "result_as_published": "APU speedup as relayed by Speedata from Table 2 of Microsoft's paper: Q07 36.7x, Q10 40.7x, Q12 49.4x, Q13 259.2x, Q16 61.5x, against single-core SQL Server",
   "cost_result_as_published": "None published",
   "ran_by": "Microsoft (paper); numbers read from Speedata's blog",
   "ran_by_is_vendor": false,
   "run_by_maker": false,
   "method_published": "Yes, in the paper; the paper itself was not fetched",
   "source_url": "https://www.speedata.io/post/259-faster-what-microsoft-s-award-winning-sigmod-paper-says-about-the-speedata-apu",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2025-08-21",
   "option": "DataPelago",
   "title": "Nucleus v3.0 against NVIDIA cuDF 25.04.00, operator tests",
   "workload": "Project, filter, aggregate, hash aggregate, hash join and K-largest selection operators on a single GPU",
   "data_size": "Not stated",
   "hardware": "AWS p5.48xlarge, AMD EPYC 7R13 (192 vCPU), 2 TB DRAM, 8x NVIDIA H100 (one GPU used)",
   "result_as_published": "Up to 10.5x on project, 10.1x on filter, 4.3x on aggregate, 4.5x on hash aggregate and 8.2x on K-largest selection, against cuDF",
   "cost_result_as_published": "None published",
   "ran_by": "DataPelago",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://www.datapelago.ai/resources/DataPelago-Nucleus-Vs-Nvidia-cuDF",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2025-05-30",
   "option": "Google Cloud Lightning Engine",
   "title": "Lightning Engine against open source Spark, TPC-H-like 10 TB",
   "workload": "TPC-H derived queries (not compliant with the TPC-H specification)",
   "data_size": "10 TB",
   "hardware": "Open source Spark on 'similar infrastructure'; cluster not published",
   "result_as_published": "3.6x acceleration over open source Spark",
   "cost_result_as_published": "None published",
   "ran_by": "Google Cloud",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://cloud.google.com/blog/products/data-analytics/introducing-lightning-engine-for-apache-spark",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2022",
   "option": "Databricks Photon",
   "title": "Photon customer speedups, SIGMOD 2022 paper",
   "workload": "Customer production queries",
   "data_size": "Not stated",
   "hardware": "Not stated",
   "result_as_published": "Average speedup of 3x over the previous Databricks Runtime; maximum speedup over 10x",
   "cost_result_as_published": "None published",
   "ran_by": "Databricks authors",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://people.eecs.berkeley.edu/~matei/papers/2022/sigmod_photon.pdf",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "2021-11-02",
   "option": "Databricks Photon",
   "title": "Databricks audited 100 TB TPC-DS result on Photon",
   "workload": "Official TPC-DS run on Databricks SQL",
   "data_size": "100 TB",
   "hardware": "Databricks SQL on AWS with Delta Lake on S3 (per the Photon paper)",
   "result_as_published": "32,941,245 QphDS at 100 TB, against 14,861,137 QphDS for the previous record holder (Alibaba)",
   "cost_result_as_published": "None published",
   "ran_by": "Databricks, audited by the TPC",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No (a TPC audit report exists under TPC rules; not a repeatable open script)",
   "source_url": "https://www.databricks.com/blog/2021/11/02/databricks-sets-official-data-warehousing-performance-record.html",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "Amazon EMR runtime for Apache Spark",
   "title": "EMR runtime claim on the AWS features page",
   "workload": "Not stated on the page (Iceberg write also cited)",
   "data_size": "Not stated",
   "hardware": "Not stated",
   "result_as_published": "AWS's wording: up to 4.5x faster than open-source equivalents; 2.7x faster Iceberg write performance",
   "cost_result_as_published": "None published",
   "ran_by": "AWS",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "Partly: AWS publishes the aws-samples/emr-spark-benchmark harness, which states no results",
   "source_url": "https://aws.amazon.com/emr/features/spark/",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "Apache Auron",
   "title": "Auron TPC-DS 1 TB against Apache Spark 3.5.6",
   "workload": "TPC-DS, Parquet with zstd, run with a spark-sql-perf derived runner",
   "data_size": "1 TB (scale factor 1000)",
   "hardware": "YARN cluster, executor cores 8; node type and count not published",
   "result_as_published": "Published as 'running ~2x faster and save ~50% cluster resources' against Apache Spark 3.5; per-query results are a chart image only",
   "cost_result_as_published": "None published (only the relative claim of about 50% fewer cluster resources)",
   "ran_by": "Apache Auron project",
   "ran_by_is_vendor": false,
   "run_by_maker": true,
   "method_published": "Yes, tools and full Spark config in the auron-project/tpcds-tools repository",
   "source_url": "https://auron.apache.org/documents/benchmarks.html",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "Apache Gluten with Velox",
   "title": "Gluten with Velox, TPC-DS 3 TB (project homepage)",
   "workload": "TPC-DS",
   "data_size": "3 TB",
   "hardware": "Intel Xeon Platinum 8592+, Spark 3.3.1 (cluster size not stated)",
   "result_as_published": "Gluten project's wording: overall 3.02x speedup with Gluten and Velox; up to 13.75x on individual queries",
   "cost_result_as_published": "None published",
   "ran_by": "Apache Gluten project",
   "ran_by_is_vendor": false,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://gluten.apache.org/",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "Apache Gluten with Velox",
   "title": "Gluten with Velox, TPC-H 3 TB (project homepage)",
   "workload": "TPC-H",
   "data_size": "3 TB",
   "hardware": "Intel Xeon Platinum 8592+, Spark 3.3.1 (cluster size not stated)",
   "result_as_published": "Gluten project's wording: overall 3.34x speedup with Gluten and Velox; up to 23.45x on individual queries",
   "cost_result_as_published": "None published",
   "ran_by": "Apache Gluten project",
   "ran_by_is_vendor": false,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://gluten.apache.org/",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "DualBird",
   "title": "DualBird 110 GB enrichment pipeline (join denormalization)",
   "workload": "Enrichment pipeline with join denormalization",
   "data_size": "110 GB",
   "hardware": "Not stated",
   "result_as_published": "DualBird's wording: '6x to 18x higher performance ceiling' on a 110 GB enrichment pipeline; baseline, hardware and absolute times not stated",
   "cost_result_as_published": "50% to 85% lower EC2 cost",
   "ran_by": "DualBird",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://dualbird.io/",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "DualBird",
   "title": "DualBird single task performance tile",
   "workload": "Single Spark task at the same cluster size (home page tile; appears to be the compaction workload)",
   "data_size": "Not stated",
   "hardware": "Not stated",
   "result_as_published": "DualBird's wording: 'X15-X20 speedup for the same cluster size'; baseline not stated on the tile (the compaction post names vanilla Spark and an unnamed C++ accelerated Spark)",
   "cost_result_as_published": "Reducing cost by over 50%",
   "ran_by": "DualBird",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://dualbird.io/",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "Google Cloud Lightning Engine",
   "title": "Lightning Engine claim on Google's pricing page",
   "workload": "Not stated",
   "data_size": "Not stated",
   "hardware": "Not stated",
   "result_as_published": "Google Cloud's wording: up to 4.9x faster than open source Spark; up to 2.5x the price-performance of a high-speed Spark alternative that the page does not name",
   "cost_result_as_published": "None published (the claim is relative price-performance)",
   "ran_by": "Google Cloud",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://docs.cloud.google.com/managed-spark/pricing",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "Google Cloud Lightning Engine",
   "title": "Native Query Execution on Axion ARM, Google's documentation",
   "workload": "Analytical workloads (not specified)",
   "data_size": "Not stated",
   "hardware": "Google Axion ARM (C4A) machines",
   "result_as_published": "Google Cloud's wording: 2.5x to 4x speedup over open source Spark",
   "cost_result_as_published": "None published",
   "ran_by": "Google Cloud",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://docs.cloud.google.com/managed-spark/docs/guides/lightning-engine",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "NVIDIA cuDF plugin for Apache Spark",
   "title": "NVIDIA comparison with open-source Spark on EC2, NDS 3 TB",
   "workload": "NDS (TPC-DS derived), 103 queries",
   "data_size": "3 TB",
   "hardware": "GPU result from the AWS EMR run; CPU runs on EC2 m7gd.8xlarge and m9gd.4xlarge with open-source Spark 3.5.6",
   "result_as_published": "NVIDIA's wording: GPU run 8.4x faster than m7gd.8xlarge and 11.0x faster than m9gd.4xlarge, as published",
   "cost_result_as_published": "77% and 72% lower reported cost per run; GPU cost includes EC2 and EMR on EKS, CPU costs include EC2 only",
   "ran_by": "NVIDIA",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No (the NDS automation is public; the comparison is described on the overview page only)",
   "source_url": "https://docs.nvidia.com/spark-rapids/user-guide/latest/overview.html",
   "read_date": "2026-11-10"
  },
  {
   "published_date": "Not stated",
   "option": "Speedata",
   "title": "Speedata production Spark workload tile",
   "workload": "Enterprise AI data preprocessing on Spark; customer not named",
   "data_size": "Not stated",
   "hardware": "Not stated",
   "result_as_published": "Speedata's wording: 62.7x acceleration on production Spark workloads; 37 servers consolidated to 3; 90 to 8 hours of processing time",
   "cost_result_as_published": "None published (the tile claims up to 90% lower TCO)",
   "ran_by": "Speedata",
   "ran_by_is_vendor": true,
   "run_by_maker": true,
   "method_published": "No",
   "source_url": "https://www.speedata.io/",
   "read_date": "2026-11-10"
  }
 ]
}