Add Hyrise

hyrise/Dockerfile

@@ -0,0 +1,97 @@
+FROM ubuntu:25.04 AS build
+
+ENV DEBIAN_FRONTEND=noninteractive
+ENV HYRISE_HEADLESS_SETUP=1
+
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends \
+        autoconf \
+        bash-completion \
+        bc \
+        ca-certificates \
+        clang-19 \
+        clang-20 \
+        cmake \
+        curl \
+        dos2unix \
+        g++-13 \
+        g++-15 \
+        gcc-13 \
+        gcc-15 \
+        git \
+        libboost-all-dev \
+        libhwloc-dev \
+        libncurses-dev \
+        libnuma-dev \
+        libnuma1 \
+        libpq-dev \
+        libreadline-dev \
+        libsqlite3-dev \
+        libtbb-dev \
+        lld-20 \
+        llvm-20 \
+        lsb-release \
+        make \
+        ninja-build \
+        parallel \
+        postgresql-server-dev-all \
+        python3 \
+        python3-pip \
+        software-properties-common \
+        sudo \
+    && update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-15 90 --slave /usr/bin/g++ g++ /usr/bin/g++-15 \
+    && apt-get clean \
+    && rm -rf /var/lib/apt/lists/*
+
+# Pin a specific Hyrise revision for reproducibility (defaults to master).
+ARG HYRISE_REF=master
+# LTO is off by default: GCC's `-flto=auto` link of libhyrise_impl.so spawns
+# one lto1 per core, and on the 32 GB c6a.4xlarge runner 16 of them exhaust
+# RAM+swap and get killed (earlyoom SIGTERM → "Terminated signal terminated
+# program lto1"). Set NO_LTO=FALSE on machines with ≥4 GB/core to re-enable.
+ARG NO_LTO=TRUE
+
+WORKDIR /opt
+RUN git clone https://github.com/hyrise/hyrise.git \
+    && cd hyrise \
+    && git checkout "${HYRISE_REF}" \
+    && git submodule update --jobs 8 --init --recursive --depth 1
+
+# hyriseConsole is shipped alongside the server because its `export` command
+# reads tables via GetTable without the MVCC Validate wrapper — SQL `COPY TO`
+# validates, and BinaryWriter materializes the resulting ReferenceSegments as
+# unencoded value segments, which would make the snapshot raw-sized and make
+# it re-import unencoded. See ./load.
+WORKDIR /opt/hyrise/cmake-build-release
+RUN cmake -GNinja -DCMAKE_BUILD_TYPE=Release -DNO_LTO=${NO_LTO} .. \
+    && ninja hyriseServer hyriseConsole
+
+FROM ubuntu:25.04
+ENV DEBIAN_FRONTEND=noninteractive
+# libboost-container/date-time: libhyrise_impl.so links them dynamically
+# (they are not dependencies of boost-system/thread, so they must be listed).
+# libreadline8t64 + libncurses6: hyriseConsole.
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends \
+        libboost-container1.83.0 \
+        libboost-date-time1.83.0 \
+        libboost-system1.83.0 \
+        libboost-thread1.83.0 \
+        libhwloc15 \
+        libncurses6 \
+        libnuma1 \
+        libreadline8t64 \
+        libsqlite3-0 \
+        libtbb12 \
+    && apt-get clean \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY --from=build /opt/hyrise/cmake-build-release/hyriseServer /usr/local/bin/hyriseServer
+COPY --from=build /opt/hyrise/cmake-build-release/hyriseConsole /usr/local/bin/hyriseConsole
+COPY --from=build /opt/hyrise/cmake-build-release/lib/libhyrise_impl.so /usr/local/lib/
+COPY --from=build /opt/hyrise/cmake-build-release/third_party/jemalloc/lib/libjemalloc.so.2 /usr/local/lib/
+RUN ldconfig
+
+EXPOSE 5432
+ENTRYPOINT ["/usr/local/bin/hyriseServer"]
+CMD ["5432"]

hyrise/benchmark.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+# Thin shim — actual flow is in lib/benchmark-common.sh.
+#
+# Hyrise has no on-disk persistence: every restart resurfaces with an empty
+# catalog and the dataset must be re-loaded into RAM. BENCH_DURABLE=no makes
+# the driver re-run ./load on every cold cycle (and roll that wall-clock into
+# the cold-try timing) so each "cold" number genuinely measures
+# load+query against a fresh in-memory dataset. The first ./load converts the
+# CSV into an encoded binary snapshot; the per-cycle reloads import that
+# snapshot, which is the closest Hyrise equivalent of a durable system
+# re-reading its own on-disk format after a restart.
+export BENCH_DOWNLOAD_SCRIPT="download-hits-csv"
+export BENCH_DURABLE=no
+exec ../lib/benchmark-common.sh

hyrise/check

@@ -0,0 +1,4 @@
+#!/bin/bash
+set -e
+
+psql -h 127.0.0.1 -p 5432 -U postgres -c 'SELECT 1' >/dev/null

hyrise/data-size

@@ -0,0 +1,10 @@
+#!/bin/bash
+set -eu
+
+# Hyrise has no on-disk persistence; report the total in-memory segment
+# size from meta_segments. Strip any whitespace so the driver's strict
+# digits-only regex check passes.
+psql -h 127.0.0.1 -p 5432 -U postgres -t -A \
+    -c "SELECT SUM(estimated_size_in_bytes) FROM meta_segments WHERE table_name = 'hits';" \
+    | tr -d '[:space:]'
+echo

hyrise/hits.csv.json

@@ -0,0 +1,116 @@
+{
+    "config": {
+        "rfc_mode": true,
+        "separator": ",",
+        "quote": "\"",
+        "escape": "\"",
+        "delimiter": "\n"
+    },
+    "columns": [
+        {"name": "WatchID",               "type": "long",   "nullable": false},
+        {"name": "JavaEnable",            "type": "int",    "nullable": false},
+        {"name": "Title",                 "type": "string", "nullable": false},
+        {"name": "GoodEvent",             "type": "int",    "nullable": false},
+        {"name": "EventTime",             "type": "string", "nullable": false},
+        {"name": "EventDate",             "type": "string", "nullable": false},
+        {"name": "CounterID",             "type": "int",    "nullable": false},
+        {"name": "ClientIP",              "type": "int",    "nullable": false},
+        {"name": "RegionID",              "type": "int",    "nullable": false},
+        {"name": "UserID",                "type": "long",   "nullable": false},
+        {"name": "CounterClass",          "type": "int",    "nullable": false},
+        {"name": "OS",                    "type": "int",    "nullable": false},
+        {"name": "UserAgent",             "type": "int",    "nullable": false},
+        {"name": "URL",                   "type": "string", "nullable": false},
+        {"name": "Referer",               "type": "string", "nullable": false},
+        {"name": "IsRefresh",             "type": "int",    "nullable": false},
+        {"name": "RefererCategoryID",     "type": "int",    "nullable": false},
+        {"name": "RefererRegionID",       "type": "int",    "nullable": false},
+        {"name": "URLCategoryID",         "type": "int",    "nullable": false},
+        {"name": "URLRegionID",           "type": "int",    "nullable": false},
+        {"name": "ResolutionWidth",       "type": "int",    "nullable": false},
+        {"name": "ResolutionHeight",      "type": "int",    "nullable": false},
+        {"name": "ResolutionDepth",       "type": "int",    "nullable": false},
+        {"name": "FlashMajor",            "type": "int",    "nullable": false},
+        {"name": "FlashMinor",            "type": "int",    "nullable": false},
+        {"name": "FlashMinor2",           "type": "string", "nullable": false},
+        {"name": "NetMajor",              "type": "int",    "nullable": false},
+        {"name": "NetMinor",              "type": "int",    "nullable": false},
+        {"name": "UserAgentMajor",        "type": "int",    "nullable": false},
+        {"name": "UserAgentMinor",        "type": "string", "nullable": false},
+        {"name": "CookieEnable",          "type": "int",    "nullable": false},
+        {"name": "JavascriptEnable",      "type": "int",    "nullable": false},
+        {"name": "IsMobile",              "type": "int",    "nullable": false},
+        {"name": "MobilePhone",           "type": "int",    "nullable": false},
+        {"name": "MobilePhoneModel",      "type": "string", "nullable": false},
+        {"name": "Params",                "type": "string", "nullable": false},
+        {"name": "IPNetworkID",           "type": "int",    "nullable": false},
+        {"name": "TraficSourceID",        "type": "int",    "nullable": false},
+        {"name": "SearchEngineID",        "type": "int",    "nullable": false},
+        {"name": "SearchPhrase",          "type": "string", "nullable": false},
+        {"name": "AdvEngineID",           "type": "int",    "nullable": false},
+        {"name": "IsArtifical",           "type": "int",    "nullable": false},
+        {"name": "WindowClientWidth",     "type": "int",    "nullable": false},
+        {"name": "WindowClientHeight",    "type": "int",    "nullable": false},
+        {"name": "ClientTimeZone",        "type": "int",    "nullable": false},
+        {"name": "ClientEventTime",       "type": "string", "nullable": false},
+        {"name": "SilverlightVersion1",   "type": "int",    "nullable": false},
+        {"name": "SilverlightVersion2",   "type": "int",    "nullable": false},
+        {"name": "SilverlightVersion3",   "type": "int",    "nullable": false},
+        {"name": "SilverlightVersion4",   "type": "int",    "nullable": false},
+        {"name": "PageCharset",           "type": "string", "nullable": false},
+        {"name": "CodeVersion",           "type": "int",    "nullable": false},
+        {"name": "IsLink",                "type": "int",    "nullable": false},
+        {"name": "IsDownload",            "type": "int",    "nullable": false},
+        {"name": "IsNotBounce",           "type": "int",    "nullable": false},
+        {"name": "FUniqID",               "type": "long",   "nullable": false},
+        {"name": "OriginalURL",           "type": "string", "nullable": false},
+        {"name": "HID",                   "type": "int",    "nullable": false},
+        {"name": "IsOldCounter",          "type": "int",    "nullable": false},
+        {"name": "IsEvent",               "type": "int",    "nullable": false},
+        {"name": "IsParameter",           "type": "int",    "nullable": false},
+        {"name": "DontCountHits",         "type": "int",    "nullable": false},
+        {"name": "WithHash",              "type": "int",    "nullable": false},
+        {"name": "HitColor",              "type": "string", "nullable": false},
+        {"name": "LocalEventTime",        "type": "string", "nullable": false},
+        {"name": "Age",                   "type": "int",    "nullable": false},
+        {"name": "Sex",                   "type": "int",    "nullable": false},
+        {"name": "Income",                "type": "int",    "nullable": false},
+        {"name": "Interests",             "type": "int",    "nullable": false},
+        {"name": "Robotness",             "type": "int",    "nullable": false},
+        {"name": "RemoteIP",              "type": "int",    "nullable": false},
+        {"name": "WindowName",            "type": "int",    "nullable": false},
+        {"name": "OpenerName",            "type": "int",    "nullable": false},
+        {"name": "HistoryLength",         "type": "int",    "nullable": false},
+        {"name": "BrowserLanguage",       "type": "string", "nullable": false},
+        {"name": "BrowserCountry",        "type": "string", "nullable": false},
+        {"name": "SocialNetwork",         "type": "string", "nullable": false},
+        {"name": "SocialAction",          "type": "string", "nullable": false},
+        {"name": "HTTPError",             "type": "int",    "nullable": false},
+        {"name": "SendTiming",            "type": "int",    "nullable": false},
+        {"name": "DNSTiming",             "type": "int",    "nullable": false},
+        {"name": "ConnectTiming",         "type": "int",    "nullable": false},
+        {"name": "ResponseStartTiming",   "type": "int",    "nullable": false},
+        {"name": "ResponseEndTiming",     "type": "int",    "nullable": false},
+        {"name": "FetchTiming",           "type": "int",    "nullable": false},
+        {"name": "SocialSourceNetworkID", "type": "int",    "nullable": false},
+        {"name": "SocialSourcePage",      "type": "string", "nullable": false},
+        {"name": "ParamPrice",            "type": "long",   "nullable": false},
+        {"name": "ParamOrderID",          "type": "string", "nullable": false},
+        {"name": "ParamCurrency",         "type": "string", "nullable": false},
+        {"name": "ParamCurrencyID",       "type": "int",    "nullable": false},
+        {"name": "OpenstatServiceName",   "type": "string", "nullable": false},
+        {"name": "OpenstatCampaignID",    "type": "string", "nullable": false},
+        {"name": "OpenstatAdID",          "type": "string", "nullable": false},
+        {"name": "OpenstatSourceID",      "type": "string", "nullable": false},
+        {"name": "UTMSource",             "type": "string", "nullable": false},
+        {"name": "UTMMedium",             "type": "string", "nullable": false},
+        {"name": "UTMCampaign",           "type": "string", "nullable": false},
+        {"name": "UTMContent",            "type": "string", "nullable": false},
+        {"name": "UTMTerm",               "type": "string", "nullable": false},
+        {"name": "FromTag",               "type": "string", "nullable": false},
+        {"name": "HasGCLID",              "type": "int",    "nullable": false},
+        {"name": "RefererHash",           "type": "long",   "nullable": false},
+        {"name": "URLHash",               "type": "long",   "nullable": false},
+        {"name": "CLID",                  "type": "int",    "nullable": false}
+    ]
+}

hyrise/install

@@ -0,0 +1,17 @@
+#!/bin/bash
+set -eu
+
+sudo apt-get update -y
+sudo apt-get install -y docker.io postgresql-client gzip
+
+# Hyrise has no upstream binary distribution and the build requires a
+# recent toolchain (gcc-15 / clang-20). Build inside Docker on top of
+# Ubuntu 25.04, then ship just the binary and its runtime libs in a slim
+# image.
+sudo docker build -t clickbench-hyrise .
+
+# The bind-mounted /data dir must exist before docker run (created here
+# instead of in ./load so ./start can be called before the first ./load
+# without erroring on a missing mount source).
+mkdir -p data
+chmod 777 data

hyrise/load

@@ -0,0 +1,57 @@
+#!/bin/bash
+set -eu
+
+# ./load runs on every cold cycle (BENCH_DURABLE=no). The first call turns
+# hits.csv into an encoded binary snapshot (data/hits.bin); every call then
+# imports that snapshot into the freshly started server.
+#
+# Why not one big `COPY hits FROM '/data/hits.csv'`: Hyrise's CSV import
+# reads the entire file into a single std::string and materializes every
+# column as unencoded ValueSegments before dictionary-encoding them —
+# roughly 200 GB peak for the 75 GB hits.csv. `COPY ... FROM` into an
+# existing table appends chunks, so importing 5M-row pieces sequentially
+# caps the transient overhead at one piece on top of the growing encoded
+# table. The pieces are imported in file order, which keeps the original
+# EventTime ordering and thus meaningful chunk-pruning statistics.
+#
+# Why the snapshot is written by hyriseConsole and not by SQL `COPY TO`:
+# a server-side SQL statement wraps the stored table in an MVCC Validate
+# operator, and BinaryWriter materializes the resulting ReferenceSegments
+# as unencoded value segments — the snapshot would be raw-sized and would
+# re-import unencoded (binary imports skip the encoding pass). The
+# console's `export` command reads the table via GetTable without
+# validation, so the dictionary-encoded segments are written — and later
+# re-imported, chunk by chunk — as-is. The console is a separate process
+# with its own catalog, so it performs the piecewise CSV import itself;
+# the server never sees the pieces, only the snapshot. Importing the
+# snapshot regenerates full-table statistics on add_table.
+if [ ! -f data/hits.bin ]; then
+    # hits.csv was delivered into cwd by download-hits-csv. Guarded so a
+    # retry after a partial first load resumes from the existing pieces.
+    if [ -f hits.csv ]; then
+        split -l 5000000 --numeric-suffixes=1 --additional-suffix=.csv \
+            hits.csv data/hits_part_
+        rm hits.csv
+    fi
+
+    # Each piece needs its own CSV meta file: Hyrise resolves it as
+    # "<filename>.json" next to the data file.
+    for piece in data/hits_part_*.csv; do
+        cp hits.csv.json "${piece}.json"
+        echo "COPY hits FROM '/data/$(basename "$piece")' WITH (FORMAT CSV);"
+    done > data/make_snapshot.script
+    echo "export hits /data/hits.bin" >> data/make_snapshot.script
+
+    # The console exits 0 even when the script aborts on an error, so the
+    # snapshot file itself is the success signal.
+    sudo docker exec hyrise hyriseConsole /data/make_snapshot.script
+    if [ ! -f data/hits.bin ]; then
+        echo "hyriseConsole did not produce data/hits.bin" >&2
+        exit 1
+    fi
+    rm -f data/hits_part_*.csv data/hits_part_*.csv.json data/make_snapshot.script
+fi
+
+psql -h 127.0.0.1 -p 5432 -U postgres -v ON_ERROR_STOP=1 -q \
+    -c "COPY hits FROM '/data/hits.bin' WITH (FORMAT BINARY);"
+sync