Dask Demo Day: Fine performance metrics and spans

DemoDay.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "070dfc8c-2bf2-45d5-b515-195e0b6f367c",
   "metadata": {},
   "source": [
    "# Dask Demo Day 2023-06-16\n",
    "## Fine Performance Metrics and Spans\n",
    "#### Presenter: Guido Imperiale (Coiled)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b0adab3a-c0d8-4bbb-af95-5c22baa55631",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import time\n",
    "\n",
    "import dask\n",
    "import dask.array as da\n",
    "from dask.datasets import timeseries\n",
    "from distributed import Client\n",
    "from distributed.metrics import context_meter\n",
    "from distributed.spans import span\n",
    "\n",
    "# Spans are built on top of annotations. In order to make \n",
    "# annotations work in dask.array, we need to disable task fusion.\n",
    "dask.config.set({\"optimization.fuse.active\": False})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "138e53af-357f-45e7-9065-1de997b90acb",
   "metadata": {},
   "outputs": [],
   "source": [
    "client = Client()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ffdcc199-ea1d-40ed-9426-dece70d88a83",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "ddf = timeseries(\"2000\", \"2020\", partition_freq=\"1w\")\n",
    "ddf.groupby(\"id\").x.mean().compute()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e8555424-22e9-4049-a1cf-666bafeb8cc0",
   "metadata": {},
   "source": [
    "(please restart notebook)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0b353ae4-37ce-4034-8a8b-2dbac7dcf8a7",
   "metadata": {},
   "outputs": [],
   "source": [
    "client = Client(n_workers=4, threads_per_worker=1, memory_limit=\"2 GiB\")\n",
    "rng = da.random.RandomState(0)\n",
    "a = rng.random((14_000, 14_000))\n",
    "b = (a @ a.T).sum()\n",
    "b.compute()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6df7dca-307d-4c76-9499-e130d8db96fb",
   "metadata": {},
   "source": [
    "(please restart notebook)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "884e37ee-d5e7-42c4-b407-064891488a4a",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "client = Client(n_workers=4, threads_per_worker=1, memory_limit=\"2 GiB\")\n",
    "\n",
    "@span(\"dot_product_self\")\n",
    "def dot_product_self(a):\n",
    "    return a @ a.T\n",
    "\n",
    "with span(\"my workload\"):\n",
    "    with span(\"data load\"):\n",
    "        rng = da.random.RandomState(0)\n",
    "        a = rng.random((14_000, 14_000))\n",
    "    b = dot_product_self(a)\n",
    "    c = b.sum()\n",
    "\n",
    "c.compute()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1650a860-9e21-4268-ba14-15045ad24577",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "with span(\"timeseries\"):\n",
    "    ddf = timeseries(\"2000\", \"2020\", partition_freq=\"1w\")\n",
    "    ddf.groupby(\"id\").x.mean().compute()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "079fbc72-e653-45cb-bffa-ee4869eb180e",
   "metadata": {},
   "source": [
    "(please restart notebook)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "071e1d13-7a7c-442d-a0fc-25d0c0a45902",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "client = Client()\n",
    "\n",
    "def io_task():\n",
    "    time.sleep(0.8)\n",
    "    with context_meter.meter(\"I/O\"):\n",
    "        time.sleep(0.5)\n",
    "    context_meter.digest_metric(\"metered I/O\", 2**30, \"bytes\")\n",
    "\n",
    "def gpu_task():\n",
    "    with context_meter.meter(\"GPU\"):\n",
    "        time.sleep(0.7)\n",
    "    context_meter.digest_metric(\"task-gpu\", 123, \"gpu_seconds\")\n",
    "    \n",
    "client.gather([\n",
    "    client.submit(io_task),\n",
    "    client.submit(gpu_task),\n",
    "])       "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "30ffb3cc-7af0-4fb9-a8ba-51f83448cdd8",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}