druid/spectator-histogram.md at beccc401e19cc36e9ba6726d359b4a9247b60d6a

21 KiB

Raw Blame History

id	title
spectator-histogram	Spectator Histogram module

Summary

This module provides Apache Druid approximate histogram aggregators and percentile post-aggregators based on Spectator fixed-bucket histograms.

Consider SpectatorHistogram to compute percentile approximations. This extension has a reduced storage footprint compared to the DataSketches extension, which results in smaller segment sizes, faster loading from deep storage, and lower memory usage. This extension provides fast and accurate queries on large datasets at low storage cost.

This aggregator only applies when your raw data contains positive long integer values. Do not use this aggregator if you have negative values in your data.

In the Druid instance shown below, the example Wikipedia dataset is loaded 3 times.

wikipedia contains the dataset ingested as is, without rollup
wikipedia_spectator contains the dataset with a single extra metric column of type spectatorHistogram for the added column
wikipedia_datasketch contains the dataset with a single extra metric column of type quantilesDoublesSketch for the added column

Spectator histograms average just 6 extra bytes per row, while the quantilesDoublesSketch adds 48 bytes per row. This represents an eightfold reduction in additional storage size for spectator histograms.

As rollup improves, so does the size savings. For example, when you ingest the Wikipedia dataset with day-grain query granularity and remove all dimensions except countryName, this results in a segment that has just 106 rows. The base segment has 87 bytes per row. Compare the following bytes per row for SpectatorHistogram versus DataSketches:

An additional spectatorHistogram column adds 27 bytes per row on average.
An additional quantilesDoublesSketch column adds 255 bytes per row.

SpectatorHistogram reduces the additional storage size by 9.4 times in this example. Storage gains will differ per dataset depending on the variance and rollup of the data.

Background

Spectator is a simple library for instrumenting code to record dimensional time series data. It was built, primarily, to work with Atlas. Atlas was developed by Netflix to manage dimensional time series data for near real-time operational insight.

With the Atlas-Druid service, it's possible to use the power of Atlas queries, backed by Druid as a data store to benefit from high-dimensionality and high-cardinality data.

SpectatorHistogram is designed for efficient parallel aggregations while still allowing for filtering and grouping by dimensions. It provides similar functionality to the built-in DataSketches quantilesDoublesSketch aggregator, but is opinionated to maintain higher absolute accuracy at smaller values. Larger values have lower absolute accuracy; however, relative accuracy is maintained across the range. See Bucket boundaries for more information. The SpectatorHistogram is optimized for typical measurements from cloud services and web apps, such as page load time, transferred bytes, response time, and request latency.

Through some trade-offs SpectatorHistogram provides a significantly more compact representation with the same aggregation performance and accuracy as DataSketches Quantiles Sketch. Note that results depend on the dataset. Also see the [limitations](#limitations] of this extension.

Limitations

Supports positive long integer values within the range of [0, 2^53). Negatives are coerced to 0.
Does not support decimals.
Does not support Druid SQL queries, only native queries.
Does not support vectorized queries.
Generates 276 fixed buckets with increasing bucket widths. In practice, the observed error of computed percentiles ranges from 0.1% to 3%, exclusive. See Bucket boundaries for the full list of bucket boundaries.

:::tip If these limitations don't work for your use case, then use DataSketches instead. :::

Functionality

The SpectatorHistogram aggregator can generate histograms from raw numeric values as well as aggregating or combining pre-aggregated histograms generated using the SpectatorHistogram aggregator itself. While you can generate histograms on the fly at query time, it is generally more performant to generate histograms during ingestion and then combine them at query time. This is especially true where rollup is enabled. It may be misleading or incorrect to generate histograms from already rolled-up summed data.

The module provides postAggregators, percentileSpectatorHistogram (singular) and percentilesSpectatorHistogram (plural), to compute approximate percentiles from histograms generated by the SpectatorHistogram aggregator. Again, these postAggregators can be used to compute percentiles from raw numeric values via the SpectatorHistogram aggregator or from pre-aggregated histograms.

If you're only using the aggregator to compute percentiles from raw numeric values, then you can use the built-in quantilesDoublesSketch aggregator instead. The performance and accuracy are comparable. However, the DataSketches aggregator supports negative values, and you don't need to download an additional extension.

An aggregated SpectatorHistogram can also be queried using a longSum or doubleSum aggregator to retrieve the population of the histogram. This is effectively the count of the number of values that were aggregated into the histogram. This flexibility can avoid the need to maintain a separate metric for the count of values.

For high-frequency measurements, you may need to pre-aggregate data at the client prior to sending into Druid. For example, if you're measuring individual image render times on an image-heavy website, you may want to aggregate the render times for a page-view into a single histogram prior to sending to Druid in real-time. This can reduce the amount of data that's needed to send from the client across the wire.

SpectatorHistogram supports ingesting pre-aggregated histograms in real-time and batch. They can be sent as a JSON map, keyed by the spectator bucket ID and the value is the count of values. This is the same format as the serialized JSON representation of the histogram. The keys need not be ordered or contiguous. For example:

{ "4":  8, "5": 15, "6": 37, "7": 9, "8": 3, "10": 1, "13": 1 }

Loading the extension

To use SpectatorHistogram, make sure you include the extension in your config file:

druid.extensions.loadList=["druid-spectator-histogram"]

Aggregators

The result of the aggregation is a histogram that is built by ingesting numeric values from the raw data, or from combining pre-aggregated histograms. The result is represented in JSON format where the keys are the bucket index and the values are the count of entries in that bucket.

The buckets are defined as per the Spectator PercentileBuckets specification. See Histogram bucket boundaries for the full list of bucket boundaries.

  // The set of buckets is generated by using powers of 4 and incrementing by one-third of the
  // previous power of 4 in between as long as the value is less than the next power of 4 minus
  // the delta.
  //
  // Base: 1, 2, 3
  //
  // 4 (4^1), delta = 1 (~1/3 of 4)
  //     5, 6, 7, ..., 14,
  //
  // 16 (4^2), delta = 5 (~1/3 of 16)
  //    21, 26, 31, ..., 56,
  //
  // 64 (4^3), delta = 21 (~1/3 of 64)
  // ...

There are multiple aggregator types included, all of which are based on the same underlying implementation. If you use the Atlas-Druid service, the different types signal the service on how to handle the resulting data from a query.

spectatorHistogramTimer signals that the histogram is representing a collection of timer values. It is recommended to normalize timer values to nanoseconds at, or prior to, ingestion. If queried via the Atlas-Druid service, it will normalize timers to second resolution at query time as a more natural unit of time for human consumption.
spectatorHistogram and spectatorHistogramDistribution are generic histograms that can be used to represent any measured value without units. No normalization is required or performed.

`spectatorHistogram` aggregator

Alias: spectatorHistogramDistribution, spectatorHistogramTimer

To aggregate at query time:

{
  "type" : "spectatorHistogram",
  "name" : <output_name>,
  "fieldName" : <column_name>
 }

Property	Description	Required?
type	This String must be one of "spectatorHistogram", "spectatorHistogramTimer", "spectatorHistogramDistribution"	yes
name	A String for the output (result) name of the aggregation.	yes
fieldName	A String for the name of the input field containing raw numeric values or pre-aggregated histograms.	yes

`longSum`, `doubleSum` and `floatSum` aggregators

To get the population size (count of events contributing to the histogram):

{
  "type" : "longSum",
  "name" : <output_name>,
  "fieldName" : <column_name_of_aggregated_histogram>
 }

Property	Description	Required?
type	Must be "longSum", "doubleSum", or "floatSum".	yes
name	A String for the output (result) name of the aggregation.	yes
fieldName	A String for the name of the input field containing pre-aggregated histograms.	yes

Post Aggregators

Percentile (singular)

This returns a single percentile calculation based on the distribution of the values in the aggregated histogram.

{
  "type": "percentileSpectatorHistogram",
  "name": <output name>,
  "field": {
    "type": "fieldAccess",
    "fieldName": <name of aggregated SpectatorHistogram>
  },
  "percentile": <decimal percentile, e.g. 50.0 for median>
}

Property	Description	Required?
type	This String should always be "percentileSpectatorHistogram"	yes
name	A String for the output (result) name of the calculation.	yes
field	A field reference pointing to the aggregated histogram.	yes
percentile	A single decimal percentile between 0.0 and 100.0	yes

Percentiles (multiple)

This returns an array of percentiles corresponding to those requested.

{
  "type": "percentilesSpectatorHistogram",
  "name": <output name>,
  "field": {
    "type": "fieldAccess",
    "fieldName": <name of aggregated SpectatorHistogram>
  },
  "percentiles": [25, 50, 75, 99.5]
}

It's more efficient to request multiple percentiles in a single query than to request individual percentiles in separate queries. This array-based helper is provided for convenience and has a marginal performance benefit over using the singular percentile post-aggregator multiple times within a query. The more expensive part of the query is the aggregation of the histogram. The post-aggregation calculations all happen on the same aggregated histogram.

The results contain arrays matching the length and order of the requested array of percentiles.

"percentilesAdded": [
    0.5504911679884643, // 25th percentile
    4.013975155279504,  // 50th percentile 
    78.89518317503394,  // 75th percentile
    8580.024999999994   // 99.5th percentile
]

Property	Description	Required?
type	This String should always be "percentilesSpectatorHistogram"	yes
name	A String for the output (result) name of the calculation.	yes
field	A field reference pointing to the aggregated histogram.	yes
percentiles	Non-empty array of decimal percentiles between 0.0 and 100.0	yes

Examples

Example Ingestion Spec

Example of ingesting the sample Wikipedia dataset with a histogram metric column:

{
  "type": "index_parallel",
  "spec": {
    "ioConfig": {
      "type": "index_parallel",
      "inputSource": {
        "type": "http",
        "uris": ["https://druid.apache.org/data/wikipedia.json.gz"]
      },
      "inputFormat": { "type": "json" }
    },
    "dataSchema": {
      "granularitySpec": {
        "segmentGranularity": "day",
        "queryGranularity": "minute",
        "rollup": true
      },
      "dataSource": "wikipedia",
      "timestampSpec": { "column": "timestamp", "format": "iso" },
      "dimensionsSpec": {
        "dimensions": [
          "isRobot",
          "channel",
          "flags",
          "isUnpatrolled",
          "page",
          "diffUrl",
          "comment",
          "isNew",
          "isMinor",
          "isAnonymous",
          "user",
          "namespace",
          "cityName",
          "countryName",
          "regionIsoCode",
          "metroCode",
          "countryIsoCode",
          "regionName"
        ]
      },
      "metricsSpec": [
        { "name": "count", "type": "count" },
        { "name": "sum_added", "type": "longSum", "fieldName": "added" },
        {
          "name": "hist_added",
          "type": "spectatorHistogram",
          "fieldName": "added"
        }
      ]
    },
    "tuningConfig": {
      "type": "index_parallel",
      "partitionsSpec": { "type": "hashed" },
      "forceGuaranteedRollup": true
    }
  }
}

Example Query

Example query using the sample Wikipedia dataset:

{
  "queryType": "timeseries",
  "dataSource": {
    "type": "table",
    "name": "wikipedia"
  },
  "intervals": {
    "type": "intervals",
    "intervals": [
      "0000-01-01/9999-12-31"
    ]
  },
  "granularity": {
    "type": "all"
  },
  "aggregations": [
    {
      "type": "spectatorHistogram",
      "name": "histogram_added",
      "fieldName": "added"
    }
  ],
  "postAggregations": [
    {
      "type": "percentileSpectatorHistogram",
      "name": "medianAdded",
      "field": {
        "type": "fieldAccess",
        "fieldName": "histogram_added"
      },
      "percentile": "50.0"
    }
  ]
}

Results in

[
  {
    "result": {
      "histogram_added": {
        "0": 11096, "1": 632, "2": 297, "3": 187, "4": 322, "5": 161,
        "6": 174, "7": 127, "8": 125, "9": 162, "10": 123, "11": 106,
        "12": 95, "13": 104, "14": 95, "15": 588, "16": 540, "17": 690,
        "18": 719, "19": 478, "20": 288, "21": 250, "22": 219, "23": 224,
        "24": 737, "25": 424, "26": 343, "27": 266, "28": 232, "29": 217,
        "30": 171, "31": 164, "32": 161, "33": 530, "34": 339, "35": 236,
        "36": 181, "37": 152, "38": 113, "39": 128, "40": 80, "41": 75,
        "42": 289, "43": 145, "44": 138, "45": 83, "46": 45, "47": 46,
        "48": 64, "49": 65, "50": 71, "51": 421, "52": 525, "53": 59,
        "54": 31, "55": 35, "56": 8, "57": 10, "58": 5, "59": 4, "60": 11,
        "61": 10, "62": 5, "63": 2, "64": 2, "65": 1, "67": 1, "68": 1,
        "69": 1, "70": 1, "71": 1, "78": 2
      },
      "medianAdded": 4.013975155279504
    },
    "timestamp": "2016-06-27T00:00:00.000Z"
  }
]

Histogram bucket boundaries

The following array lists the upper bounds of each bucket index. There are 276 buckets in total. The first bucket index is 0 and the last bucket index is 275. The bucket widths increase as the bucket index increases. This leads to a greater absolute error for larger values, but maintains a relative error of rough percentage across the number range. For example, the maximum error at value 10 is zero since the bucket width is 1 (the difference of 11-10). For a value of 16,000,000,000, the bucket width is 1,431,655,768 (from 17179869184-15748213416). This gives an error of up to ~8.9%, from 1,431,655,768/16,000,000,000*100. In practice, the observed error of computed percentiles is in the range of (0.1%, 3%).

[
  1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16, 21, 26, 31, 36, 41, 46,
  51, 56, 64, 85, 106, 127, 148, 169, 190, 211, 232, 256, 341, 426, 511, 596,
  681, 766, 851, 936, 1024, 1365, 1706, 2047, 2388, 2729, 3070, 3411, 3752,
  4096, 5461, 6826, 8191, 9556, 10921, 12286, 13651, 15016, 16384, 21845,
  27306, 32767, 38228, 43689, 49150, 54611, 60072, 65536, 87381, 109226,
  131071, 152916, 174761, 196606, 218451, 240296, 262144, 349525, 436906,
  524287, 611668, 699049, 786430, 873811, 961192, 1048576, 1398101, 1747626,
  2097151, 2446676, 2796201, 3145726, 3495251, 3844776, 4194304, 5592405,
  6990506, 8388607, 9786708, 11184809, 12582910, 13981011, 15379112, 16777216,
  22369621, 27962026, 33554431, 39146836, 44739241, 50331646, 55924051,
  61516456, 67108864, 89478485, 111848106, 134217727, 156587348, 178956969,
  201326590, 223696211, 246065832, 268435456, 357913941, 447392426, 536870911,
  626349396, 715827881, 805306366, 894784851, 984263336, 1073741824, 1431655765,
  1789569706, 2147483647, 2505397588, 2863311529, 3221225470, 3579139411,
  3937053352, 4294967296, 5726623061, 7158278826, 8589934591, 10021590356,
  11453246121, 12884901886, 14316557651, 15748213416, 17179869184, 22906492245,
  28633115306, 34359738367, 40086361428, 45812984489, 51539607550, 57266230611,
  62992853672, 68719476736, 91625968981, 114532461226, 137438953471,
  160345445716, 183251937961, 206158430206, 229064922451, 251971414696,
  274877906944, 366503875925, 458129844906, 549755813887, 641381782868,
  733007751849, 824633720830, 916259689811, 1007885658792, 1099511627776,
  1466015503701, 1832519379626, 2199023255551, 2565527131476, 2932031007401,
  3298534883326, 3665038759251, 4031542635176, 4398046511104, 5864062014805,
  7330077518506, 8796093022207, 10262108525908, 11728124029609, 13194139533310,
  14660155037011, 16126170540712, 17592186044416, 23456248059221,
  29320310074026, 35184372088831, 41048434103636, 46912496118441,
  52776558133246, 58640620148051, 64504682162856, 70368744177664,
  93824992236885, 117281240296106, 140737488355327, 164193736414548,
  187649984473769, 211106232532990, 234562480592211, 258018728651432,
  281474976710656, 375299968947541, 469124961184426, 562949953421311,
  656774945658196, 750599937895081, 844424930131966, 938249922368851,
  1032074914605736, 1125899906842624, 1501199875790165, 1876499844737706,
  2251799813685247, 2627099782632788, 3002399751580329, 3377699720527870,
  3752999689475411, 4128299658422952, 4503599627370496, 6004799503160661,
  7505999378950826, 9007199254740991, 10508399130531156, 12009599006321321,
  13510798882111486, 15011998757901651, 16513198633691816, 18014398509481984,
  24019198012642645, 30023997515803306, 36028797018963967, 42033596522124628,
  48038396025285289, 54043195528445950, 60047995031606611, 66052794534767272,
  72057594037927936, 96076792050570581, 120095990063213226, 144115188075855871,
  168134386088498516, 192153584101141161, 216172782113783806, 240191980126426451,
  264211178139069096, 288230376151711744, 384307168202282325, 480383960252852906,
  576460752303423487, 672537544353994068, 768614336404564649, 864691128455135230,
  960767920505705811, 1056844712556276392, 1152921504606846976, 1537228672809129301,
  1921535841011411626, 2305843009213693951, 2690150177415976276, 3074457345618258601,
  3458764513820540926, 3843071682022823251, 4227378850225105576, 9223372036854775807
]

21 KiB Raw Blame History