MongoDB Facts: Lightning Fast Aggregation

In my previous post, I demonstrated how fast you can insert 50 millions time-event entries with MongoDB. This time we will make use of all that data to fuel our aggregation tests.

This is how one time-event entry looks like:

{
        "_id" : ObjectId("529a2a988cccdb538932d31f"),
        "created_on" : ISODate("2012-05-02T06:08:47.835Z"),
        "value" : 0.9270193106494844
}

Beside the default primary key “_id” index, we also created one for the “created_on” field, so these are all our indexes:

[
        {
                "v" : 1,
                "key" : {
                        "_id" : 1
                },
                "ns" : "random.randomData",
                "name" : "_id_"
        },
        {
                "v" : 1,
                "key" : {
                        "created_on" : 1
                },
                "ns" : "random.randomData",
                "name" : "created_on_1"
        }
]

Now let’s use all 50 million entries to build a daily report, counting how many events were generated in a day, including the minimum, the maximum and the average value for that particular day.

This is how our script looks like:

var start = new Date();
var dataSet = db.randomData.aggregate([
    {
        $group: {
                "_id": {
                    "year" : {
                        $year : "$created_on"
                    },
                    "dayOfYear" : {
                        $dayOfYear : "$created_on"
                    }
                },
                "count": {
                    $sum: 1
                },
                "avg": {
                    $avg: "$value"
                },
                "min": {
                    $min: "$value"
                },
                "max": {
                    $max: "$value"
                }      
            }
    },
    {
        $sort: {
            "_id.year" : 1,
            "_id.dayOfYear" : 1
        }  
    }
]);
if(dataSet.result != null && dataSet.result.length > 0) {
    print("Aggregated:" + dataSet.result.length + " days.");   
    db.dailyReport.insert(dataSet.result);
}
var end = new Date();
print("Aggregation took:" + (end.getTime() - start.getTime())/1000 + "s");

After aggregating all data, the results are saved to a new dailyReport collection.
Let’s run the script and see what we get:

D:\wrk\vladmihalcea\mongodb-facts\aggregator\timeseries>mongo random aggregate_daily_report.js
MongoDB shell version: 2.4.6
connecting to: random
Aggregated:367 days.
Aggregation took:129.052s

So, in 129 seconds we managed to build our report for all this data. Let’s check the new collection and see our daily reports.

{
        "_id" : {
                "year" : 2012,
                "dayOfYear" : 1
        },
        "count" : 137244,
        "avg" : 0.5009360724400802,
        "min" : 0.0000013632234185934067,
        "max" : 0.9999953350052238
}
{
        "_id" : {
                "year" : 2012,
                "dayOfYear" : 2
        },
        "count" : 136224,
        "avg" : 0.49982110975583033,
        "min" : 0.0000023238826543092728,
        "max" : 0.9999841095414013
}

Since we generated our time-event associated values using Math.random(), the average, minimum and maximum values are what we were expecting anyway. What is really interesting is how fast MongoDB managed to mass all this data at a rate of 387440 documents per second.

Being exited about this result, let’s now check how fast we can randomly select a one-hour report. We first match one-hour span of entries, then we group and sort, to finally display the results to the Mongo shell.

var minDate = new Date(2012, 0, 1, 0, 0, 0, 0);
var maxDate = new Date(2013, 0, 1, 0, 0, 0, 0);
var delta = maxDate.getTime() - minDate.getTime();
var fromDate = new Date(minDate.getTime() + Math.random() * delta);
fromDate.setHours(0, 0, 0, 0);
var toDate = new Date(fromDate.getTime() + 60 * 60 * 1000);

print("Aggregating from " + fromDate + " to " + toDate);

var start = new Date();

var dataSet = db.randomData.aggregate([
    {
        $match: {
            "created_on" : {
                $gte: fromDate,
                $lt : toDate   
            }
        }
    },
    {
        $group: {
                "_id": {
                    "year" : {
                        $year : "$created_on"
                    },
                    "dayOfYear" : {
                        $dayOfYear : "$created_on"
                    },
                    "hour" : {
                        $hour : "$created_on"
                    }
                },
                "count": {
                    $sum: 1
                },
                "avg": {
                    $avg: "$value"
                },
                "min": {
                    $min: "$value"
                },
                "max": {
                    $max: "$value"
                }      
            }
    },
    {
        $sort: {
            "_id.year" : 1,
            "_id.dayOfYear" : 1,
            "_id.hour" : 1
        }  
    }
]);
if(dataSet.result != null && dataSet.result.length > 0) {
    dataSet.result.forEach(function(document)  {
        printjson(document);
    });
}
var end = new Date();
print("Aggregation took:" + (end.getTime() - start.getTime())/1000 + "s");

Running this script we get the following result:

D:\wrk\vladmihalcea\mongodb-facts\aggregator\timeseries>mongo random aggregate_hour_report.js
MongoDB shell version: 2.4.6
connecting to: random
Aggregating from Mon Jul 16 2012 00:00:00 GMT+0300 (GTB Daylight Time) to Mon Jul 16 2012 01:00:00 GMT+0300 (GTB Daylight Time)
{
        "_id" : {
                "year" : 2012,
                "dayOfYear" : 197,
                "hour" : 21
        },
        "count" : 5808,
        "avg" : 0.5015344015735451,
        "min" : 0.00005716201849281788,
        "max" : 0.9998941225931048
}
Aggregation took:0.209s

If you enjoy reading this article, you might want to subscribe to my newsletter and get a discount for my book as well.

This is so fast that I don’t even have to pre-calculate the hour-based reports, meaning I could easily generate it on demand, at run-time.

If you enjoyed this article, I bet you are going to love my book as well.

MongoDB aggregation framework is extremely useful and its performances can’t go unnoticed. What I showed you were only simple examples, that didn’t require any extra optimization, aiming to demonstrate the out-of-the-box performance of MongoDB.

Code available on GitHub.