Node.js API for storing and querying time series in Redis

Convenient module for storing and querying time series statistics in Redis using Node.js.

The design (and even parts of the implementation) were picked from the ApiAxle project.

You can find basic usage examples in examples. This module also powers a real-time dashboard written in Node.js. Check the sources out for more insight.

Dependencies

redis-timeseries has no dependencies, and will work along the redis module you'll install in your own project. redis@~0.9.0 versions are compatible.

Usage

	var TimeSeries = require('redis-timeseries'),
		redis = require('redis').createClient();
		
	// Create the TimeSeries client
	//
	// "stats" is the Redis namespace which will be used
	// for storing all the TimeSeries related keys
    //
	// "granularities" encodes the granularities at which
	// you want to store statistics. More on that in the next section
	//
	var ts = new TimeSeries(redis, "stats", granularities);
	
	// Recording hits
	//
	// This increments the counters for the
	// stats keys you provide
	//
	// "timestamp" defaults to the current time
    // "increment" defaults to 1
	//
	ts.recordHit('your_stats_key')
	  .recordHit('another_stats_key', timestamp)
      .recordHit('another_stats_key', timestamp2, increment)
	  …
	  .exec();

    // Removing hits
    //
    // It's also possible to decrement the hits counter for
    // some key
    ts.removeHit('your_stats_key', [timestamp]).exec();

    // Recording values
	//
	// This sets the value for the
	// stats keys you provide
	//
	// "timestamp" defaults to the current time
    // "increment" defaults to 1
	//
	ts.recordValue('your_stats_key', value)
	  .recordValue('another_stats_key', value, timestamp)
      .recordValue('another_stats_key', value, timestamp2, increment)
	  …
	  .exec();
	  
	// Querying statistics
	//
	// Returns "count" chunks of counters at the precision described by
	// "granularity_label"
	// 
	ts.getHits('your_stats_key', granularity_label, count, function(err, data) {
		// data.length == count
		// data = [ [ts1, count1], [ts2, count2]... ]
	});

	// getValues is identical to getHits except that it returns a null value for unset timestamps
	ts.getValues('your_stats_key', granularity_label, count, function(err, data) {
		// data.length == count
		// data = [ [ts1, count1], [ts2, count2]... ]
	});

Defining custom statistics granularities

For each key, TimeSeries stores statistics at different granularities. For further information about this, please refer to the detailed blog post from the ApiAxle project.

The default granularities are:

{
    '1second'  : { ttl: this.minutes(5), duration: 1 },
    '1minute'  : { ttl: this.hours(1)  , duration: this.minutes(1) },
    '5minutes' : { ttl: this.days(1)   , duration: this.minutes(5) },
    '10minutes': { ttl: this.days(1)   , duration: this.minutes(10) },
    '1hour'    : { ttl: this.days(7)   , duration: this.hours(1) },
    '1day'     : { ttl: this.weeks(52) , duration: this.days(1) }
}

This means that the number of hits per second will be stored for 5 minutes, and the corresponding hashset will expire afterwards. Likewise, the number of hits per minute for a given key will be kept for an hour. Daily counters on the other hand are kept for a full year.

When querying for statistics, a granularity label is expected:

	// Give me the hits/second for the last 3 minutes
	ts.getHits('your_stats_key', '1second', ts.minutes(3), callback);
	
	// Give me the number of hits per day for the last 2 weeks
	ts.getHits('your_stats_key', '1day', 14, callback);
	
	// And so on

When creating the TimeSeries client, you can override the default granularities with your own.