madbence/node-round-robin-database

round-robin-database

In-memory Round Robin Database implementation for storing time-series data. It's pertty much the same as whisper.

Usage

var rrd = require('round-robin-database')
var DB = rrd.DB;
var Layer = rrd.Layer;

var db = new DB({
  layers: [
    new Layer('15s', '1w'), // 15s resolution for a week
    new Layer('1m', '3w'),  // 1m resolution for 3 weeks
    new Layer('1h', '5y')   // 1h resolution for 5 years
  ]
});

console.log(db.getSize()); // 914880, less than 1MB!

// ...

db.write(time, value);

Under the Hood

Data is stored in a single Buffer, which holds {time,value} pairs, every point has a size of 8 bytes (32 bit timestamp, 32 bit float value).

If you wan to store 15s precision values for a week, 1m precision values for 3 weeks, 1h precision values for 5 years, the Buffers size will be (7*24*3600/15+21*24*3600/60+5*365*24*3600/3600)*8/1024 = 893KB. That's pretty good.

On my 3 year old laptop (i5-m450) ~30 records/ms can be written.

License

MIT

node-round-robin-database's Issues

Handle old records

Old records = older than the retention of the highest precision layer.

Add x-files factor

Aggregate NaN when unknown value ratio is higher than xff

Add other aggregation methods

max, min, last, sum, i dunno