Basically a re-implementation of hasning.py in the Carbon-project.

The current way Graphite is built, only one-second resolution of data since the UNIX epoch is allowed. But should that keep us from doing better internally?

• It takes up a wee bit more space per metric.
• It does give us a lot of freedom internally.
• Lots of exciting time-manipulation options for free.

From reading about Graphite (and reading implementations) it does seem there is a heavy emphasis on batching a lot of data for one metric.

GraphiteBase should have a sub-system for that.

type BulkMetric struct {
    Name string
    Data []MetricValues
}

And then a system to make a channel of Metrics into a channel of BulkMetrics.

• Setting how many metrics it should buffer before spewing out bulk data
• Allow for querying