RFC: Version 8.1 planning - hackystat-dev

Message from discussion RFC: Version 8.1 planning

Philip Johnson

More options Jan 21 2008, 8:00 pm

From: Philip Johnson <john...@hawaii.edu>

Date: Mon, 21 Jan 2008 15:00:44 -1000

Local: Mon, Jan 21 2008 8:00 pm

Subject: Re: [hackystat-dev] RFC: Version 8.1 planning

  Reply to author
  |
  Forward
  | Print
  | View thread
  | Show original
  | Report this message
  | Find messages by this author
  

--On January 21, 2008 1:54:13 PM -1000 Daniel N Port

<dp...@hawaii.edu> wrote:
> I'm anxious to write a FFT or Bayesian network telemetry analysis
> package because I'm curious to see what might pop out. Do you think
> v.81 is a good target for this?

yes, it's perfect.

> Hopefully it will be fairly straightforward to implement with the
> v.8 architecture (seems so) but I've kind of been waiting for the
> architecture to stabilize a bit before launching into it. Is there
> an analogous example package I might adapt to get things started?

Well, it depends. The most "natural" approach is as two hackystat
services:

hackystat-analysis-fft
- retrieves data from telemetry service, performs FFT, exposes its
results through a REST API.

hackystat-ui-fft
- a user interface that queries hackystat-analysis-fft and displays
the results.

Splitting things up this way is nice because:
- we can easily implement multiple UIs.
- we can integrate FFT results into other UIs.

If you want to follow this approach, then the best examples are
hackystat-analysis-telemetry and hackystat-ui-telemetryviewer.

If you want to simply hack something together quickly (i.e. the
"spike solution"), then just build something in PHP or whatever you
want that makes HTTP calls to the telemetry analysis service to get
the data and then passes it off to whatever FFT library you can come
up with. Don't even bother with google project hosting if you don't
want. Just start hacking.

> Also, is there a decent amount of telemetry data to analyze at this

point?

No, but don't let that stop you. Look at the
hackystat-sensorbase-simdata package, which contains a
"simpletelemetry" package which generates the sensor data that gets
processed for the telemetry tutorial:

<http://code.google.com/p/hackystat/wiki/Tutorial_SoftwareProjectTelem...>

What you can do is write your own package in
hackystat-sensorbase-simdata called "ffttelemetry", which would
generate simulated data that you could use to play with your system.

That's actually a really nice way to do development; get something
that seems to work under "controlled conditions" before testing it
out in vivo.

Cheers,
Philip

Reply to author Forward