Project Neuron

 

Introduction

An important drop in the BOINC ocean

This project aims to provide a trial BOINC environment in which a set of dummy applications will run. The purpose of this being to record, observe and understand BOINC activity and data with a view to developing metrics that will establish or otherwise the quality/reliability/dependability of particular BOINC projects. A central reference point will be developed and updated automatically to which users can refer. User feedback may also be permitted at this reference point.

Background

Many will know that some BOINC projects are better resourced, run and managed than others. It is inevitable that some research will be better funded than other research. A concern of a few BOINC users is that its hard to distinguish between a well run project where volunteer CPU resources are used effectively and a badly run project where volunteer CPU time is being wasted. The environmental cost alone says volunteers will want some reassurances their resources are not wasted. Equally volunteers want to see feedback on how the science is progressing. It is often the case with BOINC projects that no news really is bad news. This project is an attempt to differentiate between the well run and the badly run. What this is NOT is an attempt to determine if a project and its aims are worthwhile or laudable or acceptable. What is IS is an attempt to collect data from a BOINC project with a view to interpreting that data to help a user work out if that project is sufficiently well run for them to join or continue with it.

There are 4 distinct phases to the project.

Phase 1

This research needs volunteers to run this BOINC project to produce work, download it, process the data, return result, validate, assimilate and delete it just as any BOINC project would normally do. It needs to run a forum, have news, send mails and generally act as a regular project, The data gathered in the project's BOINC database and or static and dynamic information on the server will be examined to see what data is available for analysis use and what tools might need to be created to exploit this data. It may be there is sufficient data to achieve the goal or some part of it but if not then the project will offer advice to UCB on what might be collected in order to better inform users.

Phase 2

Tools will be built that will be project admin based but will provide a page on the project web site that users can refer to in order to get some quality indicator. Questions might be for example:

Do the project team post regular updates about their work to keep volunteers in touch with

  • project developments?
  • project problems?
  • application fixes?
  • News and RSS feed up to date and regular?

Does the project keep up to date with BOINC versions to bring users a better service?

Is there a regular supply of work?

Is work validated and assimilated on a timely basis giving users their credits in turn on timely basis?

Do back-logs build up that slow down the delivery of work and subsequent results?

Is server status available?

Are backups of database taken to secure the volunteer effort?

What is the error rate for work against applications and application  versions and do developers respond to problems to make all results "valuable"?

Phase 2 is a discovery process to explore what questions can be asked and what answers provided by the system's data. From the data collected it may be possible to provide an automated answer to these and other questions.

Phase 3

A scoring system will be developed which will weight, score and rank these reliability criteria to produce a result. This result will help users make choices over which projects they may wish to join.

Phase 4

The final phase is to:

To persuade and get sign up from BOINC and other projects to the approach of quality/reliability metrics in their projects.

To see the tools  produced, or variations of them, become part of BOINC per se.

To establish the central reference point to collate and distribute project data to users.

How long will this all take?

Not at all sure to be frank. Certainly several months and may be as long as a year as it is part time activity.

Is this being done with David Anderson's agreement?

No. Dr Anderson never approves nor disapproves of any project. Matters relating to the quality and bone fides of BOINC projects have been raised with Dr Anderson but he has not endorsed this work at all. He is sympathetic to seeing BOINC used well and exploited for what it can achieve for science. He sees the arrangements made for a project by each project team as their own affair. Some of us differ slightly with that view and are concerned about wasted CPU time and rogue projects albeit this work does not address rogue project activity.

 



Return to Project Neuron main page


Copyright © 2008 neuron.mine.nu