What we store

The key items we store in PerformDB are:

Models

the descriptions of individual systems, or components which may be combined to make larger systems.

Measures

queries we can pose of models in the system. They may be tied to a particular model in the system, or phrased so that they may apply to many different models.

Parameters

A measure may stand on its own, but may also have some free parameters. These allow us to represent a range of values for which a particular model has been analysed. Storing these succinctly is important as although we may wish to extract the result from an individual set of parameters, very often we will want to treat the sweep as a single evaluation and display that collected result together.

Results

The actual numeric results from a calculation. These are typically stored as tab-separated values in ASCII, which is a convenient input for many graphing tools.

We are exploring suitable formats to store symbolic results.

Each of these four items has its own list of extensible types. Any user can add, for example, a new model type, and then add their models marked as that type. In order that we can fully support you new model format (with syntax checking and static analysis), please get in touch.

Who can change what?

Roughly, anyone can change anything. We keep track of which user performed an action, so we can revert if anyone abuses the system.

You can also make changes without creating an account, and your contributions will be assigned to 'nobody', the anonymous user.

More to follow :

• using performance tools on our cluster from the site
• using PerformDB from inside tools on your own systems