MEADEP

Sophisticated System Reliability - Availability Modelling Made Simple!

MEADEP - Measurement-Based Dependability Analysis Tool.

MEADEP consists of four modules.
These modules are:

Data Pre-Processor (DPP)
Data Editor and Analyzer (DEA)
Model Generator (MG)
Model Evaluator (ME)

Figure 1 below shows the overview of the data flow of the program and how each of MEADEP's modules interact with each other.

Results generated by MEADEP are either directly obtained from data or evaluated from dependability models. Thus, the two basic types of input to MEADEP are:

Data: structured reports containing information on event time, location, impact and other event characteristics, and Models: graphical specifications of dependability models consisting of reliability blocks and Markov chains. In order for MEADEP to work on data, the data must be in the MEADEP required format. Therefore, MEADEP includes the Data Pre-Processor (DPP) module which converts existing databases (which can be in a variety of formats) into the MEADEP data format.

However, if the data is not contained in a file, and you need to input data manually, then the Data Editor and Analyzer (DEA) module can be used. The DEA module enables you to created MEADEP formatted databases and input the data directly into it without any conversion. This is useful if the data is contained in hand written event logs. After converting the data to the MEADEP required format, the Data Editor and Analyzer (DEA) module can be used to work on the data and perform graphical and statistical analysis on it.

As mentioned above, an alternative way to input data is by building dependability models. The Model Generator (MG) module can be used to do this. From these models, MG can generate text modelling files which contain model descriptions and parameters that the Model Evaluator (ME) module uses to generate the desired results. Results evaluated from models include: Mean Time Between Failures (MTBF), Reliability for a given time period, and Steady-state availability. ME enables you to generate these results using a user-specified range of values for a selected parameter, and the results can then be displayed graphically.

For all of its functions, MEADEP provides you with a graphical user interface (GUI) on Windows 95, 98, NT, 2000 or Me featuring menus, dialogs, pictures, printing previews, and extensive on-line help information.

The Data Pre-Processor (DPP) Module

The Data Pre-Processor (DPP) Module enables you to translate existing reliability and failure databases (not in MEADEP format) to the MEADEP required format. Data formats supported by MEADEP include ASCII delimited text and a variety of databases such as Access, dBASE, and Paradox. MEADEP also supports other formats on your system which have existing Open DataBase Connectivity (ODBC) Drivers attached. MEADEP data are stored in records, where each record represents a single event, in the Access database format.

The Data Editor and Analyzer (DEA) Module

The Data Editor and Analyzer (DEA) module works on data and performs statistical analysis on it. The results of DEA's statistical analysis can be graphed or bound to parameters that are found in text modelling files (created in the Model Generator module).

DEA has three major functions:

1. Data Editing- Data Editing includes correctness checking for data formatting, querying records, sorting records, adding records, deleting records, undo changes, consolidating fields, saving records, assigning a constant string to a field and more.

2. Graphical Analysis (click to see examples)- Graphical Analysis can generate pie charts for event distribution, histograms for Time Between Events (TBE) or Time To Recovery/Repair (TTR) distributions with the option to superimpose typical analytical functions accompanied by the results of their goodness-of-fit tests, and progressive curves over time for Mean Time Between Events (MTBE) and its confidence interval. Graphical analysis can be specified by the user through multiple window dialogs.

3. Parameter Estimation- Parameter Estimation provides the mean, upper and lower bounds at a specified confidence level for the following: Mean Time Between Failures (MTBF), Mean Time To Recovery/Repair (MTTR), failure rate, recovery/repair rate, and fault-tolerance coverage. Estimates are also given even if failures are rare. These estimates can then be inserted into a text modelling file for binding to model parameters. Parameter estimation can be specified by the user through multiple window dialogs and can also be specified by a predefined query command file.

The Model Generator (MG) Module

The Model Generator (MG) module is a graphical drag and drop interface for constructing reliability and availability models. A model is developed hierarchically, from the top level down. Each level can be one of the following:

A diagram of serial or parallel reliability blocks (block diagram), A k-out-of-n model (block diagram) or A Markov chain (Markov diagram).

A reliability block diagram is a graphical method of depicting the components in a system and their connections in terms of functioning requirements. Each component can be represented by a block. Block diagrams must be in the following format:

At least one block, exactly two terminals (source and destination) and at least two links. The figure below shows a parallel system block diagram with appropriate links and terminals.

A Markov Model consists of system states and transitions from one state to another. A system state represents a combination of both operational and failed components in the system. The system stays in a state for a random time, defined by an exponential distribution, and then transitions into another state. A transition from one state to another state is characterized by a transition rate. A Markov model can be solved mathematically to obtain reliability and availability measures. For example, the expected proportion of time that the system spends in the failure states, which is called the system unavailability, can be calculated.

For a Markov diagram, you can:

Draw states and transition arcs between states, specify a reward value for each state, specify a transition rate for each transition arc and specify the initial state and the failure state for the model. For example, as you can see in the Markov diagram below, there are at least two states, at least one transition with its appropriate transition rate, an initial state and a final state. In this example, the parameter "lambda" represents the failure rate and the parameter ?represents the recovery rate.

When the model construction is completed, the diagrams are saved in a graphical modelling file (*.mdg) for reuse. From the model, MG can generate a text modelling file (*.mdt). This modelling file contains the model specifications which the Model Evaluator (ME) module uses to evaluate the model and obtain results.

MG is also capable of using pre-designed library files (*.mdl) for increasing productivity. A library file is a graphical modelling file that defines the structure of a dependability model but does not contain parameter values. This capability allows you to re-use previously developed and tested models and can greatly reduce model construction time.

MG also allows you to save a model diagram as a Microsoft Windows metafile. This metafile can then be imported into word processors and other Windows programs.

The Model Evaluator (ME) Module

The MEADEP Model Evaluator (ME) module has two major functions:

Editing text modelling files and Evaluating models defined in the text modelling file. ME enables you to revise parameters and models and then to calculate results based on these revisions. In addition, ME can also perform parametric analysis on the data and can graphically display the results. For this analysis, you can choose from the following four loop types:

Data Pre-Processor (DPP)
Data Editor and Analyzer (DEA)
Model Generator (MG)
Model Evaluator (ME)

ME also allows you to create and edit parameter files and include them in the model evaluation process. This provision allows you to include a standard parameter list in multiple modelling files without having to input these parameter values into each file.