A3 architecture

Intro

The A3 architecture is one of the possible Assothink architectures.

It is an extension / rewriting of the previous (less powerfull) versions of Alex and Assothink (2 names for the same project), developped from 2009.

It was developed from 2013 and it is still under development in 2015 ...

The A3 architecture divides the Assothink work in a trilogy of main working layers, plus 3 complementary side parts.

Part	Technology	Function
Engine Layer	Linux, C project name : engine (main: engine.c)	Excitation propagation model implemented (goal: computation speed, sizing efficiency)
Transmission Layer	Linux, Java project name : alexLib (main: central.java)	Transmission of information between engine layer and presentation layer; engine launching (goal: network efficiency)
Presentation Layer	Linux, Java, JavaFX project Name : alexFX (main: alexFX.java)	User interface, working as a client to the central process. Access to passive jelly. Access to active jelly and engines. (goal: user friendliness, visibility on data and process)
Resource Builder	Linux, Java project name : alexRB (main: builder.java)	Construction of passive jelly database. Organized in numerous steps. (goal: data coherence, data availability)
Java Utilities	Linux, java, javaFX project names: mslib, techLib, structureLib, drawFXLib	Generic tools for advanced java application development. Usable not only with Assothink. (goal: ease of development and deployment)
Bash Scripts	Linux, bash main bash scripts: a	terminal-based front-end utilities to launch, monitor, test, debug... all other parts (goal: convenience)

Engine layer (engine project) notes

This layer implements the Assothink Excitation Propagation Model (EPM), or Assothink active Jelly, or Assothink brain.

The program is a C program: engine.c.

Only concepts, links, signals (only the active jelly).

Load concept and links data from the transmission layer.

Runs autonomously, saving state .

A brain uses a brain slot: a host/port combination.

Many brains may run at the same time in a network, and even on a single machine.

Interaction

The engine layer interacts mainly with the transmission layer.

It is also controlled through the bash scripts components.

Suggested hardware

• standard PC
• mainly accessible through network (remote shells)
• with Linux OS
  
• with strong processor (fast) and gigabit ethernet!
• 4 Gb RAM, USB port, small HD, simple graphic controller
• compact
• without (no need) CD/DVD, advanced graphic, monitor, keyboard, mouse...

The EPM engine machine setup is standardized, simple, fast,

Transmission layer (AlexLib project) notes

The transmission layer works as a daemon.

Its main class is central.java.

The transmission layer is a java program, launched from the a script.

Interaction

This layer has two sides.

Interacting with its presentation layer clients, it

• delivers (begin of user session) the whole data set describing the passive jelly
• continuously delivers information produced by the engines
• receives orders from clients, and transmits then to engines

Interacting with the engine layer, it

• creates, kills and controls engine processes
• delivers orders defined by clients
• receives state information from the engines, propagates them to all subscribing clients
  

Functions

The controller is able to:

• read, analyse, display (or distribute) state data generated by the EPM engine
• send orders to the EPM engine

Orders

The transmited orders are:

• Wake Up (launch a new instance of EPM engine)
• Halt (the EPM engine process is terminated)
• Stop (the EPM engine is stopped: it continues to run, but does not perform any cycle)
• Resume (the EPM engine resumes: it performs repeated cycles until new orders are sent)
• Clear EPM memory (all excitation values are set to 0)
• N cycle (the EPM engine performs N cycles, then it is stopped)
• Set one-shot excitation for some nodes
• Set permanent excitation for some nodes
• Modify EPM parameters
• Modify permeabilities of specified EPM links
  

Suggested hardware

Roughly the same as the computing layer.

More server oriented: more disk and network efficiency, less CPU needs, high availablity.

Presentation layer (AlexFX project) notes

This is a javaFX front-end client applic running

• as stand-alone application
• within browsers

Many clients (presentation layer) may run and interact with the transmission layer simultaneously.

The presentation layer client allows

• to list all running engines to select any of them
• to display the state of the selected EPM engine
• to send to the selected EPM engine all orders listed above.

Suggested hardware

No specific hardware needed.

A standard PC with browser having good web access and good visualization hardware.

• The screen resolution should by high (at least 1500x1000 points).
• The graphic card should have average or high performance.
• The preferred browsers are google chrome (chromium) and firefox.
• The web access should be fast (>20Kbit/sec for download) and the network throughput should be high.
  
• OS? Linux, Windows, Android and MacOS OK.

Resource Builder (AlexRB project) notes

The resource building part of Alex is a set of batch programs running sequentially.

They are described in details here.

Lauching Alex triple level components: what's running where?

We assume here that 'MMU' is bith ethe dataHost and the centralHost. This may be checked with the a info comman (see Bash Scripts notes)

A typical Alex system includes data located in /media/server.Alex.MMU/  (assuming MMU is the dataHost machine).

A typical running set includes:

• a ceds (unique script running on MMU, mainly reporting the creation of central process, engine process, and client process)
• central (unique ava program running on MMU)
• one or many engines (runing on several hosts)
• one or many UI clients (running on several hosts)

The active process may be listes using a ps or a psa. They are also visible in the UI interface.

The list of hosts likely to run the processes, known as hostSet is defined in java source mscp.alex.msg.hostSet .

Here is a typical output of a info (mar 2013)

Java utility libraries notes

The java utility libraries are used

1. by the transmission layer
2. by the presentation layer
3. by the resource builder

and generally speaking by all java programs included in the Assothink projec t (and possibly by other programs).

Project-library mslib

General purpose libraries, complementing the standard java packages.

Project-library techLib

Specialized technical tools.

• byting: byte encoding/decoding of various simple data types
• clock: nanosecond-level time measurement
• pack: data compression
• rounding: creation of rounded data for grafs
• sort: sorting utilities
• tObs & tSerie: time series
• topSort: partial sorting
• wordDistance: distance between langage words

Project-library structureLib

Advanced structures based mainly on Java collections, mainly matrix and indexing tools.

Project-library drawFXLib

JavaFX animation library, plus various javaFX utilitie.

There is also a set of simple graphic programs used to test drwFXLib, located in the animFXLab project.

Bash Scripts notes

The a script covers various maintenance and convenience functions.

It is located in ~/Scripts/a .

It uses function and settings defined in ~/Scripts/bashrc and ~/Scripts/bashrcAlex .

It covers very different functions, as visible in its help display (feb 2013)

Alex Main host is MMU Local host is MMM

Usage a ceds (launch monitoring CEDS; only on MMU)
Usage a l(og) (view CEDS logFile)
CEDS is central-exec-daemon-script

Usage a fx (javafx front end)

Usage a u(p) (up central)
Usage a h(alt) (halt central)

Usage a gcc (compile engine)
Usage a gcc <args> (compile engine, then execute with args)
Usage a engi <args> (run engine with args)

Usage a ps (involved processes on this host)
Usage a psa(ll) (involved processes on all hosts)
Usage a f(ollow) (tail -f central stdout)
Usage a e(rror) (tail -f central stderr)
Usage a t(est) (test central and CEDS)
Usage a c(heck ssh cap) (check ssh capabilities; only on MMU)
Usage a engp(erf) (engine performance)
Usage a ba(ckup) (create a backup of critical source files)

Usage a info (general working context info)
Usage a wiki (open Alex wiki)
Usage a h(elp) (this help)