..index..

2/
everythingwireless/
infosys/
infosys2formalmethods/
infosys3softeng/
infosys4oo/
infosys5thesis/
is/
jav/
jonathan/
jonstoyota85/
men/
mmp/
nzph/
pics/
projects/
Publish/
sap/
saudi/
ske/
snd/
sto/
vyvien/
wit/

dlv:
2003-09-22 Software Project Management - Lessons Learned
2003-09-16 Wireless Software Engineering

log:

Grid Computing

Cardiff University: http://www.cf.ac.uk/
Omar Rana : http://www.cs.cf.ac.uk/User/O.F.Rana/
Oracle 10 Grid enabled Grid Economic Unit
"Semantic Grids"
UDDI - RDF
ontology repositories - FIPA
Can I belong to 2 grids? Yes, you can belong to as many grids as you like.
A grid is like an institution (Colombetti)
grid must be a government lab - linux cluster

Coloured Petri Nets
Petrinets for workflow
Fraunhofer Resource Grid - globally distributed hardware and software resources
http://www.fhrg.fhg.de/index_en.html

type checking
consistency checking

ontology UML
sylvian villani
meta-ontology RDF REST http cranefield
(PUT POST GET) - computer ... (applet) - human
"lightweight" - repository

ontology for SKEN
Grid computing - repository - agents - FIPA - REST architecture - ontology architecture
OKBC at the backend - otago - UML for Ontology Development
"everyone knows http" - more users of http than fipa

FIPA SL0 and SL1 - specifying constraints - like first order logic
KQML - Belief State Representation in agents in FIPA

Realism and Instrumentalism

Realism - Belief State Representation required for messages
Instrumentalism - Messages do not require internal belief states

Research Areas:
service discovery - ontology

MathML - a language aim at a particular domain - what do we do for a more generalised language?

Trust and reputation? - how do we get a trust model in a grid? - something like ebay, for example.

What properties do we say that a service has?

1. Functional properties
2. non-functional properties ... eg. trust and reputatio...

example, what is result of the 8 point decimal service turns out only to deliver the 4 point decimal service?

old example - using corba wrappers, request brokers, in order to manage a GIS system, also there is the agent paradigm

example of a matrix transfer, not possible in the standard CORBA system

Scientific Computing is not possible in Java because it cannot match the performance of Fortran, for example Airbus still using fortran for computation

Web services examples, like UDDI SOAP, WSDL are planned to be used by grid computing, however the specification available is very weak froma scientific point of view, a DUNNS Dun and Bradstreet Category Bag is not enough to finely separate mathematics operations

Should we use Corba? We can try web services, but the performance will probably be worse. WS gives interoperability, but the penalty is probably performance

One solution is to use agents, but then to move away from them for large transfers of data, for example, we can use agents for the discovery and the linkages and then when we have a high bandwidth application, we send our references "out of band"

Grid community requires expressiveness in services, but also ease of use, this is apart from the basic performance requirement

Jini- leases - future services

Triana - P2P JXTA FIPA ACL no general advertisement concept - stripped down JXTA - single destination

Peer to Peer online Wireless PDAs - TUM Arena - Infrastructure - wireless computers - apple powerbooks - wireless computting

Java - memory managment ian taylor - PS=small

jxta used in 2 ways:
1 - using the jxta protocols
2 - java implementation of jxta

the push for new research leads to a multiplicity of standards
- similar to the software engineering provision - a multiplicity of methods are used in order to advertise different methods and implmentations

24 people in the GRID centre in Cardiff

Science and Computer ENgineering Department

British Aerospace - Bristol

Representation of Ontologies in UML

Agent Game Theory Strategies

tit-4-tat

possible 'roles' for a pitgame agent:

- Selfish agent: Tries to find out what other agent needs, and tries NOT to trade that cards to them.
- Conventional/blinkered agent: Tries to stick with the corner he wanted to go for in the beginning.
- Fetish agent: Loves one (or more) kind(s) of commodity, always wants a corner on this.
- Friend agent: Tries to find a buddy, will give his buddy the cards his buddy needs.
- Kamize agent: Tries to prevent all other agents from getting the right cards, at his own cost.
- Sacrifice agent: Descendant of the friend agent, will give his friend cards he needs, at his own cost.
- Hero agent: Will be helped by buddy and sacrifice agents.

GRANULARITY, THREADS - Willem

granularity of computer science

SAP LUW - LOGICAL UNIT OF WORK

... http://sap.mis.cmich.edu/sap-abap/abap09/sld006.htm

Funding

FP6

Building a lightweight distributed objects system based on Java

Status: In progress

Issues:

Risks:
1. Tight schedule
2. Development and Test path is not ideal
3. Test cases not fully mapped out

Scenarios
1. Peer 1 starts game, peer 2 starts game, peer 3 starts game - all in game
2. Peer 1 starts game, peer 2 starts game, peer 1 leaves game, peer 3 starts game - all in same game

Issue object ownership for lost peers

Peer 1 Owner
Peer 2 Observer

Peer 2 loses communication with Peer 1

What does Peer 2 do?

Issue 1 - Category of Game

Category A of Game: Communication must be maintained
Category B of Game: Communication can be dropped

Issue 2 - Category B Threshold

Category B:
- communication loss - need to determine a threshold when we consider a communication

Peer "Lost"

If threshold met, and agreed to by other pstarties that the peer is lost, then the objects need to be reststructured

Books

Toby Young

Rhetoric

is the counterpart of Dialectic.

The Billion Dollar Question

- questions to improve the internal functioning of mega-corporations - global multinational bodies with heavy financial muscle and consequent political clout
Estimates of the benefit of the improvement of internal processing:
Global Spend on IT within corporations
estimate 2003 - top 1000 companies - .... IT department spend 20 million dollars x 1000 = 20 billion dollars. 50% improvement would save 10 billion dollars, which at a long term interest rate of 7%, means a net present value of 140 billion dollars.
So a single change in processing to enable such a productivity improvment would be worth the value of a company like General Motors for example.
For US companies, the top 10 companies have a total capitalisation (capitalization) of around 2,200 billion dollars. For The top 100 companies the total cap is about 8,300 billion dollars.
Assuming that IT spend represents 5% of that total, then total IT spend is 415 billion dollars for the top 100 companies.
A 50% cut in IT costs at 415 billion dollars is 207.5 million dollars, which assuming our 7% return, means an NPV of 2,964 billion, more than the top capitalisation of the top 10 US companies.
So this is the trillion dollar question ... what methods can be implemented to attain this productivity improvement. It is little known how critical this issue is to the financial health of all countries.

Semantic Reuse, ebXML, SAP, UBL, Patterns in ebXML, Reification in SAP

Abstraction

3 types: programming language abstraction, like C Perl AWK Java ABAP COBOL BASIC
Application Software Abstraction
Other Abstractions

TIME AND COST: EFFORT / REUSE / REPETITION

TRIANGLE - QUALITY, TIME, COST

Using JPEGs to describe the world: 80 pixels square - 6400 pixels - 256 colours - 1,638,400 million variations enough to describe all known concepts in the Oxford English Dictionary

NP Completeness - Garey and John - Problems being impossible

Company Process - Law - CPE - LPE - MBA - IP - Philosophy - SENG
Company Process - Knowledge Management - IP - Philosophy
Company Process - Knowledge Managment - Process Management - Operations Management - SENG - MBA
Company Process - Technical - Economic Input Model
Company Process - Technical - Computer Technology - Logic - Optimisation

Transformation

Process Transformation HASKELL and PERL

Function (x) F(x) x
IN and OUT

Iterative Process

Exponential Growth

Types of growth

Population growth charts - S shape to fit resources

http://www.silicon.com/comment/petercochrane/
The economist Adam Smith was wrong. He was just too linear in his thinking. So says Peter Cochrane in his latest column, where he explodes the myths surrounding exponential growth... There are two commonly misunderstood expressions used by the media and politicians that mildly amuse me for their inaccuracy. The first is quantum leap - which is, in fact, an infinitesimally small change. The second is exponential growth.
For the past three years I have been asking audiences of business leaders, planners, educators and politicians if they understand what an exponential function is? In an audience of 500 people I generally get less than 10 people put up their hands and the rest admit that they do not really understand. Surprisingly, when I move on to explain that an exponential function is exactly like compound interest (what you pay on a loan or overdraft) I find that the vast majority still don't understand. They have a mortgage and a bank account but they don't really don't understand compound interest and therefore, exponential functions.
The function ex (which = EXP(x) in computer speak) is so delightfully simple and so very deceptive! Here is a simple explanation. Suppose you to work for me on the following basis: on day one I pay you $2, on day two I pay you $4, on day three $8, day four $16 and so on. How much will I pay you on the tenth day? The answer is $1,024. This is counter intuitive outcome, even more so on the twentieth day I will pay you slightly over $1m and the thirtieth day slightly over $1bn. This is exponential growth.
Let's look at it another way - if you were to invest $1 at 10 per cent interest compounded for 10 years, you would receive $2.59, but if it were $1 for ten years at 100 percent, then you receive $1,024.
An old conundrum says a king is asked to pay for work by placing a grain of wheat on a square of a chessboard and then on day two two grains, on the third four grains and so on. By the last square of the chess board the grains of wheat more than fill a throne room. This, by the way, is how a viral infection rapidly becomes an epidemic - sneezing or skin contact is an exponential spreading mechanism! And so is a computer virus that can now span the planet in much less than a day via the internet.
Probably the most frightening exponential experiences would be to sit on a beach and notice a wave on the horizon but by the time you have realise it is a tsunami (a tidal wave) it is too late to run. You will be swallowed up by the advancing wave and you will die.
Similarly, driving a car at great speed sees our perception framework distorted and we become unbelievably tolerant and confident on an open road. Only when we pull onto the off ramp do we suddenly perceive we are travelling extremely fast.
Technology is the same. It may appear to be insignificant and on the horizon but by the time it is perceived as a threat it is too late. Moore's Law is the most celebrated exponential law in the IT industry. It accurately predicted that integrated circuit density, and hence computer power, would double in power every 12-18 months from 1960 onwards. But there are other, similar laws for optical fibre bandwidth, network capacity and so on. In fact it is difficult to find anything in IT that isn't governed by exponential growth.
Adam Smith was wrong in his own time and he is even more in error today. In his economic model of the universe there is a finite source of material with limited production, routes to market, finance and communication. This led to the linear channel model of economics with a finite population of limited appetite, expectation and money. This was a model that worked well when the world was a slow moving place but not today,
It is now apparent that in the new economy the source of raw materials is unlimited in term of bits, production, routes to market, finance and communication, and there is no limit to what customers will purchase and use or expect and communicate. All are now linked by highly non-linear channels involving fixed and mobile networks, computers and people networks - eyes and ears.
How different to 100 years ago - and how dangerous when planners, politicians, decision makers and leaders do not understand the most fundamental of functions that now governs the growth of trade, communication, wealth and risk.

State Explosion Problem

Luckham

Event Patterns and Event Management - Formal Methods

BULLSHIT
Building UML Lightweight Agile Lifecycle Systems Hypertext Interfacing Technology

Try my new BULLSHIT Process. With Bullshit you get:
- Interoperability Optimisation
- Insubstantiation Optimisation
- Inter-overobjectification Optimisation
-

Scott Ambler

UP - RUP

UP - EUP

Software Nonsense

esperanto = ebxml?

ADL = Architecture Definition Language

Event ADL - JEFFREY L. VAGLE CMU BBN

regex = regular expression

Probes - Gauges