1 lessons from the crisis: enterprise software lifecycle optimization the 2nd international...

1

Lessons from the Crisis: Lessons from the Crisis: Enterprise Software Enterprise Software

Lifecycle Optimization Lifecycle Optimization

The 2nd International Conference on Complexity, Informatics and Cybernetics (IMCIC'11)

March 27–30, 2011, Orlando, Florida, USA

Dr. Sergey V. Zykov, Ph.D.State University –

Higher School of Economics

2

Lessons from the Crisis: Enterprise Software Lifecycle OptimizationLessons from the Crisis: Enterprise Software Lifecycle Optimization

Heterogeneous ESE: a challenge, esp. under the crisis The methodology combines formal models (incl. modified spiral

model) and SDK for class-level association-based relationships Problem domain features: heavy data burden - in 2005 total data size of Intel Corp. exceeded

3.2 petabytes (over 120,000 employees in 57 countries) - high object classes complexity - incomplete information on the structure of certain instantiations

of the classes; - the set of class attributes and operations can be determined

rigorously. Reasons for methodology application (besides crisis): - variety of heterogeneous classes, - importance of association-based inter-class relationships - class inference possible even under certain % of weak-structured

class instancesThe 2nd International Conference on Complexity, Informatics and Cybernetics (IMCIC'11)


3


Software production crisis and SE era advent 1960-s: the software production crisis begins Problems/Disproportions: - anarchic SDL vs. growing system complexity - no SDL methodologies (except “build-and fix” approach) - the crisis is in our minds, not just in economics - an adequate (R&D-based, adaptive) SDL methodology required!

SE – the sci-&-tech discipline to overcome the crisis

SE development stages: - 1960-s: “hand-made” art – unique and precious masterpieces - 1970-s: “manufacturing” – mission-critical systems - 1990-s: “machine-tool conveyor” – CASE + team development



4


Comparing SDL and material production 1967 - The NATO Conference on SE Q.: Can SW be constructed/produced as a physical/material

object? Analysis: - step-by-step elaboration - maintenance (bridges may cost $0 for years, not SW) - “intellectual degeneration” (complex SW platforms change fast) - prototype reliability - “brute force” / “bulletproof “ approach (a bridge twice thicker) - residual faults (NASA flight simulation) Conclusions: - SW production is similar to material one in certain aspects - SW production is entirely different from material one in others A.: NO, Lifecycles are fundamentally different



5


Optimizing lifecycle: the answer to the crisis Every SSDL stage can be optimized (analysis, req.spec., design,

implementation, maintenance, etc.) Documenting should be optimized as well Optimization basis = models + methods + tools integration: - discrete metrics (residual faults, KLOC, etc.) + heuristics - practical applicability vs. mathematically “single-best” solution SSDL model basic features: - iterative nature - sequential elaboration - incremental development - risk analysis - modified incremental/spiral model - best applicable for enterprise-level projects (CASE + risks)



6


Heterogeneous ESE: a challenge, esp. under the crisis The methodology combines formal models (incl. modified spiral

model) and SDK for class-level association-based relationships Problem domain features: heavy data burden - in 2005 total data size of Intel Corp. exceeded

3.2 petabytes (over 120,000 employees in 57 countries) - high object classes complexity - incomplete information on the structure of certain instantiations

of the classes; - the set of class attributes and operations can be determined

rigorously. Reasons for methodology application (besides crisis): - variety of heterogeneous classes, - importance of association-based inter-class relationships - class inference possible even under certain % of weak-structured

class instancesThe 2nd International Conference on Complexity, Informatics and Cybernetics (IMCIC'11)


7


The methodology vs. ontology-based approaches (OBA):

OBA (e.g. Cyc) efficiency is comparable only under a total class-level uncertainty, which is a different problem domain than ECM

Thesaurus needed for the OBA to meet the relevance required

The methodology uses similar foundations and tools as OBA (UML and XML-based tools, predicate calculus-based CycL, “conceptual model” etc.) for data modelling and integration

OBA lack a balanced combination of formal models and industry-level SDKs (incl. visualization) for ECM lifecycle, resulting in low scalability and non-suitability for the major enterprise-level tasks



8

Objective, tasks, theoretical background

Objective: to make a software development methodology, which supports entire lifecycle of the enterprise software in the global computational environment

Tasks: - formalizing stages and levels of the methodology; - mathematical modeling; - creating CASE- and RAD-tools ; - implementing the methodology (prototype, full-scale).

Background: finite sequence, category, computation (D.Scott), semantic networks.




9

Innovations – the integrate methodology includes :

1. a set of data models for problem domain objects and for computational environment (CM, AMCM);

2. algorithm of the new component integration into the software implemented;

3. personalization procedure for enterprise content access;

4. SDKs: ConceptModeller, Content Management System




10

EDW problems solved by the methodology




11

Enterprise software lifecycle support by the methodology




12

Problem domain modeling

Data object modeling: “class object value” Class – collection of data objects of the integrated problem domain; Object – class instantiation by CMS template (metadata partial evaluation); Value – static HTML page generated by CMS (full evaluation).

Benefits: - evolves from the object-oriented approach; - develops the existing models ([V.E.Wolfengagen’s CM] et al.) in relation to global computational environment




13

Modeling classes of data objects

Classes C of problem domain data objects are modeled by domains: C = Iw:[D] v:D (w(v) ) = {v:D|}, where: 1) C and D are in a partial order relation (C ISA D); 2) is a criterion of data object w belonging to class C from the viewpoint of a problem domain expert.

Class of “n-dimensional” data objects is modeled by an n-arity relation:Rn = Iw: [V1,..,Vn] v1:V1 … vn:Vn (w [v1,…,vn] ) = {[ v1:V1 , …, vn:Vn ] | }, where: – “n-dimensional” criterion of data object w belonging to class Rn

Class is a collection of ordered pairs (vi,Vi), where vi is its i-th attribute (either of data or of metadata); Vi – attribute type.




14

From problem domain to computational environment (1)

Under class C instantiation with assignment a1 and template k of CMS HTML page, evaluation of the template collection M sets into “true” value its element mi, which index (k) equals the template number:

M = (m1,…, mk,…, mN), i=1,…,N mi{0,1};

[M|k] = (m1*,…, mk*,…, mN*), где mi*=1, i=k и mi*=0, ik.

Certain attributes of class metadata v1,…,vn are evaluated according to ti conditions of template:

[(v1:V1,…,vn:Vn)]ti= ([v1]|(t1),…, [vn]|(tn)) = (v1’:V1’,…,vn’:Vn’),

причем V1’ ISA V1,…, Vn’ ISA Vn.




15

From problem domain to computational environment (2)

The second assignment a2 instatiates non-evaluated

template elements (v1’,…,vn’) of CMS HTML-page

by content values (c1,…, cn) :

[(v1’:V1’,…,vn’:Vn’)]c = (v1’/c1,…, vn’/cn),

where c1:С1,…, cn:Сn,

and C1 ISA V1’,…, Cn ISA Vn’.

Сi class template is Ti = (i,(t1,…,tn)),

where (t1,…,tn) is the vector of the evaluated class metadata




a1

Value (portal )

Class (UML)

-Name : char-ColorDepth : int-Resolution : int-ID : long-Width : int-Height : int-TemplMask : long double

Photo

Object (CMS )

a2

16

Formal syntax of the CMS abstract machine

Let us collect all the CMS abstract machine language identifiers into Ide domain, commands – into Com domain, and expressions – into Exp domain:

Ide ={I | I – identifier}; Com ={C | C – command}; Exp ={E | E – expression}.




17

Formal semantics of the CMS abstract machine (1)Order of construction:

– standard domains (most often used);– finite domains (including explicitly enumerable elements);– domain constructors – operations of building new domains out of

existing ones;– composite domain formalization based on standard domains and

domain constructors.

Domain constructors :

- functional space: [D1D2];

- Cartesian product: [D1D2…Dn]; - sequence: D*;

- disjunctive sum:[D1+ D2+… +Dn].




18

Formal semantics of the CMS abstract machine (2)

State = Memory Input Output; Memory = Ide [Value + {unbound}];

Input = Value*; Output = Value*; Value = Type1 + Type2 + …

Constant denotate: <variable, value>

Identifier denotate: <variable_in_memory, identifier, state>




19

Formal semantics of the CMS abstract machine (3)

Semantic function for expression:E: Exp [ State [[Value State] + {error}]];

Semantic function for command:С : Com[State[State+{error}]].

Semantic statement for identifier:E [I] s = (m, I = unbound) error, (m, I, s).

Semantic statement for assignment command:C [I=E] = E [E] * v (m , i, o) . (m [v/I], i, o).




20

Bi-directional software development

in ConceptModeller CASE-toolkit


FO

RM

AL

IZA

TIO

NNATURAL

LANGUAGE

Business situations

in terms of

natural language

FORMAL LANGUAGEOF FRAME

DESCRIPTION(XML)

Visual frame

representation

C#.NET

FRAMETRANSLATION

INTO UML DIAGRAMS

Mapping

function from

frames to UML

diagrams

C#.NET

FORMAL LANGUAGEOF DIAGRAMS

(XML / RATIONAL)

UML diagrams

visualization

C#.NET

FORMAL LANGUAGEOF DIAGRAMS

(IBM RATIONAL / MS VISIO)

IBM RATIONAL,ORACLE DEVELOPER,

MS VISUAL STUDIO



VIS

UA

LIZ

AT

ION

TR

AN

SL

AT

ION

VIS

UA

LIZ

AT

ION

Business situations

in terms of

UML diagrams

21


Software Solution Arcitecture

в

Oracle InterOffice

Oracle Appls

Media

ArchiveOracleDBMS Cluster

WWW server

SD

Catalyst8500

Power Supply 0CISCO YSTEMSS Power Supply 1

SwitchProcessor

SERIES

Content DB

Notebook

Mobile phone

PocketPC /PDA

Home desktop PC

RS CS TR RD TD CDTALK / DATA

TALK

Enterprise LAN

Oracle Portal/Intranet server

Content DB



22


CMS logical structure

Keywords

IM

ImagesReferences, keywords

Image parameters

Special section parameters,

Message status

KeywordsModulePress Release

Module

Media News

NCNews Columns

Module

MenuModulePages

CMConfiguration management

ModuleEvents

ModuleSpeeches

SSСпециальные разделы

ModuleSynchronize

ModulePublish Cycle

ADAdministration

Menu parameters

Keywords

Keywords

Page status, publishing cycle

params

Params and states for news pages

Page status, publishing cycle params

References, keywords

Menu items paramsMedia news parameters

(menu items, etc.)

Keywords

Image state

parameters

Special section parameters

(menu items etc.)



Params and states for menus and messages



23


Structure of the integrated enterprise program system

Deductions and Bonuses

Module

VacanciesModule

LeavesModule

Personnel Movement

ModuleAssets

Responsibility

Module

Appraisal and Testing

Module

Personal Data

Module

TrainingsModule

PR

Payroll Information System

FAFixed Assets Management

Information System

AP

Accounts Payable IS

DC

Document Control Information System

AR

Accounts Receivable IS

Deductions, payments

Address book , organization

structure

Asset Features

Параметры аттестации

Appraisal params

Appraisal reports

Параметры выплат

Responsibility Reports

Assets Refixed

Leave params Testing

params

NecessaryTrainings

Appraisal params

Appraisal params

Appraisal params



24


Comparing the software development methodology

to the commercial methodologies available



Methodology Mathematical model

Integrated methodology

25


Implementation features comparison



Software

ITERA CMS

Mu

lti-

lan

gu

age

pu

blic

atio

n

Java

ser

vlet

s

.NE

T w

eb s

ervi

ces

UM

L d

iag

ram

s

WY

SIW

YG

-map

pin

g

Inte

gra

ted

ER

P

info

rmat

ion

sys

tem

re

po

rts

Inte

gra

tio

n w

ith

leg

acy

info

rmat

ion

sys

tem

s

Sm

art,

tem

pla

te-

bas

ed d

esig

n

Co

mp

lex

dat

a o

bje

ct

emb

edd

ing

26


TCO comparison results



/person

CMS + ConceptModeller

27


ROI comparison results



CMS + ConceptModeller

, yrs

28


Implementation terms comparison results



CMS +

Conce

ptM

odell

er

Optimistic scenario

Pessimistic scenario

29

Theoretical results:

1) A system of formal models for problem domain and computational environment (rigorous semantics, entire lifecycle support, content management orientation);

2) Algorithm of integrating new components to the enterprise software system (problem-oriented, heterogeneous software architecture support);

3) personalization procedure for accessing enterprise content

(flexible, reliable)




30

Engineering results:

1) CASE- and RAD-toolkits: a) ConceptModeller (rigorous semantics; compatible to up-to-

date CASE-tools, ERP and и legacy systems; re-engineering; XML/BPR/UML standard support);

b) ITERA CMS (rigorous semantics; rapid publishing of complex content; WYSIWYG interface; office products integration).

2) Architecture (environment unification of heterogeneous enterprise

applications; role personalization with situation dynamics)




31

Practical value of the results obtained:

1) implementation term-and-cost reduction (TCO, ROI) as compared to commercially available

software by 30% (average);

2) major enterprise software features improvement:- scalability; - reliability; - ergonomics.




32


Research results approbation:

Over 30 presentations on international conferences,4 books and over 50 papers

Research grants: MSR (2002-03), RFBR (1996-2006) and HSE (2008-11).

ITERA implementation (150 companies, 10,000 employees): CMS (2002); Internet-portal (2003); Intranet-portal (2004)

Other implementations: ICP (RAS), Sterkh Foundation, Ashihara Karate Association, Russian Orthodox Church, etc.

Curricula (HSE, MEPhI, MSUFI, INTUIT, LANIT, SoftLine, TEKAMA, CareerLab) – over 3000 graduates



1 lessons from the crisis: enterprise software lifecycle optimization the 2nd international...

Documents

crisis se development

international conference

usa slide

material production

cybernetics imcic11

optimization basis

nato conference

growing system complexity