help-desk systems stephen lee-urban november 17, 2006
TRANSCRIPT
Help-Desk Systems
Stephen Lee-Urban
November 17, 2006
2
References
• Experience Management: Foundations, Development Methodology, and Internet-Based Applications, Ralph Bergmann.– Chapter 9: Developing and Maintaining
Experience Management Applications– Chapter 11: Experience Management for Self-
Service and Help-Desk Support
3
Outline
• Motivation / Goals / Background
• The HOMER architecture
• Developing & Managing EM* Applications
• Evaluation of HOMER
*EM = Experience Management
4
Motivation
• Complexity of technology increasing– Operation, Maintenance, Repair– Probability of failure grows exponentially with
complexity– Expertise in all features unreasonable
• Problem solving experience distributed– Experience overlap is variable
Experience Management System (EMS)
5
Goals of EMS
• Create knowledge repository– Store problem solving experience– Simplify trouble-shooting complex technical
domain– Domain is likely to change over time
• Use knowledge repository– By varying levels of expertise– In time-critical situations
6
Experience Management vs CBR
Experience Management
CBR
(Organization)
(IDSS)2. Reuse3. Revise
4. Retain
Case Library
1. RetrieveBackground Knowledge
Experience base
Reuse-related
knowledge
Problem acquisition
Experience evaluation and retrieval
Experience adaptation
Experience presentation
Complex problem solving
Developm
ent and M
anagement
Methodologies
BO
OK
Slide: Dr. Munoz-Avila
7
Help Desk Support Levels
• Key-user
• Specific technician
• Vendor
• Help-Desk sys.
• Bottlenecks at each level. – Problems recur.– Frustration for all parties, and wasteful of resources!
• “Managerial” Aids to Help-Desk: – Inventory systems, trouble-ticket tools, call-tracking software– Don’t support diagnosis nor serve as knowledge repository
http://www.simpsonstrivia.com.ar/simpsons-photos/wallpapers/homer-simpson-wallpaper-brain-1024.jpg
CBR
8
Outline
• Motivation / Goals / Background
• The HOMER architecture
• Developing & Managing EM* Applications
• Evaluation of HOMER
9
The HOMER System
• German: “HOtline Mit ERfahrung”• Developed as part of INRECA-II project• For CAD/CAM help-desk at
DaimlerChrysler in Sindelfingen• Generic experience management
architecture & set of related tools– Stores experience in CB for access, reuse
and extension– Drastically decreases problem solving time
http://www.aeria.phil.uni-erlangen.de/photo_html/portraet/griechisch/dichter/homer/homer3.JPGhttp://gattaca.com.ar/weblog/wp-content/homer.gif
10
Structure and Representation• Determines experience management approach• HOMER uses an object oriented approach to model experience. • Advantages of OO approach:
– Structure of system to diagnose is representable in detail– Symptoms (attr.) easily related to owning object– Semantics of prob. description can be captured and used for selecting
appropriate prior exp.– High retrieval accuracy can be achieved– Particularly suited for diagnosis help-desks w/ complex equipment
Can show many hard to diagnose faults– Can be used to help guide help-desk operator while describing and
entering cases– Better domain model = easier maintenance and use
• Disadvantages of OO approach: – Harder to create model– Additional effort in beginning of knowledge acquisition
Representation should be discussed with system admin. Shallow?
11
Case Structure
• Modeled according to approach used by help-desk operators to solve problems
• Feasible for most complex help-desks
• Topic: problem area (e.g. hardware)• Subject: failing object (e.g. printer)• Behavior: way (miss-) behaves (e.g crashes)
• Minimum set of attribute-value pairs describing symptoms necessary to fault diagnosis.• In OO, all attributes are related to certain object.
• Fault: problem cause• Remedy: problem fix• Represented as (hyper-) text
Case: Complete path from prob. to solution
12
Experience Base Partitioning
• Domain size & complexity + demanded accuracy & consistency = partitioning
• Two kinds of cases:– Approved: experience reviewed by top-level experts.
“Best Practice” for problem solving– Open: recently captured, pending validation
(correctness, completeness, & utility)
• Case base separated into:– Case Buffer – all open cases– Main Case Base – all approved cases
All available to help-desk operators
13
HOMER Architecture
• Client-Server– Shared domain model (DM) + case base (CB)– Eases maintenance of DM + CB– Server accessible via intranet / internet– “Fat” client reduces network traffic
• Object-Oriented experience model– Not “shallow” users are experienced– For non-trivial problems
14
• Lowest access rights• Daily user• Retrieval & acquisition
• Middle rights• Case maint. & case approval• Redundancy & consistency checks• Buffer to base• May modify value ranges of attributes
• Highest rights• Creates & maint. domain and case model• +/- attr & concepts• Admin users and rights
15
The Server
• CBR-Works server stores model & CB,
• CBR-Works modeling tools provided by server for domain modeling, case & model maintenance, & initial case acquisition
• Domain model snapshot…
16
Hierarchical domain
concepts
Help Desk Case
Attributes
• Problem, holds failure• Situation, holds symptoms - Hierarchical, sub-concepts - Structure helps maint. & retrieval• Loesung, holds solution• Administrativa, holds organizational & statistical information
17
The Client
• Interface to server for case retrieval, case acquisition, & case browsing
• Written in Java• “Hotline” component for help-desk operator
– Only shows relevant information– Assists via two modes: user/system driven
• “Case Browser” component for experience author– For access to CB and case buffer– Revision, extension, approval of “open” cases– Removal of outdated cases
18
Hotline Component (HC)
• Used mainly during hotline operations
• Supports help-desk operator during problem-solving process
• Four main modes of execution– Problem description (initialization/refinement)– Situation description (user/system driven)– Solution retrieval (manual/automatic)– Retain w/ feedback
questions
answers to questions
solutions (cases)
problems
19
HC: Problem Description
• Gives initial info. on user’s problem
• Answer following questions consecutively:– Problem’s topic? (e.g. output, software, …)– Topic’s subject? (e.g. output: plotter or printer)– Behavior? (e.g. “no output at all”)
problems
question (free text)
values
value range units
20
HC: Situation Description
• Problem determines relevant case structure (i.e. situation template)– Contains possible symptoms for prob.– Each symptom associated w/ question
• Symptom attribute values complex/simple• Two modes of operation
– User-driven: (no guidance, direct entry)– System-driven:
• Shows sorted view of most relevant questions • Based on attributes with highest info gain
21
HC: Solution Retrieval
• Experience retrieved based on problem and situation descriptions
• Manual (batch) / automatic (per question)
• Each row a solution, sorted by relevance (similarity)
• Solutions viewed (read-only)
22
HC: Retain/Feedback
• Retain case when it – Has new experience– Requires case model extension (e.g. new
value to type)
• Case entry interface allows– Final modifications (e.g. finish case descr.)– Annotating why op. thinks case should be
retained
23
Case Browser Component
• Used by experience author to manage CB
• Works on both approved and open cases
• Can perform following operations:– Case creation– Case copy– Case deletion– Case approval
Problem, situation, solution, administrativa
question (free text)
values
value range units
24
Outline
• Motivation / Goals / Background
• The HOMER architecture
• Developing & Managing EM* Applications
• Evaluation of HOMER
25
Developing & Maintaining EM Applications
• IT companies require efficient, effective application development– Guidelines/Methods of implementing apps– Also seek to preserve past project experience
• INRECA methodology– Targeted at industrial EM applications– Developed in the INCRECA-II European
ESPRIT project (1996-9)
26
EM Model
Knowledge Kernel
Problem Solving Cycle
• Knowledge Acquisition & Knowledge Maintenance• Technical, organizational & managerial aspects
27
Three Process Types
• Technical– Describe development of system & required documentation– E.g. requirements analysis, system design, implementation and
testing
• Organizational– Address parts of business process in which software will be
embedded– E.g. training end users, archiving request records, updating &
maintaining the help-desk system
• Managerial– Provides environment and services for developing software that
meets product requirements and project goals– E.g. project planning, monitoring, quality assurance
28
Methodologies
• Makes development an engineering activity, rather than an art
• Use of methodology provides benefits:– Productivity, Quality, Communication, Management
decision making• INRECA methodology based on software
engineering principles, covering:– Project management (cost assessment, schedules,
project plans, etc.)– Product/Deliverable specification– Product development and maintenance– Analysis/Organization of target env. for CBR system
29
INRECA
• Basic philosophy: experience based construction of experience management applications
• Self-application of principles of experience reuse
• Experience modeled as structured text documents, hyper-text linked.– Origin: Software process modeling
30
Software Process Modeling
• Well defined terminology to describe the engineering of a product– Process, “what”: basic step to carry out, transforming input into
output. Defined by a goal, a set of alternative methods, input/output/modified products, & required resources (agent/tool)
– Method (simple/complex) “how”: Detailed specification of a way to achieve a process’ goal
– Product: goal of the process
INRECA notation
31
INRECA Process Models
• Make explicit all processes, products, methods, resources and interactions
• Diagrams insufficient Each element must be described in detail– Called a process model– Solid basis for project planning (e.g. effort
calculable based on processes involved)
• General vs. Specific descriptions– Common Generic/Cookbook vs. Project level
32
The INRECA Experience Base• Collection of processes, products, & methods common across EM apps.
• Processes, products, & methods tailored to class of applications (e.g. help-desk). • Each class has recipe• Recipe: has process models describing how app. of that class be developed & maintained
• Describes experience of an instance of a completed project• Project-specific info. (e.g. processes carried out, effort of processes, etc.)
33
The Common Generic Level (CGL)
• Overview of top-level processes and products in Bergmann, chapter 9– General enough to occur in many projects– Problem statement “vision document”
goal checklist feasibility study detailed analysis of organization project plan software development evaluation
– Meanwhile, experience base can be consulted for all of the above processes
• Consult book for thorough exposition
34
Go / No-go(analysis &elicitation)
Organizational(planning)
Implementation
Maintenance
35
CGL: The Three Processes
• Managing an AI / EM project differs from other IT projects– Concepts & technologies altogether new– Emphasize early awareness training– Ensure user-participation in iterative prototyping
process
• Technical processes very typical of software development projects
• Organizational includes identifying perceived problems and opportunities at the human & organizational levels
36
Documenting in INRECA
• Processes, products, & methods documented and stored in experience base
• “Sheets” structured page w/ all relevant information in a predefined format– Standardize documentation– Created as web pages for easy access & use– CGL has more than 150 linked sheets– 8 type of description sheets:
[ generic | specific ] [process | product | simple method | complex method]
– Contains references to respective input, output, and modified products of the process
37
Process Description Sheets
• Recipe/Project Name• Process Name• Process Goal• Input/Output/Modified Products• Set of Applicable Methods (names)• Agents (e.g. personnel) • Required Tools• Administrative Information (e.g. sheet author,
version, last modified)
38
Product Description Sheets
• Module/Project Name
• Product Name (e.g. “requirements doc”)
• Product Description
• Administrative Information
39
Simple Method Description Sheets
• Module/Project Name
• Method Name
• Method Description
• Administrative Information
40
Complex Method Description Sheets
• Module/Project Name
• Method Name
• Method Description
• Details– Links to a product flow description which
contains the relevant sub-processes (byproducts)
• Administrative Information
41
The INRECA Reuse Procedure
• Recipes at cookbook level are most useful for building a new application
• Even projects not fitting a recipe can use processes described in CGL
42
Don’t Forget to Remember!
• After completing project, include it in the experience base continuous improvement of the EM software development process
43
Tool Support for INRECA
• Experience modeling methodology tool implemented in Visio®– Natural choice: shapes, hierarchical, HTML, VBA, database
access
• Knowledge modeling tools– Integrated in the CBR-Works tool (tec:inno GmbH 1999)– CBR-Works Concept & Type Hierarchy Editor for vocabulary
modeling – Similarity modeling tools integrated in the Concept & Type
editors
• Also has an editor for adaptation and completion rules
44
Outline
• Motivation / Goals / Background
• The HOMER architecture
• Developing & Managing EM* Applications
• Evaluation of HOMER
45
Evaluation of HOMER
• Evaluation performed by INCRECA-II project partners to identify benefits of EMS
• Incoming calls monitored. Help-desk operator 1st solves conventionally, then with HOMER– No solution new case created– Two month test period: 102 calls; 45 unsuitable for
HOMER
46
Evaluation (II)
• Of the 57 (/102) suitable problems– 18 solved (i.e. 32%)
• Ave. resolution time w/out HOMER: 141 min.• Ave. resolution time w/ HOMER: 9 min.• Correct result a limited bias HOMER sys. Mode
• HOMER + CB transferred to a new site– Operators have no experience with the process chain
so the cases are of great value– Initial knowledge acquisition gave operators insight
into domain.
47
Evaluation (III)
• Methodology recipe created and used– Impact: productivity, quality, communication, &
management decision making– Customer and developer both benefited
• Creation of project definition from scratch– Took three months– New development team reused recipe to define three
new projects, each in < one week (12x speedup!)• Were sure all relevant aspects taken into account• Basic recipe available developers focus on domain
peculiarities quality/detail of descriptions greatly enhanced
48
Evaluation (IV)
• Development & testing of 1st prototype– About 6 months– Subsequent efforts: 2 weeks (13x speedup)– Less qualified personnel w/out sacrifice of quality
• Useful as training tool for maintenance & use• Message: development of EM system a science,
not art– Each process describable– Validity of approach defensible– Realistic expectations of effort by whom and when– Measurable project process
49
Thank You!
(Questions?) (Comments?)