A User -Perceived Availability Evaluation of a Web-based Travel Agency Mohamed Kaâniche, Karama Kanoun, Magnos Martinello Partially supported by the European Community, DSoS - Project IST-1999-11585 DSN-2003, IPDS, 22-25 June 2003, San Francisco, CA, USA
Context and motivations External Assess the impact of failures on Supplier 1 Suppliers the quality of service provided to the users User 2 Supplier 2 User 1 Support e-business system Supplier n designers in making objective architectural decisions Internet Hierarchical modeling framework e-Business service provider Case study: Travel agency e-Business key players 2
Context and motivations External Flight reservation Flight reservation Assess the impact of failures on Flight reservation Suppliers system system system the quality of service provided to the users AF, KLM, BA, … Hotel reservation Hotel reservation Hotel reservation system system system Support e-business system Sheraton, Sofitel … User 2 designers in making objective Hotel reservation Hotel reservation Car reservation architectural decisions system User 1 system system Hertz, Visa … Internet Hierarchical modeling framework Case study: Travel agency Web-based Travel Agency 3
Hierarchical modeling framework A (user 1 ) A (user N ) … User operational profile User level … User 1 User 2 User k Availability modeling at the user level functions invoked by the users Function A (F 1 A (F n … level ) ) F1 F2 … Fn Availability modeling at the function level Internal & external services implementing Service functions and their interactions A (Si 1 ) A (Si k … level e-business provider External suppliers A (Se 1 ) … A (Se p ) ) Si 1 Si 2 Si m Se 2 Se p … Se 1 … Availability modeling at the service level Architecture on which the services A (Ri 1 ) A (Ri m A (Re p ) … Resource are implemented A (Re 1 ) … ) level Availability modeling e-business provider External suppliers at the resource level Ri 1 Ri 2 Ri k Re 2 Re p … Re 1 … E-business system description E-business availability modeling 4
Function and User levels Home p 32 p 12 p 47 p 24 p 27 p 54 Start Exit Book Search p 57 p 45 p 44 p 56 p 34 p 37 p 67 p 23 Pay p 13 User level Browse User 1 User 2 … User k Function level p 33 F1 F2 … Fn Service level p ij Probability of invoking Fj after Fi internal external Si 1 Si 2 Si m Se 2 Se p Se 1 … … Functions invoked from “Start” to “Exit” ⇔ User scenario Resource level internal external 5 Ri 1 Ri 2 Ri k Re 2 Re p … Re 1 …
User execution scenarios User scenario Class A (%) Class B (%) 1: Start-Home-Exit 10.0 10.0 2: Start-Browse-Exit 26.7 6.6 3: Start-{Home; Browse}*-Exit 11.3 4.2 4: Start-Home-Search-Exit 18.4 13.9 5: Start-Browse-Search-Exit 12.2 20.4 6: Start-{Home; Browse}*-Search-Exit 7.6 9.7 7: Start-Home- {Search-Book}*-Exit 3.0 4.7 8: Start-Browse- {Search-Book}*-Exit 2.0 6.9 9: Start-{Home; Browse}*- {Search-Book}*-Exit 1.3 3.3 10: Start-Home- {Search-Book}*-Pay-Exit 3.6 6.4 11: Start-Browse- {Search-Book}*- Pay-Exit 2.4 9.4 12: Start-{Home - Browse}*- {Search-Book}*- Pay-Exit 1.5 4.5 6
User execution scenarios User scenario Class A (%) Class B (%) 1: Start-Home-Exit 10.0 10.0 } SC1 2: Start-Browse-Exit 26.7 6.6 3: Start-{Home; Browse}*-Exit 11.3 4.2 } 4: Start-Home-Search-Exit 18.4 13.9 SC2 5: Start-Browse-Search-Exit 12.2 20.4 6: Start-{Home; Browse}*-Search-Exit 7.6 9.7 7: Start-Home- {Search-Book}*-Exit } 3.0 4.7 SC3 8: Start-Browse- {Search-Book}*-Exit 2.0 6.9 9: Start-{Home; Browse}*- {Search-Book}*-Exit 1.3 3.3 10: Start-Home- {Search-Book}*-Pay-Exit } 3.6 6.4 11: Start-Browse- {Search-Book}*- Pay-Exit 2.4 9.4 SC4 12: Start-{Home - Browse}*- {Search-Book}*- Pay-Exit 1.5 4.5 7.5% 20.3% 7
Service and function levels Internal services External services Web Application Database Flight Hotel Car Payment Home Browse Search Book Pay 5 3 6 WS End End User level Browse q 5,6 User 1 User 2 … User k Function level F1 F2 … Fn q 2,3 q 4,5 7 8 9 2 1 4 10 Service level internal external WS AS DS AS WS Begin End Si 1 Si 2 Si m Se 2 Se p Se 1 … … q 1,2 q 2,4 q 4,7 q 7,8 q 8,9 q 9,10 Resource level internal external 8 Ri 1 Ri 2 Ri k Re 2 Re p … Re 1 …
Service and Resource levels External service: modeled as a black box Internal services: detailed analysis of e-business provider architecture solutions o Organization of services on the hardware support Dedicated hosts for each server ↔ Multiple servers on the same host o Fault tolerance strategies Non-redundant servers ↔ Replicated servers o Maintenance policies User level Dedicated ↔ Shared repair resources User 1 User 2 … User k Immediate ↔ Delayed maintenance Function level F1 F2 … Fn o Fault assumptions Service level permanent, transient internes externes Si 1 Si 2 Si m Se 2 Se p Se 1 … … Resource level internes externes 9 Ri 1 Ri 2 Ri k Re 2 Re p … Re 1 …
Resource level: TA example Flight reservation Flight reservation Car reservation component system #N C Flight reservation component system #N F E-business Provider site component system component system #1 #1 Database Disk server Hotel reservation Web component system #1 Internet server #N H Application server Payment server Basic architecture Web Database server 1 Application D1 server 1 server 1 Web Redundant server 2 architecture Database Application server 2 D2 server 2 Web server N W 10
TA availability modeling User level Two operational profiles: A, B Function level Home, Browse, Search, Book, Pay Service level Internal External WS, AS, DS Flight, Hotel, Car, Pay Resource level Internal External Flight res. system: A Fi i = 1.. N F Basic architecture Hotel res. system: A Hi Redundant architecture i = 1.. N H Car res. system: A Ci i = 1.. N C Perfect coverage Payment: A PS , Network: A net Imperfect coverage 11
Web service: Modeling approach Composite Availability-Performance model System states resulting from System behaviour resulting from hardware and software request arrival/service processes failures and recoveries conditioned on system states A(Web service) = 1 - Prob. {WS in a down state or overloaded} 12
Web service: Modeling example Redundant architecture: imperfect coverage o Availability model N w c λ (N w -2 ) c λ 2c λ λ N w N w -1 N w -2 1 0 µ ……. µ µ µ µ β β β N w (1-c) λ 2 (1-c) λ y Nw -1 y Nw -2 y 2 o Performance model: M/M/i/K p K (i): probability a request rejected in state i due to WS buffer full N w N w � 2 � � A (Web service) = 1 � � � p K (i) + + � o � i � i y � � i=1 i=1 13
Evaluation results: example 1 10 -1 α = 50/sec 10 -2 α = 100/sec 1- A (WS) 10 -3 λ = 10 -2 /h 10 -4 λ = 10 -3 /h 10 -5 λ = 10 -4 /h 10 -6 10 -7 1 2 3 4 5 6 7 8 9 10 Number of web servers ( Nw) Redundant architecture: imperfect coverage 14
User level availability Composition o Service and Function level availabilities o Operational profile A (A users) A A A (B users) N F = N H = N C 0.842355482 0.768755879 1 0.965090024 0.955291581 2 0.97867277 0.975935054 3 0.980044591 0.978019988 4 0.980181909 0.978228688 5 0.980197168 0.978251879 10 173 hours/year User perceived unavailability ☞ 190 hours/year 15
User level availability 1 1 User perceived Unavailability SC4: Scenarios that end up with a payment 10 -1 10 -1 UA(B users) UA(A users) UA(SC2) UA(SC2) 10 -2 10 -2 UA(SC4) UA(SC1) UA(SC3) UA(SC4) UA(SC1) UA(SC3) 10 -3 10 -3 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 NF=NH=NC NF=NH=NC 16 hours downtime per year 43 hours downtime per year Loss of revenue 570 Million € 1.55 Billion € Transaction rate: 100per sec Revenue per transaction: 100 € 16
Conclusion Contribution o Hierarchical modeling framework ↔ user perceived availability o Example: travel agency Illustration of the main concepts of the proposed framework Examples of dependability analysis results to support design solutions State of the art o Measurement-based evaluation o Modelling:performance oriented Extensions to handle more complex assumptions 17
Recommend
More recommend
