Docstoc

Methodologies to Support FT for CORBA Objects

Document Sample
Methodologies to Support FT for CORBA Objects Powered By Docstoc
					 Principles and Patterns for QoS-
enabled Fault Tolerant Middleware

            Aniruddha Gokhale
       Bell Labs, Lucent Technologies
      agokhale@research.bell-labs.com




                      Presented at
        ISIS, Vanderbilt University, Nashville, TN
                Friday, August 10, 2001
       Aniruddha Gokhale                                          Patterns for FT Middleware

           Motivating Technology Forces
Total Ship Computing Environments
                                             TRACKING         • Hardware is becoming faster
  IOM                                               IOM
          SATELLITES Total Ship C&C Center
                                              STATION           & cheaper
                                               PEERS
 IOM     BSE            BSE                   BSE   IOM
                                                              • Ubiquitous & affordable
 IOM                                                IOM         wireless/wireline broadband
                                                                internet connectivity
                STATUS INFO
 IOM                                                IOM
    WIDE AREA                                                 • Distributed real-time &
    NETWORK
                                                                embedded (DRE) system are
 IOM     BSE COMMANDS BSE                     BSE DATA
                                              BULK
                                                    IOM
                                              TRANSFER
                                                                becoming more complex &
 IOM                                                IOM         mission-critical
       GATEWAY
   Modalities                                                 • Increasing demand for
 IOM                                                IOM
 e.g., MRI, CT, CR,                                             COTS-based multi-
                 LOCAL AREA NETWORK
  Ultrasound, etc. BSE
 IOM BSE                            BSE             IOM
                                                                dimensional quality-of-service
  GROUND                                                        (QoS) support
 STATION                                                         • e.g., simultaneous requirements for
 IOM
   PEERS                                            IOM
                                                                   efficiency, predictability, scalability,
                                                                   security, & dependability
                                                          2
Aniruddha Gokhale                    Patterns for FT Middleware
                               IOM                                IOM
                               IOM   BSE       BSE        BSE     IOM
                               IOM                                IOM
                               IOM                                IOM
                               IOM   BSE       BSE        BSE     IOM
                               IOM                                IOM
                               IOM                                IOM
                               IOM   BSE       BSE        BSE     IOM
                               IOM                                IOM

                Key open challenge
       QoS-enabled dependability




                           3
    Aniruddha Gokhale                              Patterns for FT Middleware

                      Research Synopsis                               Applications
 Create the new generation of middleware technologies for
 distributed real-time & embedded (DRE) systems that enable            Middleware
  1. Simultaneous control of multiple QoS properties &
  2. Composable & customizable common technology bases               OS & Protocols
                                                                       Hardware

   Research Challenge         Research Approach                 Impact
• Domain-specific           •Model description          •Model-driven
  middleware services       •Code generation             approach for network
  shielding lower-level      mapping to lower            mgmt of 3G wireless
  middleware                 level middleware            components
• Standardized              •Influence standards        •Fault Tolerant
  middleware for fault       body from experience        middleware Standard
  tolerance                  with prototype
•High-Performance           •Iterative process          •QoS Enabled
  Middleware                 involving                   Middleware
                             benchmarking,
                             quantifying/profiling,
                             and optimizing
•Robust Middleware          •Based on patterns          •Dependable,
                                          4
                                                         Extensible
Aniruddha Gokhale                     Patterns for FT Middleware
     FT CORBA Contributions and Impact
                                  • Influenced the Fault
                                    tolerant CORBA
                                    standard by
                                    incorporating principles
                                    and patterns from a
                                    prototype called DOORS

                                  • Patterns-based
                                    optimizations in DOORS
                                    provides high
                                    performance &
                                    predictability to DRE
                                    systems


          www.bell-labs.com/~agokhale/DOORS.html
                              5
Aniruddha Gokhale         Patterns for FT Middleware

  Overview of Fault Tolerant CORBA
                    Overview
                    •Provides a standard set of
                     CORBA interfaces, policies, &
                     services
                    •Entity Redundancy of objects
                     is used for fault tolerance via
                      •Replication
                      •Fault detection &
                      •Recovery from failure
                    Features
                    •Inter-Operable Group
                     References (IOGR)
                    •Replication Manager
                    •Fault Detector & Notifier
                    •Message Logging for recovery
                    •Fault tolerance Domains

                    6
Aniruddha Gokhale       Patterns for FT Middleware

      FT CORBA Use Case Scenario
                              Base stations
                              (Node B) manage
                              several thousand
                              subscribers
                              modeled as
                              objects
                             Radio Network
                             Controllers
                             manage several
                             Node Bs modeled
                             as objects
                             requiring high
                             reliability and
                             availability (3 min
                             per year downtime)
                    7
Aniruddha Gokhale                    Patterns for FT Middleware

          Effect of Polling Interval on
             Total Recovery Time
                            Analysis
                            • Average failure detection time is
                              half the polling interval
                            • Replica Group Management time
                              is constant


                            Key Challenge
                            • Minimize replica group
                              management time
                            • Setting the right polling interval

 Recovery Time = Failure detection time + Replica Group
 Management Time
                            8
    Aniruddha Gokhale                       Patterns for FT Middleware

  Optimization Opportunities to Improve
   Fault Tolerant CORBA Performance
ORB Core Optimizations
•Efficient IOGR parsing &
 connection establishment
•Reliable handling & ordering
 of GIOP messages
•Predictable behavior during
 transparent connection
 establishment &
 retransmission
•Tracking requests with
 respect to the server object
 group
CORBA Service Optimizations
•Support for dynamic system configuration
•Bounded recovery time
•Minimize overhead of FT CORBA components
                                   9
     Aniruddha Gokhale                           Patterns for FT Middleware

                   Overview of Patterns
 Patterns formalize
 expert knowledge to
 help generate software
 architectures by
 capturing recurring
 structures & dynamics
 and resolving common
 design forces

Design patterns capture the static          www.posa.uci.edu/
& dynamic roles & relationships in   Architectural patterns express a
solutions that occur repeatedly      fundamental structural organization for
Optimization principle patterns      software systems that provide a set of
document rules for avoiding          predefined subsystems, specify their
common design & implementation       relationships, & include the rules and
mistakes that degrade performance    guidelines for organizing the relationships
                                     between them
                                       10
    Aniruddha Gokhale                            Patterns for FT Middleware
 Challenge: Predictable and Scalable Fault
                Monitoring
Naïve implementation of Polling Thread                              Replica

while (true) {                         Fault Detector
                                                              HANGS
                                                                 Replica
    for each obj to be monitored {          Polling
                                            Thread
        poll the object;
                                                                    Replica
        if (no response) {
            report fault to notifier
                                                                    Replica
        }                                    HANGS
    }                                        Fault
                                            Notifier
    sleep (polling_interval);
}
                                       Allocate new thread per
                                       monitored object?
                                       11
Aniruddha Gokhale              Patterns for FT Middleware

Predictable & Scalable Fault Monitoring
                         Context
                         •Periodic polling of several objects
                          & fault notification done in the
                          same polling thread can block the
                          thread
AMI                       Problem
                          •Blocking can cause missed polls
                          Forces
                          •Must guarantee polling of all
                           registered objects
                          •Must minimize concurrency
                           overhead
                          •Must be scalable
                         Solution
                         •Leader-Followers and ACT
                          architectural pattern

                    12
   Aniruddha Gokhale                     Patterns for FT Middleware

Challenge: Scalable, Prioritized Fault Recovery
                                            Context
                                            •RM needs to handle
                                             bursts of fault reports
                                            •Handle faults based on
                                             importance

                                            Problem
                                            •Reduced responsiveness
                                            •Wasted resources during
                                             normal conditions
                                            •Lack of handling faults
                                             based on importance
                                            Forces
           Solution                         •Predictable amount of
           •Apply the Thread Pool            time for failure recovery
            Active Object design             irrespective of burst size
            pattern                         •Prioritize fault handling
                                    15
   Aniruddha Gokhale                     Patterns for FT Middleware

 Challenges using QoS-enabled Middleware
  Domain-Specific           Network mgmt, SIP, Location-based
Middleware Services         services, web hosting services

     Common                Distributed Services
Middleware Services        • Fault tolerance, Notification, Naming,

    Distribution
    Middleware
                       }     Event, Logging, Transaction, etc
                           QoS-enabled Distribution Middleware
                           • RT/FT CORBA, RT Java, .NET

  Infrastructure             Platform independent: e.g., ACE, Java
                             VMs, RogueWave, Microsoft CLR
   Middleware
Operating Systems          Realtime OS e.g., VxWorks, PSOS
   & Protocols             General OS e.g., Solaris, Win2K, Linux

    Hardware               IntServ/DiffServ using COTS Networks


                                18
   Aniruddha Gokhale              Patterns for FT Middleware

 Challenges using QoS-enabled Middleware
Meta programming       • Configuring middleware QoS
                         parameters is hard
  Domain-Specific
Middleware Services • Needs expert understanding
     Common           of middleware layers and their
Middleware Services   APIs
    Distribution
    Middleware      • Big learning curve for
   Infrastructure     application developers
    Middleware
Operating Systems • Applications get tightly
    & Protocols       coupled with the underlying
      Hardware        middleware technology
                           19
  Aniruddha Gokhale                         Patterns for FT Middleware

Challenge: Resolving Economic Forces
            Context                              Solution
 • Less spending by customers          • Reduce/amortize costs
   => Shrinking revenues and             associated with existing
   profit margins                        product lifecycles

 • Maintaining competitive
   advantage and higher returns        • Rapid Time-to-market of high
   on investment                         quality products

 • Plethora of competing               • Platform independent, highly
   technologies e.g., CORBA              interoperable design
   CCM, EJB/RMI, COM+,
   XML/SOAP
 • Dealing with disruptive             • Minimal effort in incorporation
   technologies or                       of upcoming/new enabling
   unprecendented hype                   technologies
                                  20
  Aniruddha Gokhale                       Patterns for FT Middleware

Candidate Solution: Model Driven Architecture
                      • Consequence of codifying years of R&D
                        efforts in:
                                    • Meta-programming
                                      techniques, e.g., UML/XML
                                     • Model-integrated computing
                                     • Middleware-centric pattern
                                       languages
                                     • Platform independence,
                                       interoperability
                                     • Sophisticated compiler
                                       techniques & code generation
   www.omg.org/mda
                                       tools
Open R&D question:
How can we apply MDA to DRE systems most effectively
                                21
Aniruddha Gokhale             Patterns for FT Middleware

             Challenges using MDA




                         Problems
                         •Models used primarily for
                          feasibility and system
                          schedulability analysis
                         •High software lifecycle costs
                           •e.g.,mapping models to
                            platforms incurs many
                            “accidental complexities” &
                            low-level platform
                       22   dependencies
       Aniruddha Gokhale                        Patterns for FT Middleware

                  AMW Project: MDA in Action
                                                Aurora library management software
                                                Aurora generated management code
                  (1)                     (4)   Application-specific management s/w
                                                Application non-management software
   .


  (2)                                           (3)
            Code
 input    Generator
specifi
cation
1. Software libraries provide application-independent management building
   blocks
2. Code generators from input model descriptions customize building
   blocks for specific applications
3. Application developers implement application-specific management
   building blocks
4. Building blocks cooperate to implement management functionality

                                     23
        Aniruddha Gokhale                               Patterns for FT Middleware

              Aurora Management Workbench
                                                        Aurora library management software
                                                        Aurora generated management code
                    (1)                           (4)   Application-specific management s/w
                                                        Application non-management software
    .


    (2)                                                 (3)
              Code
 input      Generator
specifi
cation
•       Contributions
        •   Fault Escalation strategies during initialization and recovery
        •   Component status querying
        •   Event Notification
        •   Dynamic growth/degrowth
•       Impact
        •   Management of network elements in 3G (UMTS) wireless architecture

                                             24
Aniruddha Gokhale               Patterns for FT Middleware

               Crossing the Chasm




                          Model-driven multidimensional
                           QoS management (performance,
                           scalability, predictability,
                           dependability, security) in
                           heterogenous networks
                          Possible Application Domains
                             Defense
                             Transportation
                             Healthcare
                             Finance
                             Consumer Products
                         25
    Aniruddha Gokhale                           Patterns for FT Middleware

      Concluding Remarks & Future Work
                              • Model Driven Architecture intends to
                                resolve the technological and
                                economic forces, however …
                               • Researchers & developers of
                                 distributed systems face common
                                 challenges, e.g.:
                                  • Connection management, service
                                    initialization, error handling, flow control,
                                    event demuxing, distribution, concurrency
• The application of patterns,      control, fault tolerance, synchronization,
  frameworks, & components          scheduling, & persistence
  can help to resolve these •Carefully applying these techniques
  challenges                   can yield efficient, scalable,
                             predictable, dependable, & flexible
Key challenge                middleware & applications
• Patterns bridging the gap between models and middleware
• Model-driven multidimensional QoS management
                                      26
Aniruddha Gokhale              Patterns for FT Middleware




                    EXTRA SLIDES




                          27
        Aniruddha Gokhale                                                  Patterns for FT Middleware

           Concluding Remarks & Future Work
            PROBLEM                    PATTERN               CATEGORY      • Researchers &
1    Missed Polls in              Leader/Followers, AMI Architectural        developers of distributed
     FaultDetector                                                           systems face common
2    Excessive Overhead of        Active Object             Design
     recovery                                                                challenges, e.g.:
3    Excessive overhead of        Optimize for common       Optimization
     service lookup               case
                                                                             • Connection management,
                                  Store extra information   Optimization       service initialization, error
                                  Eliminate gratuitous      Optimization       handling, flow control,
                                  waste
4    Tight coupling of data       Strategy                  Design
                                                                               event demuxing,
     structures                   Abstract Factory          Design             distribution, concurrency
5    Lack of dynamic              Component                 Design             control, fault tolerance,
     configuration                Configurator                                 synchronization,
6    Inefficient Property value   Chain of responsibility   Design
     lookup
                                                                               scheduling, & persistence
                                  Perfect hash Functions    Optimization

• The application of patterns •Carefully applying these techniques can
  can help to resolve these    yield efficient, scalable, predictable,
  challenges                   dependable, & flexible middleware &
                               applications

    Model-driven multidimensional QoS management
                                                               28
Aniruddha Gokhale           Patterns for FT Middleware

    Middleware for Broadband Cable




                         • Video-on-demand
                         • Telephony and other
                           services over cable
                         • QoS management of
                           infrastructure components

                    29
Aniruddha Gokhale            Patterns for FT Middleware

  Middleware for Broadband Wireless




                         •    Voice-over-IP-over-wireless
                         •    M-Commerce applications
                         •    Messaging Services
                         •    Location Services
                         •    Security
                         •    Billing
                         •    QoS management of
                              infrastructure components


                    30
Aniruddha Gokhale              Patterns for FT Middleware

     Middleware for Adhoc Networks




                         •   Peer-to-peer communication
                         •   Efficient routing
                         •   Fault tolerance and QoS
                         •   Seamless interoperability
                             between Bluetooth, 802.11,
                             3G



                    31
       Aniruddha Gokhale                                                                       Patterns for FT Middleware

          Problems with Modeling Approaches
• Dynamic DRE combat system • Likewise, the proprietary multiple technology bases in
  QoS requirements historically     DRE combat systems today limit effectiveness by impeding
  not supported by COTS              • Assurability (of QoS),
   • i.e., COTS is too big, slow,    • Adaptability, &                Today, each combat
     buggy, incapable, & inflexible  • Affordability                  system brings its own:
                                                                                                                     networks
                                                                                                                    computers
                                                                                                                     displays
                                                                                                                     software
                                                                                                                      people

 Applications                                                                    Applications             Problems
  Sensor           Command            Engagement           Weapon                        Weapon           • Non-scalable tactical
  Systems          & Control            System             Control                       Systems            performance
                    System                                 Systems
                                                                Kill
                                                                                                          • Inadequate QoS control
                                                           EO
                                                                Eval
                                                                       Sched

                                                                 Network
                                                                               Illum
                                                                                                            for joint operations
                                                          AAW
                                                           EG
                                                                AAW AAW
                                                                TBM
                                                                 AAW MG
                                                                 EG   AAW
                                                                     AAW
                                                                               TMB
                                                                               MG
                                                                                                            • e.g., distributed weapons
                                                                                                              control
Technology base:
 Proprietary MW
                   Technology base:
                       DII-COE
                                      Technology base:
                                       Proprietary MW
                                                         Technology base:
                                                          Proprietary MW
                                                                                       Technology base:
                                                                                        Proprietary MW
                                                                                                          • High software lifecycle
    Mercury             POSIX              POSIX             VxWorks                        POSIX           costs
   Link16/11/4       ATM/Ethernet          NTDS             FDDI/LANS                      VME/1553
                                                                                                            • e.g., many “accidental
                                                                                                              complexities” & low-level
Operating                                                                                  Operating          platform dependencies
System                                                                                       System
                                                                           32
    Endsystem                  Wireless/Wireline Networks                              Endsystem
    Aniruddha Gokhale                             Patterns for FT Middleware

                 Research Contributions
      Context              Technical          Approach             Impact
                           Challenge
• Meta Programming • N/W elements          • High-level        N/W Mgmt of
• Domain-specific    need mgmt               Config file        3G Wireless
  middleware       • Not reinvent          • Code               Components
  services           mgmt code               generation
• Core Middleware  • Standardized          • Influence         Fault Tolerant
  Services           middleware for          standards body     CORBA
                     fault tolerance         from               Standard
                                             experience with
                                             prototype
• Distribution          • High-            • Iterative         QoS Enabled
  middleware              Performance,       process            Middleware
                          Realtime           involving
                          Middleware         benchmarking,
                                             quantifying/pro
                                             filing, and
                                             optimizing
• Pattern languages     • Robust,          • Based on          Dependable,
                          configurable       middleware-        Extensible,
                          middleware     35 centric patterns   Flexible code
  Aniruddha Gokhale         Patterns for FT Middleware

Distribution Middleware Research Contributions
  Meta programming
   Domain-Specific
 Middleware Services
      Common
 Middleware Services
     Distribution                TAO Real-time
     Middleware                  CORBA ORB
                                 Contributions
   Infrastructure
    Middleware
 Operating Systems
    & Protocols
     Hardware
                       36
  Aniruddha Gokhale                       Patterns for FT Middleware

Distribution Middleware Research Contributions
           Andy, since this is older work, I recommend
           that you move it to later in the talk, I.e.,
           after the DOORS stuff. Moreover, I
           recommend that you add a slide or two that
           shows an example of the kinds of work you
           did on this topic, e.g., the principle patterns
           + optimizations.




         www.cs.wustl.edu/~schmidt/TAO.html
                                  37
   Aniruddha Gokhale          Patterns for FT Middleware
Core Middleware Services Research Contributions
    Meta programming
     Domain-Specific
   Middleware Services
        Common                   Fault tolerant
   Middleware Services           CORBA (DOORS)
       Distribution
       Middleware
      Infrastructure
       Middleware
    Operating Systems
       & Protocols
        Hardware
                         38
  Aniruddha Gokhale        Patterns for FT Middleware

Domain-specific Middleware Contributions
Meta programming           HARP/AMW
  Domain-Specific          Contributions
Middleware Services
     Common
Middleware Services
    Distribution
    Middleware
  Infrastructure
   Middleware
Operating Systems
   & Protocols
    Hardware
                      39
Aniruddha Gokhale                       Patterns for FT Middleware

    Domain-specific Middleware and
    Meta-programming Contributions
      Andy, it’s not clear how the title of this slide
         corresponds to what the diagram is and
         what the text bullet say. • I don’t understand
                                       Majority of network
                                       element software
         what value this slide has, so please try to is
                                       management software (60
         motivate it better.
                                       – 80 %)
                                   • Includes sequenced
                                       initialization, fault
         Management                    tolerance, load balancing,
           Software                    growth/degrowth,
                                       maintenance, component
                                       querying, and event
                                       notification
    Network Element Code

                               40
Aniruddha Gokhale            Patterns for FT Middleware

Supporting Interchangeable Behavior
                    Context
                    •FT properties can be set statically
                     (as defaults) or set dynamically
                    Problem
                    •Hard-coding properties make the FT-
                     CORBA design inflexible & non-
                     extensible
                    Forces
                    •Need highly extensible
                     services that can be composed
                     transparently from configurable
                     properties
                    Solution
                    •Apply the Strategy design
                     pattern
                    41
Aniruddha Gokhale             Patterns for FT Middleware

            Consolidating Strategies
                             Context.
                             •FT CORBA implementations
                              can have many properties.
                              • e.g.,membership, replication,
                                consistency, monitoring, # of
                                replicas, etc.
                             Problem
                             •Risk of combining semantically
                              incompatible properties
                             Forces
                             •Ensure semantically
                              compatible properties
                             •Simplify management of
                              properties
                             Solution
                             •Apply the Abstract
                              Factory design pattern
                        42
Aniruddha Gokhale               Patterns for FT Middleware

             Dynamic Configuration
                            Context
                            •There are many potential FT
                             properties that can be used

                            Problem
                            •Static configuration of
                             properties is inflexible & overly
                             resource intensive
                            Forces
                            •The behavior of FT-CORBA
                             properties should be
                             decoupled from the time when
                             they are actually configured
                            Solution
                            •Apply the Component
                             Configurator design pattern
                       43
Aniruddha Gokhale             Patterns for FT Middleware

       Research Directions (TO-DO)
                      Middleware for broadband cable
                       (set top boxes)
                      3G wireless
                       services/applications
                      Intelligent networking, billing,
                       network management, real-time
                       monitoring, load balancing
                      Web services/E services
                      E/M commerce
                      FT+RT systems
                      Smart transducer RFP
                      Wireless CORBA




                    44
Aniruddha Gokhale        Patterns for FT Middleware

     Benchmarking Results (TO-DO)




                    45
Aniruddha Gokhale                Patterns for FT Middleware

         Caching Object References
                          Context
                          •Replication Manager queries
                           Naming Service for Fault
                           Detector object reference

                          Problem
                          •Adds overhead of querying

                          Forces
                          •Minimize time spent in
                           querying for object reference

                          Solution
                          •Optimize for common case,
                           store redundant info, eliminate
                           gratuitous waste

                     46
     Aniruddha Gokhale                      Patterns for FT Middleware

    Interoperable Object Group References
•Composite & enhanced
 Interoperable Object Reference
 (IOR) for referencing server object
 groups
•Comprises one or more
 TAG_INTERNET_IOP profiles,
 which in turn must contain a
 TAG_FT_GROUP and zero or more
 TAG_IIOP_ALTERNATE_ADDRESS
 components
•TAG_PRIMARY component in at
 most one TAG_INTERNET_IOP
 profile
•Client ORBS operate on IOGRs in
 the same way as with IORs

                                       47
 Aniruddha Gokhale          Patterns for FT Middleware

 Promising Solution: Fault Tolerant (FT)
Distributed Object Computing Middleware
                     Challenges
                     •Limitations of non-OO FT
                      strategies that focus on
                      application processes
                     •Techniques based on process-
                      based failure detection &
                      recovery are not applicable to
                      distributed object computing
                      applications due to:
                       1.Overly coarse granularity
                       2.Inability to restore complex
                         object relationships
                       3.Restrictions on process
                         checkpointing & recovery
                     48
Aniruddha Gokhale              Patterns for FT Middleware

 Fault Tolerant Middleware Strategies
                    Integration Strategy
                     •Modify middleware to support FT, e.g.:
                       •Orbix+Isis
                        www.iona.com
                       •Electra
                        www.softwired.ch/people/maffeis/
                        electra.html
                     •Obtrusive


                    Interception Strategy
                     •Intercepts messages outside the
                      ORB, e.g.:
                       •Eternal
                         www.eternal-systems.com
                       •AQuA
                        www.crhc.uiuc.edu/PERFORM/
                        AQuA.html
                     •Unobtrusive, but complex
                      49
   Aniruddha Gokhale                      Patterns for FT Middleware

Fault Tolerant Middleware Strategies (cont’d)
                              Service Strategy
                               • FT as a higher-layer service, e.g.,
                                  •DOORS
                                  •www.bell-labs.com/
                                   org/11356/html/doors.html
                               • Unobtrusive, but requires standard ORB
                                 support

                       Summary of FT Strategies
                        • Integration strategy requires extensive non-
                          portable & non-standard modifications to an
                          ORB
                        • Interception strategy provides out-of-band
                          solution that can be inefficient (due to
                          duplication of effort) and is hard to port
                          across operating systems
                        • Service strategy requires some ORB
                          modifications, but is now standardized…
                                50
Aniruddha Gokhale             Patterns for FT Middleware

             DOORS & FT-CORBA
                          •DOORS is a “Distributed OO
                           Reliable Service” developed
                           prior to FT-CORBA
                           •Uses the service strategy to
                            provide FT to CORBA objects
                          •Patterns and mechanisms in
                           DOORS were integrated into
                           FT-CORBA standard
                          •DOORS implements most of
                           FT-CORBA standard
                           •Focus on passive-replication
                           •Available as open-source for
                            non-commercial use from
                            Lucent
                           •Runs atop the TAO open-source
                            real-time ORB
                             •www.theaceorb.com
                     51
Aniruddha Gokhale                   Patterns for FT Middleware

               Replication Manager
                             Components
                             •Property manager
                              • Allows properties to be set for
                                an object group
                              • Properties include replication
                                style, membership style,
                                consistency style, monitoring
                                style, & number of replicas
                             •Generic factory
                              • Creates object groups & each
                                member of the object group
                              • Used when the membership-
                                style is infrastructure-controlled
                             •Object group manager
                              • Used by applications to create,
                                add or delete members of an
                                object group

                        52
Aniruddha Gokhale                Patterns for FT Middleware

        Fault Detection & Notification
                       Components
                       •Fault Detectors
                        •Detect faults using a pull-
                         based or a push-based
                         mechanism
                       •Fault Notifier
                            •Notified of any faults by
                             the Fault Detector
                            •A Fault Notifier
                             notifies its Replication
                             Manager when it detects
                             faults


                       53
Aniruddha Gokhale              Patterns for FT Middleware

               Logging & Recovery
                        •Used in a infrastructure-
                         controlled consistency
                         style
                        •The mechanism
                         intercepts & logs GIOP
                         messages
                        •On failure detection, the
                         messages can be “played
                         back” for recovery
                        •Mechanism is transparent
                         to the client application




                        54
Aniruddha Gokhale         Patterns for FT Middleware

     FT CORBA Spec Requirements
                         •Preserving the CORBA
                          Object Model with
                          enhancements
                         •No single point of failure
                         •Transparent failover
                         •Transparent client
                          redirection & reinvocation
                         •FT CORBA cannot
                          handle commision faults
                          or correlated faults




                    55
   Aniruddha Gokhale                   Patterns for FT Middleware

             Effect of Polling Interval on
              Failure Detection Times
                                Analysis
                                • Failure detection time
                                  increases with the polling
                                  interval
                                • Average failure detection time
                                  is half the polling interval
                                Challenge
                                • Choosing small polling
                                  interval
                                • Minimize message overhead

Fault detection time measured as the time between the failure
of replica & the FaultDetector detecting failure
                               56
   Aniruddha Gokhale                          Patterns for FT Middleware

                   Research Synopsis
   Technical Challenge      DOORS Approach                 Impact
• Standardized           •Influence standards       Fault Tolerant
  middleware for fault    body from experience       CORBA Standard
  tolerance               with prototype
•High-Performance        •Iterative process         QoS Enabled
  Middleware              involving                  Middleware
                          benchmarking,
                          quantifying/profiling,
                          and optimizing
•Robust Middleware       •Based on patterns         Dependable,
                                                     Extensible

•DOORS is a “Distributed OO Reliable Service” developed prior
 to FT-CORBA
 •Uses the service strategy to provide FT to CORBA objects
•Patterns and mechanisms in DOORS were integrated into FT-
 CORBA standard
                                    57
Aniruddha Gokhale              Patterns for FT Middleware

             What is CORRIDOR ?
 CORRIDOR = Component Oriented
 Reliable Realtime Infrastructure for
  Distributed Object MiddlewaRe
    Next generation of meta middleware
    technology for distributed real-time and
    embedded (DRE) telecom systems that
    enable
     1.Simultaneous control of multiple
      Quality of Services properties
     2.Composable & customizable common
      platform bases for telecommunications
      and data networking applications
                        58
Aniruddha Gokhale              Patterns for FT Middleware

                    Architecture




                          59
   Aniruddha Gokhale                     Patterns for FT Middleware

            CORRIDOR Dependencies
   Applications
 Domain-Specific                                will leverage from
    Services                                    lower layers
    Common                 QoS-enabled Distribution Middleware
    Services               • RT/FT CORBA, RT Java, .NET

   Distribution
   Middleware
                       }   Services
                           • Fault tolerance, Notification, Naming,
                             Event, Logging, etc

  Infrastructure             Platform independent: e.g., ACE, JVM,
                             RogueWave
   Middleware
Operating Systems          Realtime OS e.g., VxWorks, PSOS
   & Protocols             General OS e.g., Solaris, Win2K, Linux

    Hardware               Network elements using COTS hardware
                                60
      Aniruddha Gokhale                                                 Patterns for FT Middleware

                                      CORRIDOR Impact
       Technical Challenge                   CORRIDOR Approach                                Impact
• Generating optimized middleware        • Algorithms & tools to auto-generate   Time-to-Market
                                           optimized code for repetitive tasks   • Saves time and resources to
                                                                                   implement code



• Supporting multiple QoS Properties     • Components providing fault            Configurability
                                           tolerance, real-time scheduling,      • Rapid development of
                                           throughput and/or latency, dynamic      applications requiring multiple
                                           growth/degrowth, and runtime            QoS properties by composition of
                                           upgrades                                CORRIDOR components
• Assuring dynamic flexibility and QoS   • Robust, yet dynamically               Adaptability
  in telecom systems                       reconfigurable, meta-programming      • Support varying workloads &
                                           middleware                              environment over telecom system
                                                                                   lifecycles

• Interoperability with third party      • Based on standards-based              Competitive Edge
  components                               middleware                            • Rapidly integrate newer or third
                                                                                   party components at lower costs
                                                                                   over life cycle of installed base


• Formalizing QoS-related design         • Pattern languages & frameworks to     Affordability
  expertise for telecom systems            encapsulate complexity in meta        • Reduce recurring maintenance
                                           middleware                              costs
                                                                                 • Reduced initial costs for new
                                                                                   products


                                                          61
    Aniruddha Gokhale                 Patterns for FT Middleware

                        Novel Features
•    Tools for automatic code generation for repetitive
     features that require domain specific knowledge (e.g.
     fault tolerance, real-time, security)

•    Composition of service-enabled components based on
     declarative description in XML or similar formats

•    Adaptive and Reflective capabilities to fine tune the
     system vertically in order to optimize resource usage

•    Abstraction over distribution middleware allows newer
     middleware technology to be plugged underneath

•    Support for run-time upgrades and growth/degrowth
                               62
  Aniruddha Gokhale                  Patterns for FT Middleware

                Technical Challenges
• How to effectively decouple application-specific software
  from management software in different dimensions (fault
  tolerance, security, real-time)?

• How to design an extensible and flexible infrastructure of
  components that enables the rapid incorporation of new
  technologies?

• How to specify QoS policies and requirements in
  different dimensions? Is XML suitable?

• How to monitor resources at runtime and adapt without
  causing performance degradation?

• How to handle run-time growth and software upgrades?
                              63
Aniruddha Gokhale                             Patterns for FT Middleware

                Research Directions
 CORRIDOR: QoS-enabled framework of middleware components
    Higher level middleware framework shielding applications from
     lower level middleware
    Multi-dimensional QoS support
    Patterns-based architecture of plug & play components
    Code generation tools for repetitive tasks
 Middleware for Ad hoc/Wireless networks
    FT CORBA enhancements for JINI-like systems
    CORBA Pluggable Protocol for Bluetooth devices
    Middleware enhancements for 3G wireless/mobile internet
 Fault tolerance and FT CORBA
    Sequenced Initialization and Recovery (dealing with object
     dependencies)
    Handling failure groups and collocated groups
    Fault Escalation strategies and Fault Analysis
    Growth/degrowth, runtime upgrades

                                    64
Aniruddha Gokhale                Patterns for FT Middleware

               Acknowledgments
FT CORBA (DOORS)
• Douglas C. Schmidt (schmidt@uci.edu)
• Balachandran Natarajan (bala@cs.wustl.edu)
• Shalini Yajnik (shalini@precache.com)


Network Management (HARP/AMW)
• Richard Buskens (rbuskens@lucent.com)
• Ali Siddiqui (alisidd@lucent.com)
• Oscar Gonzalez (ojgonzale@lucent.com)



                          65

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:3
posted:11/16/2011
language:English
pages:59