workshop ta2: open logistics interconnection model protocols · open logistics interconnection...
TRANSCRIPT
Workshop TA2: Open Logistics Interconnection Model
& Protocols
In telecommunications, interconnection is the physical linking of a carrier's network withequipment or facilities not belonging to thatnetwork. The term may refer to a connectionbetween a carrier's facilities and the equipment belonging to its customer, or to a connection between two (or more) carriers.
Open Logistics InterconnectionModel & Protocols
Sources: wikipedia
Open Logistics Interconnection Model & Protocols – OLI model
Weight,(Volume,(Temperature,(Hazard5Level(
Single(vs.(mul<ple(des<na<ons,(Single(vs.(mul<ple(sourcing(loca<ons(
Delivery(<me(specifica<ons(
(Number(and(types(of(π5containers((Assignment(of(products(to(π5containers((
Number(of(shipments(Assignment(of(π5containers(to(shipments(Logis<cs(service(class(for(each(shipment(
Target(route(for(each(shipment,(defined(as(sequences(of(segments/nodes,(with(<ming(specifica<ons(
s:(Assignment(of(π5containers(to(π5means(Assignment(of(π5means(to(π5links(
(Genera<on(of(loading,(moving,(sor<ng,(storing,(retrieving(and(unloading(orders(
Montreuil, B., E. Ballot, and F. Fontane. An Open LogisticsInterconnection Model for the Physical Internet. in INCOM 12 Conference. 2012. Bucharest, Romania: IFAC.
Ballot, E., B. Montreuil, and M. Thémans, OPENFRET: contribution à la conceptualisation et à la réalisation d'un hub rail‐route de l'Internet Physique, PREDIT, Editor. 2010, MEDDAT: Paris. p. 114.
Open Logistics InterconnectionModel & Protocols
Agenda: • Presentations
• Discussion
Evaluating five typologies on costs and requirements for hyperconnectedlogistics networksWout Hofman
Networking in the Real World: Unified Modeling of Information and Commodity Distribution NetworksAmitangshu Pal and Krishna Kant
Open Logistics InterconnectionModel & Protocols
Discussion topics• What is the purpose of a protocol suite?• What is interconnection of logistics services?• How to use these frameworks? • …
NETWORKING IN THE REAL WORLD: UNIFIED MODELING OF INFORMATION AND PERISHABLE COMMODITY DISTRIBUTION NETWORKS
Amitangshu PalandKrishnaKantComputerandInformationSciencesTempleUniversity
The Perishability Challenge PerishableInformationDistribution
Increasingamountsoftimesensitiveinformation,includingstreaming&otherreal‐timedata.
Notwellservedbytraditional“besteffort”Internet. Newarchitecturestohandlethese(e.g.,contentcentricnetworks)
PerishableCommodityDistribution Perishability Efficiencytradeoff VaryingperishabilityofproductsMixingproductsverydifficult
Localfoodmovementleadstouniquelocallogisticschallenges. Integrationbetweenlocal&nonlocallogisticsamust
Goal: StudysynergiesbetweenPerishablecommodityvs.Informationdistributiontoadvancebothfields
Information vs. Commodity DistributionManysimilarities Movementofpackagesorpacketsbetween“nodes”withvariouscapabilities
• Change of mode (or “media”), e.g., truck to railcar or barge• Store and forward w/ potential product remix.
Accentuatedbyemergingstandardization(e.g.,GS1)&capacitysharing(e.g.,3PL)ideas
MultipleflowswithvariedSLArequirements Privacyandsecurityrequirements(e.g.,privacypreservingdistribution)
Dynamicallychangingavailabilityanddemand
Information vs. Commodity Distribution
Butmanykeydifferencesaswell Recursivebundling(boxeswithinboxes) Physicalgoodscannotbe“cloned”(exceptatsource) Packet“carrier”(e.g.,truck)andauxiliaryresources(e.g.,driver,loadingequipment)
Complexrequirements(e.g.,longtermcontracts).
Canwecreateaunifiedmodelthatcomprehends InformationNetworks(IN) PerishabilityCommodityDistributionNetworks(PCDN)
A Unified View 5layermodelsimilartoTCP/IPprotocolstack Twotypesofpackets
Non‐clonable Transportationofperishablegoods ClonableComputerNetworks
PacketscanbeofdifferentQoS classes Basedonperishabilityorotherhandlingcharacteristics Eachclasshasaperishabilityfunctionwithin0and1
Physical Layer
Media Switching Layer
Routing & Distribution
Transport/Delivery
Virtualization Layer
Physical layer Layer 1 Dealswithactualmovementofpacketsalongamediasegment/channel IN:Transports“frames”overawired/wirelesschannel
• Max frame size, smaller sizes less efficient PCDN:Physicaltransportoveramediasegment(road,rail,waterway,orair)
• Transport in units (e.g., PI containers) that fit the carrier (e.g., truck)• Carrying smaller PI containers less efficient
PCDNChannelisfarmorecomplex MultipleroadwaysegmentsfrompointAtoBwithdifferentcapacities.
Alternateroutesevenatlayer1– HWYvs.normalroad PCDNtrafficofinterestjustasmallpartofalltrafficthat’sflowing
• Much higher uncertainties & little controllability
Whyisabstractionuseful? Statisticaldescriptionofthepath(e.g.,netcapacity) Canbecomparedtowirelesslinkw/fading&interference
Media Switching layer Layer 2 Mediaaccesscontrol
Assignsthe“channel”(e.g.,pathtonexttransferpoint)• Transfer point: A switch in IN, and a physical transfer point for PCDN
Maydoreframing• IN: Break up into smaller frames, make use of jumbo frames• PCDN: Truck full of PI containers Rail car full of PI containers• Damaged (rotten) frames Discard & request replacement
ComplexitiesinPCDN PCDNneedstoassigncarriers (e.g.,trucks,railcars)tothepackets PCDNneedsadditionalresources(e.g.,drivers) PCDNinvariablyneedstoreturndamagedgoods
• Reverse logistics added complication Layer2routing
IN:Transportbetweensuccessiverouters(layer3endpoints)• Intermediate transfer point (switches & protocol bridges) with change in
media (wired, wireless, optical, …) PCDN:TransferofPIcontainersfromadistributioncentertonext
• Intermediate transfer points may involve media changes (e.g., truck to railcar)
Routing & Distribution layer : Layer 3 End‐to‐endtransferofpackets/packages
IN:Transferfromsrc todest viamultiplerouters• May fragment a TCP stream into multiple IP datagrams• Different types of packets may be mixed before framing
PCDN:Transferfromsrc todest viamultipledistributioncenters• May fragment stream of goods into PI containers• Bundling of different types of contents together quite common
ComplexitiesinPCDN Mayinvolverecursivebundling(boxeswithinboxes) Involvesallocation&trackingL3resourcelikecontainers Bufferspacemanagementlotmorecomplex(contracts?)
RoutingneedsuniqueIDs ForIN,thismaybeamessageordatagramsequencenumber ForPCDNitmaybetheGTIN,GSIN,SSCCnumber
Whyistheabstractionuseful? NumerousroutingalgorithmsexploredinIN Emergingnotionsofcontentcentricnetworking
Transport/Delivery Layer : Layer 4 Endtoend(src todest)deliveryofpackets
IN:Userendpointtouserendpoint PCDN:Farmtoretailer/business(donotconsiderretailpurchases)
Similarities Flowcontrol,packetization,resourceallocation,retransmissionofdamagedgoods.
PCDNcomplexities Batching(accumulatingenoughgoodsfortransport)anessentialcomponent
• Needs to deal with tradeoffs between perishability, cost, efficiency Contractbaseddeliveryscheduling(lessflexibleflowcontrol) Qualitydegradationwithtime&productmixingmuchmorechallenging
• Lateral distribution to handle perishability Reverselogisticsforreturnsandreplacements
Whyistheabstractionuseful? Contentcentric&just‐in‐timemediadelivery
Virtualization Layer : Layer 5 VirtualizationgoalsinIN&PCDN
Sharenetworkcapacityefficientlyamongdifferentapplications Providestablecapacityallocationandisolation
ExamplesforPCDN Definea“HPTransport”asaVSfortransportinghighlyperishable(HP)betweenasrc &dest
• Similar VSs for moderate and low perishable items SeparateVSsfordifferenttypesofcustomers VSforpremiumcustomersorotherlow‐endcustomers
KeyChallenges Themappingofvirtualresourcesonphysicalresources Lackofvisibilityintotheentirenetworkandthedifficultyoftrackingtheentirenetworkstate
Whyistheabstractionuseful? Increasingcomplexityofbothcloudcomputingandlogistics
Resource Management in UNMTheunifiednetworkinvolvesacquisitionofcertainresourcesateachlayerofthenetwork Needtoobtaincarriers&containerstocarryproducts
Returnofcarriers&containers(fullorempty) Availability&properdistributionofcarriersandcontainersimpactstimeliness&freshnessofperishableproductdelivered.
Resourcemanagementcrucialformodelingperformance. Detailsofrepresentationandusageinthepaper
Asimplemodelingillustration
Analytical Model for Commodity Distribution in UNM
Assumedabulkqueuingtheoreticmodel: Trucksareservingthedistributioncenters(DCs)thatareplaceduniformly truckservicetimeisdeterministic
PacketsarearrivingatthedistributioncenterasaPoissonprocess
Distributioncentershavefinitebuffers( packages) PacketsarewastedwhentheDCbufferisfull DCqueueisserveduponarrivalofatruck Thetruckloadsalmost packagesataDC
• If <= packages entire queue is loaded onto the truck truck leaves without waiting for other packages
Thetruckcapacityisassumedtobe . eachDCreservesaspaceof unitsinthetruck
Thisqueuingdisciplinefallsunderthecategoryof / /1/queue
Analytical Model for Commodity Distribution in UNM
100nodesareplacedinanareaof100x100sq.m. Packagefreshnessdegradeslinearlywithtimeatarateof0.25%(atdeliverycenters)and0.35%(ontruck)
Packagedelayincrease deliveryqualitydecrease Withlesserbatchsize highertruckaccesstime WithhigherpackagearrivalrateatDC higherwaitingtime
Package delay against packet arrival rate Package delivery quality against packet arrival rate
Analytical Model for Commodity Distribution in UNM
Tradeoffbetweentransportationefficiencyanddeliveryquality: Increaseinnumberoftrucks improvesthe deliveryqualityaswaiting
timeofthepackagesreduces Howeverthetransportationefficiencyreducesduetolesseravailable
packagesateachDC
IncreasingBloadsmorenumberofpackagesatanyparticularDCimprovesthedeliveryqualityespeciallyincaseofsmallernumberoftrucks
Conclusions Weconsideredthesynergiesbetweeninformationandcommoditydistributiondisciplines Devisedaunifiedmodeltocaptureboth Discussedananalyticalframeworktogetaninsightregardingthekeyperformanceparameters
Thereisalwaysatradeoffbetweenthetransportationefficiencyanddeliveryquality
Ongoingworks: Wedesignedadatacenteropticalnetworkinspiredbyintegrationoflocal&nonlocallogistics
Usetheperishablelogisticsconceptsincontentcentricnetworks
Makingtheperishablelogisticsmoreresilienttospoilageandcontaminationbysensingandlocaldelivery