Decoupling QoS Control from Core Routers: A Novel Bandwidth Broker Architecture for Scalable Support of Guaranteed Services*

Size: px
Start display at page:

Download "Decoupling QoS Control from Core Routers: A Novel Bandwidth Broker Architecture for Scalable Support of Guaranteed Services*"

Transcription

1 Decouplng QoS Control from Core Routers: A Novel Bandwdth Broker Archtecture for Scalable Support of Guaranteed Servces* Zh-L Zhang y, Zhenha Duan y, Lxn Gao z, and Ywe Thomas Hou x y Unversty of Mnnesota z Smth College x Fujtsu Labs of Amerca Mnneapols, MN Northampton, MA 0060 Sunnyvale, CA fzhzhang,duang@csumnedu gao@cssmthedu thou@flafujtsucom ABSTRACT We present anovel bandwdth broker archtecture for scalable support of guaranteed servces that decouples the QoS control plane from the packet forwardng plane More specfcally, under ths archtecture, core routers do not mantan any QoS reservaton states, whether per-flow or aggregate Instead, QoS reservaton states are stored at and managed by bandwdth broker(s) There are several advantages of such a bandwdth broker archtecture Among others, t releves core routers of QoS control functons such as admsson control and QoS state management, and thus enables a network servce provder to ntroduce new (guaranteed) servces wthout necessarly requrng software/hardware upgrades at core routers Furthermore, t allows us to desgn effcent admsson control algorthms wthout ncurrng any overhead at core routers The proposed bandwdth broker archtecture s desgned based on a core stateless vrtual tme reference system developed n [20] In ths paper we focus on the desgn of effcent admsson control algorthms under the proposed bandwdth broker archtecture We consder both per-flow guaranteed delay servces and class-based guaranteed delay servces wth flow aggregaton We demonstrate how admsson control can be done on an entre path bass, nstead of on a hop-byhop" bass Such an approach may sgnfcantly reduce the complexty of the admsson control algorthms We also study the mpact of dynamc flow aggregaton on the desgn of class-based admsson control algorthms Based on the proposed bandwdth broker archtecture, we devse effectve mechansms to crcumvent the problem caused by dynamc flow aggregaton Λ Ths work was supported n part by NSF under the CA- REER Award NCR Any opnons, fndngs, and conclusons or recommendatons expressed n ths paper are those of the authors and do not necessarly reflect the vews of NSF Permsson to make dgtal or hard copes of all or part of ths work for personal or classroom use s granted wthout fee provded that copes are not made or dstrbuted for proft or commercal advantage and that copes bear ths notce and the full ctaton on the frst page To copy otherwse, or republsh, to post on servers or to redstrbute to lsts, requres pror specfc permsson and/or a fee SIGCOMM '00, Stockholm, Sweden Copyrght 2000 ACM /00/0008 $500 INTRODUCTION The ablty to provde end-to-end guaranteed servces (eg, guaranteed rate or delay) for networked applcatons s a desrable feature of the future Internet To enable such servces, Qualty-of-Servce (QoS) support from both the network data plane (eg packet schedulng) and the control plane (eg, admsson control and resource reservaton) s needed For example, under the Internet IETF Integrated Servces (IntServ) archtecture, schedulng algorthms such as Weghted Far Queueng (WFQ), Vrtual Clock (VC) and Rate-Controlled Earlest Deadlne Frst (RC-EDF) [5, 8] were developed to support the Guaranteed Servce [] Furthermore, a sgnalng protocol, RSVP, for settng up end-toend QoS reservaton along a flow's path was also proposed and standardzed [4, 9] However, due to ts need for performng per-flow management at core routers, the scalablty of the IntServ archtecture has been questoned To address the ssue of scalablty, several alternatve archtectures have been proposed n recent years, among others, the IETF Dff- Serv model [2] and the more recent core stateless approach usng dynamc packet state [2, 3] In addressng the ssue of scalablty n QoS provsonng, the majorty of the recent works have focused on elmnatng per-flow router state management n the data plane For example, under the DffServ archtecture no per-flow QoS state s mantaned at core routers, and user flows are aggregated and processed based on a number of servce bts carred n the packet headers Consequently, only coarse-gran QoS support s provded to users In contrast, the core stateless approach amsatprovdng end-to-end per-flow guarantees wthout the complextyof per-flow QoS management To accomplsh ths goal, a novel schedulng mechansm the Core Jtter Vrtual Clock (CJVC) s developed n [2] to support end-to-end per-flow delay guarantees wthout requrng per-flow states n core routers Attempts at reducng the complexty of QoS control plane have mostly followed the conventonal hop-by-hop reservaton set-up approach adopted by RSVP through QoS control state aggregaton In the hop-by-hop reservaton set-up approach, each router mantans ts own QoS state database and admnsters a local admsson control test to determne whether a flow reservaton set-up request can be honored or not To ensure consstency among the QoS state databases mantaned by ndvdual routers, RSVP uses soft QoS states, whch requres perodc state exchange among routers, thus 7

2 ncurrng addtonal communcaton and processng overheads These overheads can be reduced through a number of control state reducton technques [6, 6, 7] Other hop-by-hop reservaton set-up protocols have also been proposed [8, 5] In general, the conventonal hop-by-hop reservaton set-up approach tes such QoS control functons as admsson control, resource reservaton and QoS state managementtocore routers, whether per-flow or aggregate QoS states are mantaned at core routers Therefore t requres admsson control and QoS state management modules to be nstalled at every router to support QoS provsonng In the context of the DffServ archtecture, an alternatve, and perhaps more attractve, approach the bandwdth broker approach s proposed n [7] to support the so-called Premum Servce In ths approach, admsson control, resource provsonng and other polcy decsons are performed by a centralzed bandwdth broker (BB) n each network doman Although several mplementaton efforts n buldng bandwdth brokers are under way (see, eg, [4]), so far many ssues regardng the desgn of bandwdth brokers, such as admsson control and QoS provsonng, have not been addressed adequately n the lterature For example, under the proposed BB archtecture n [7], t appears that only coarse-gran QoS provsonng can be supported and that explct confguraton of core routers s stll needed to provde QoS guarantees [4] Clearly, the delverable QoS performance wll hnge on how frequently such confguraton s performed In addton, t s not clear how admsson control should be performed under such a BB archtecture, n partcular, whether core routers are stll requred to perform local admsson control and QoS state management In ths paper we present a novel bandwdth broker archtecture for support of QoS provsonng that decouples the QoS control plane from the data forwardng plane Ths BB archtecture s desgned under the core stateless framework, usng the technque of dynamc packet state [3] Under our proposed BB archtecture, the QoS reservaton states are stored at and managed solely by the BB(s) n a network doman, and no or mnmal confguraton of core routers s requred In other words, core routers are completely releved of QoS control functons such as admsson control and state management (whether per-flow or aggregate), makng them potentally more effcent Ths decouplng of data plane and QoS control plane s appealng n several aspects Among others, t allows anetwork servce provder to ntroduce new QoS servces wthout necessarly requrng software/hardware upgrades at core routers Furthermore, as we wll demonstrate, t also enables the deployment of sophstcated QoS provsonng and admsson control algorthms to optmze network utlzaton n a network-wde fashon Such network-wde optmzaton s dffcult, f not mpossble, under the conventonal hop-by-hop reservaton set-up approach The objectve of ths paper s to demonstrate how guaranteed servces (both per-flow and class-based) can be supported usng the proposed BB archtecture, despte the fact that no QoS states are mantaned at core routers We focus on guaranteed servces partly because these servces are well-defned and understood, and partly because a unfyng core stateless framework the vrtual tme reference system [20] has been developed that provdes a QoS abstracton of the data plane for supportng guaranteed delay and rate servces Usng guaranteed servces as examples, we llustrate how admsson control can be performed usng the proposed BB archtecture wthout the assstance of core routers In partcular, we develop a path-orented approach to the desgn of effcent admsson control algorthms under the proposed BB archtecture We establsh that the proposed bandwdth broker archtecture s capable of supportng guaranteed servces wth the same granularty and expressve power (f not more) as the IntServ/Guaranteed Servce model, despte the fact that all QoS reservaton states are removed from core routers and mantaned solely at the bandwdth broker Ths s acheved wthout the potental complexty and scalablty problems of the IntServ model However, the bandwdth broker approach to QoS provsonng ntroduces a set of new ssues In partcular, the scalablty of the BB archtecture ts ablty to manage a large number of QoS control states and process a large volume of user flow QoS requests s an mportant ssue that must be nvestgated To partally address ths ssue, n ths paper we also consder the support of coarse-gran class-based guaranteed servces usng the proposed BB archtecture Va flow aggregaton, the number of QoS states mantaned by the BB can be reduced, and the complexty of admsson control operatons can be lowered, thereby enhancng the processng capacty of the BB However, n the context of guaranteed servces, (dynamc) flow aggregaton can havean undesrable transent effect that may result n delay bound volaton, f proper care s not taken We llustrate how ths problem can be solved usng relatvely smple mechansms under the proposed BB archtecture Our work s only a frst step towards addressng many problems that stll reman n the desgn and mplementaton of the proposed BB archtecture For example, to further mprove scalablty, a dstrbuted (or herarchcal) archtecture consstng of multple BBs may be necessary to support QoS provsonng n a large network doman Such an archtecture ntroduces many new desgn and mplementaton ssues The problem of nter-doman QoS reservaton and servcelevel agreement [2, 7] s another mportant ssue that must be addressed In addton, supportng statstcal or other forms of QoS guarantees usng the proposed BB archtecture s also a challengng problem Clearly, all these ssues must be satsfactorly resolved before the proposed BB archtecture can be deployed n practce The remander of ths paper s structured as follows Secton 2 outlnes the proposed bandwdth broker archtecture, where we also brefly revew the vrtual tme reference system developed n [20], and present a hgh-level overvew of the admsson control operatons under the proposed BB archtecture In Secton 3, we desgn effcent path-orented admsson control algorthms for per-flow guaranteed servces These admsson control algorthms are extended n Secton 4 to support class-based guaranteed servces wth dynamc flow aggregaton Smulaton nvestgaton s conducted n Secton 5, and the paper s concluded n Secton 6 Ths phenomenon s not unque to the core stateless framework, and may occur n any guaranteed servces wth dynamc flow aggregaton 72

3 Step : new flow servce request new flow arrval Edge condtoner Step 2: admsson control process Control plane BB S S 2 S S h Core router A network doman Data plane Step 3: decson (accept/reject) Management nformaton bases (MIBs) - topology nformaton base - polcy nformaton base - flow nformaton base - path QoS state nformaton base - node QoS state nformaton base Servce modules - admsson control module - QoS routng module - polcy control module Fgure : Illustraton of a bandwdth broker (BB) and ts operaton n a VTRS network doman 2 OVERVIEW OF THE PROPOSED BAND- WIDTH BROKER ARCHITECTURE In ths secton we outlne the proposed bandwdth broker archtecture for support of guaranteed servces wthn a sngle network doman Ths bandwdth broker archtecture s bult upon the vrtual tme reference system (VTRS) developed n [20], a core stateless framework that provdes a QoS abstracton of the data plane Under the proposed BB archtecture, core routers perform only data plane functons such aspacket schedulng and forwardng, usng the dynamc packet state carred n packet headers No QoS states, whether per-flow or aggregate, are mantaned at any core routers Instead, all QoS states are solely mantaned at and managed by the BB(s), no or mnmal confguraton of core routers s needed Furthermore, all QoS control plane functons such as admsson control and resource reservaton are performed by the BB(s), wthout nvolvement of core routers Fgure llustrates the basc components and operatons of the proposed BB archtecture as well as ts relatonshp to the data plane As shown n Fgure, a bandwdth broker 2 (BB) conssts of several servce modules (e, servers) such as admsson control, routng and polcy control The routng module peers wth routers to obtan the topology nformaton of the network doman [] and s responsble for path selecton and set-up (usng, eg, MPLS [0]) The polcy control module mantans a polcy nformaton base and s responsble for network polcy admnstraton The admsson control module mantans the QoS states of the network doman, whch are stored n several management nformaton bases (see Secton 22), and s responsble for admsson control and resource reservaton In ths paper we wll focus prmarly on the operatons of the admsson control module for support of guaranteed ser- 2 In ths paper we assume that there s a sngle centralzed BB for a network doman In practce, a dstrbuted or herarchcal archtecture consstng of multple BBs can be employed to mprove relablty and scalablty We wll explore these ssues n future research Note that an mportant advantage of the BB approach s that the relablty and scalablty ssues of the QoS control plane (e, the bandwdth broker archtecture) can be addressed separately from, and wthout ncurrng addtonal complexty to, the data plane vces In partcular, we wll demonstrate how effcent admsson control can be performed under the proposed BB archtecture Before we embark on the problem of admsson control usng the proposed BB archtecture, we need to understand the basc operatons of the vrtual tme reference system and ts mplcaton on QoS provsonng In the followng, we frst brefly revew the vrtual tme reference system, and then present a hgh-level descrpton of the admsson control module and the QoS state nformaton bases t mantans 2 The Vrtual Tme Reference System The vrtual tme reference system s desgned as a unfyng schedulng framework based on whch both the per-hop behavors of core routers (n terms of ther abltes to provde delay and bandwdth guarantees) and the end-to-end propertes of ther concatenaton can be characterzed Ths unfyng schedulng framework enables core routers to focus on packet schedulng and forwardng functons based on the packet state carred n packet headers wthout drect nvolvement n QoS control The key construct n the vrtual tme reference system s the noton of packet vrtual tme stamps, whch, as part of the packet state, are referenced and updated as packets traverse each core router Packet vrtual tme stamps are computed usng only the packet state carred by packets (plus a couple of fxed parameters assocated wth core routers), and thus the vrtual tme reference system s core stateless Conceptually, the vrtual tme reference system conssts of three logcal components: packet state carred by packets, edge traffc condtonng at the network edge, and per-hop vrtual tme reference/update mechansm at core routers (see Fgures 2 and 3) These three components are brefly descrbed below A more detaled descrpton can be found n [20] Edge Traffc Condtonng Edge traffc condtonng playsakey role n the VTRS, as t ensures that the packets of a flow 3 wll never be njected nto the network core at a rate exceedng ts reserved rate Formally, foraflow j wth a reserved rate r j,thenter-arrval tme of two consecutve packets of the flow at the frst hop core router s such that ^a j;k+ ^a j;k Lj;k+ r j, where ^a j;k denotes the arrval tme of the kth packet p j;k of flow j at the network core, L j;k the sze of packet p j;k, and r j the reserved rate of flow j Packet State After gong through the edge condtoner at the network edge, packets enterng the network core carry n ther packet headers certan packet state nformaton that s ntalzed and nserted at the network edge The packet state carred bypacket p j;k of a flow j contans three types of nformaton: ) a rate-delay parameter par hr j ;d j of the flow, determned by the bandwdth broker based on flow j's QoS requrement; 2) the vrtual tme stamp ~! j;k of the packet that s assocated wth the router currently beng traversed (t s ntalzed to ^a j;k, the actual tme t leaves the edge condtoner and enters the frst core router along the flow's path); and 3) the vrtual tme adjustment term 3 Here a flow can be ether an ndvdual user flow, or an aggregate traffc flow of multple user flows, defned n any approprate fashon 73

4 Core router Core Stateless Vrtual Tme Plane ~ j,k ~ j,k ~ j,k w v w + Edge condtoner Packet state Vrtual Fnsh Tme Computaton + ~ d j,k Core Stateless Vrtual Tme Update + Ψ + π,+ Schedulng Blackbox Real Tme Plane Network core a ^ j,k ^j,k f Fgure 2: A conceptual network model n the vrtual tme reference system Fgure 3: Vrtual tme reference system: per-hop behavor and operatons ff j;k of the packet, a parameter that s computed at the edge and used to ensure that the vrtual spacng property (see below) of vrtual tme stamps s satsfed [20] Vrtual Tme Reference/Update Mechansm and Per- Hop Router Behavor Characterzaton In the conceptual framework of the vrtual tme reference system, each core router s equpped wth a per-hop vrtual tme reference/update mechansm to mantan the contnual progresson of the vrtual tme emboded by the packet vrtual tme stamps Ths vrtual tme stamp ~! j;k represents the arrval tme of packet p j;k of flow j at the th core router n the vrtual tme, and thus t s also referred to as the vrtual arrval tme of the packet at the core router The vrtual tme stamps ~! j;k 's assocated wth packets of flow j satsfy the followng two mportant propertes: ) vrtual spacng property: ~! j;k+ ~! j;k Lj;k+ r j, and 2) the realty check property: ^a j;k» ~! j;k, where ^a j;k denotes the actual arrval tme of packet p j;k at router In order to ensure that these two propertes are satsfed, the vrtual tme stamps must be approprately referenced or updated as packets enter or leave a core router The referencng/updatng rule depends on the schedulng algorthm (or scheduler) employed by a core router We dstngush two types of schedulers: rate-based and delay-based, dependng on how thevrtual deadlne and vrtual fnsh tme are computed for packets traversng t For example, f the scheduler S at the th router s rate-based, packet p j;k s assocated wth the vrtual deadlne ~ d j;k fnsh tme s defned as ~ν j;k delay-based, ~ d j;k = d j and ~ν j;k = L j;k =r j +ff j;k and ts vrtual + d ~ j;k Whereas, f S s + d ~ j;k =~! j;k =~! j;k The per-hop behavor of a core router (or rather, ts scheduler) s characterzed by an error term, whch s defned wth respect to the vrtual fnsh tme and actual fnsh tme j;k of packets at the router Let ^f denote the actual tme packet p j;k departs the scheduler S We say that S can guarantee flow j ts reserved rate r j (f S s rate-based) or ts delay parameter d j (f S s delay-based) wth an error j;k term Ψ, f for any k, ^f» ~ν j;k +Ψ In other words, each packet of flow j s guaranteed to depart S by the tme ~ν j;k +Ψ =~! j;k + d ~ j;k +Ψ Gven the error term Ψ of the scheduler S, the vrtual tme stamp of packet p j;k after t has traversed S s updated usng the followng concatenaton rule: ~! j;k + = ~ν j;k +Ψ + ß =~! j;k + d ~ j;k +Ψ + ß () where ß denotes the propagaton delay from the th router to the next-hop router along the flow's path In [20] t s shown that usng the concatenaton rule () the vrtual spacng and realty check propertes of vrtual tme stamps are satsfed at every router End-to-End Delay Bound and QoS Abstracton of Data Plane Usng the vrtual tme reference system outlned above, the delay experenced by packets of a flow across the network core can be upper bounded n terms of the ratedelay parameter par of a flow and the error terms of the routers along the flow's path Suppose there are total h hops along the path of flow j, of whch q routers employ ratebased schedulers, and h q delay-based schedulers Then for each packet p j;k of flow j, wehave ^f h j;k ^aj;k» d j core = q Lj;max r j +(h q)d j + X (Ψ + ß); (2) where L j;max s the maxmum packet sze of flow j Suppose the traffc profle of flow j s specfed usng the standard dual-token bucket regulator (ff j ;ρ j ;P j ;L j;max ) where ff j L j;max s the maxmum burst sze of flow j, ρ j s the sustaned rate of flow j, P j s the peak rate of flow j Then the maxmum delay packets of flow j experenced at the edge shaper s bounded by d j edge = Ton j P j r j r j + Lj;max r j ; (3) where T j on =(ff j L j;max )=(P j ρ j ) Hence the end-to-end delay bound for flow j s gven by d j e2e = dj edge + dj core = Ton j P j r j r j X +(q+) Lj;max r j +(h q)d j + (Ψ + ß):(4) Observe that the end-to-end delay formula s qute smlar to that specfed n the IETF Guaranteed Servce usng the WFQ as the reference system In ths sense, the vrtual tme reference system provdes a conceptual core stateless 74

5 framework based on whch per-hop behavor (e, ts ablty to support delay guarantees) of a core router can be characterzed usng the noton of error term Ths smple abstracton enables us to derve end-to-end delay bounds for flows traversng an arbtrary concatenaton of core routers wthn a network doman Furthermore, the vrtual tme reference system does not mandate any specfc schedulng mechansms to be mplemented n a network doman as long as ther abltes to provde delay guarantees can be characterzed usng the noton of error term In fact, n [20] t s shown that almost all known schedulng algorthms can thus be characterzed, be they core stateless or stateful In addton, the vrtual tme reference system leads to the desgn of a set of new core stateless schedulng algorthms (both rate-based and delay-based) Two representatve examples of such core stateless schedulng algorthms are: the ratebased core stateless vrtual clock (C 6S VC) and delay-based vrtual tme earlest deadlne frst (VT-EDF) schedulng algorthms The core stateless vrtual clock(c 6S VC) s a work-conservng counterpart of the CJVC schedulng algorthm developed n [2] It servces packets n the order of ther vrtual fnsh tmes It s shown n [20] that as long as the total reserved rate of flows traversng P a C 6S VCscheduler does not exceed ts capacty (e, j rj» C), then the C 6S VC scheduler can guarantee each flow ts reserved rate r j wth the mnmum error term Ψ = L Λ;max =C, where L Λ;max s the largest packet sze among all flows traversng the C 6S VCscheduler Unlke the conventonal rate-controlled EDF, VT-EDF supports delay guarantees wthout per-flow rate control, and thus s core stateless Lke C 6S VC, VT-EDF also servces packets n the order of ther vrtual fnsh tmes The VT- EDF scheduler can guarantee each flow ts delay parameter d j wth the mnmum error term Ψ = L Λ;max =C, f the followng schedulablty condton s satsfed [20]: NX j= [r j (t d j )+L j;max ] ft d j g» Ct; for any t 0 (5) where we assume that there are N flows traversng the VT- EDF scheduler wth 0» d» d 2»»d N The ndcator functon ft d j g =ft d j, 0 otherwse 22 The Admsson Control Module: QoS State Informaton Bases and Basc Operatons In order to support guaranteed servces under the proposed bandwdth broker archtecture, the admsson control module mantans several QoS state nformaton bases They nclude: ) a flow nformaton base, where nformaton regardng ndvdual flows, such asflow d, traffc profle (eg, (ff j ;ρ j ;P j ;L j;max )), servce profle (eg, end-to-end delay requrement D j ), and QoS reservaton (eg, the rate-delay parameter par hr j ;d j ), s mantaned; 2) a node QoS state nformaton base, where nformaton regardng each router n the network doman, such as the bandwdth, buffer capacty, type of scheduler used (e, rate- or delay-based) and ts error term, and current QoS reservaton on each outgong lnk of the router, s mantaned ; and 3) a path QoS state nformaton base, where nformaton regardng the characterstcs of paths between ngress and egress routers s mantaned Examples of path characterstcs are the hop number of a path, the sum of the router error terms and propagaton delay along the path, and the mnmal resdual bandwdth along the path As we shall see n the next secton, mantanng separate path-level QoS state nformaton allows us to perform effcent admssblty test on an entre path bass We now brefly dscrbe the basc operatons of the BB, focusng, n partcular, on those of the admsson control module (see Fgure ) When a new flow wth a traffc profle (ff j ;ρ j ;P j ;L j;max ) and an end-to-end delay requrement D j;req arrves at an ngress router, the ngress router sends a new flow servce request message to the BB Upon recevng the servce request, the BB frst checks for polcy nformaton base to determne whether the new flow s admssble If not, the request s mmedately rejected Otherwse, the BB selects a path (from the ngress router to an approprate egress router n the network doman) for the new flow, and decde whether the flow can be admtted Generally speakng, the admsson control procedure conssts of two phases: ) admssblty test phase durng whch tsdeter- mned whether the new flow servce request can be accommodated and how much network resources must be reserved f t can be accommodated; and 2) bookkeepng phase durng whch the relevant management nformaton bases wll be updated, f the flow s admtted If the flow s admtted, the BB wll also pass (usng, eg, COPS [3]) the QoS reservaton nformaton such ashr j ;d j to the ngress router, so that t can set up a new or re-confgure an exstng edge condtoner (whch s assumed to be co-located at the ngress router) for the new flow As the packets of the flow arrve at the ngress router, the edge condtoner wll approprately ntalze and nsert these packet states, before njectng them nto the core of the network doman In the remander of ths paper we demonstrate how effcent admsson control can be performed under the proposed BB archtecture In Secton 3 we present a path-orented approach to perform effcent admsson control operatons for support of per-flow guaranteed servces Unlke the conventonal hop-by-hop approach whch performs admsson control ndvdually based on the local QoS state at each router along a path, ths path-orented approach examnes the resource constrants along the entre path smultaneously, and makes admsson control decson accordngly Clearly, such a path-orented approach s only possble because the avalablty of QoS state nformaton of the entre path at the BB Note that although any bandwdth broker archtecture can adopt a smlar path-orented approach toresource allocaton, under our BB archtecture no local admsson control test or explct resource (re-)confguraton s needed at any core router, thanks to the use of dynamc packet state n the data path As a result, we can sgnfcantly reduce the tme of conductng admsson control and resource reservaton Furthermore, we can also perform path-wde optmzaton when determnng resource allocaton for a new flow In Secton 4, we study the problem of admsson control for class-based guaranteed delay servces, where a fxed number of guaranteed delay servce classes are offered n a network doman As n the DffServ, there are two mportant benefts n provdng class-based servces under our framework Frst, aggregatng flows n a small number of servce classes greatly smplfes the mplementaton of the data plane (n partcular, the edge condtoners) Second, t also enhances 75

6 the scalablty of the QoS control plane (e, the BB archtecture) For example, the number of QoS states that need to be mantaned by a BB can be sgnfcantly reduced Ths leads to faster admsson control operatons, thereby ncreasng the number of flow servce requests a BB can handle However, as ponted out n the ntroducton, n class-based guaranteed servces certan transent behavor of dynamc flow aggregaton must be taken care of to avod potental delay bound volaton In Secton 4, we nvestgate the problem of dynamc flow aggregaton and devce smple mechansms to effectvely crcumvent ths problem 3 ADMISSION CONTROL FOR PER-FLOW GUARANTEED SERVICES In ths secton, we study the problem of admsson control for support of per-flow guaranteed servces under the proposed bandwdth broker archtecture In partcular, we present a path-orented approach to perform effcent admsson control operatons We llustrate how ths approach works by frst consderng a smple case, where only rate-based schedulers are employed along the path of a new flow We then look at the general case where the path of a new flow conssts of both rate-based and delay-based schedulers We present an effcent admsson control algorthm that ) determnes whether a new flow s admssble by examnng the resource constrants along the entre path smultaneously, and 2) mnmzes the bandwdth allocaton along the path for the new flow, f t s admssble 3 Path wth Only Rate-based Schedulers We frst consder the smple case where we assume that the path P for a new flow ν conssts of only rate-based schedulers Hence n ths case, we only need to determne whether a reserved rate r ν can be found for the new flow for t to be admtted The delay parameter d ν wll not be used For smplcty of exposton, we assume that a scheduler suchascore-stateless vrtual clock (C 6S VC) or corejtter vrtual clock (CJVC) s employed at the routers S along P Let F be the set of flows currently traversng S, and C be the total bandwdth at S Then as long as Pj2F rj» C, S can guarantee each flow j ts reserved rate r j We usec S res to denote the resdual bandwdth at P S,e,C S res = C j2f rj Let (ff ν ;ρ ν ;P ν ;L ν;max ) be the traffc profle of a new flow ν, and D ν;req be ts end-to-end delay requrement Let h be the number of hops n P, the path for the new flow From (4), n order to meet ts end-to-end delay requrement D ν;req, the reserved rate r ν for the new flow ν must satsfy: and ρ ν» r ν» P ν D ν;req Ton ν P ν r ν r ν +(h+) Lν;max r ν +Dtot; P (6) where T ν on =(ff ν L ν;max )=(P ν ρ ν ) and D P tot = P 2P (Ψ+ ß ) Furthermore, r ν must not exceed the mnmal resdual bandwdth Cres P along path P, where Cres P =mn 2P C S res s mantaned, as a path QoS parameter assocated wth P, n the path QoS state MIB Let r ν mn be the smallest r ν that satsfes (6), e, r ν mn = [T ν onp ν +(h +)L ν;max ]=[D ν;req D P tot + T ν on] Defne r low fea =maxfρ ν ;r ν mng and r up fea = mnfp ν ;C P resg: Then R Λ fea = [r low ] s the feasble rate range, from whch a feasble reserved rate r ν can be selected Clearly, f R Λ fea s empty, the servce request of the new flow ν must fea;rfea up be rejected Otherwse, t s admssble, and r ν = r low fea s the mnmal feasble reserved rate for the new flow ν Gven that the path QoS parameters D P tot and C P res assocated wth P are mantaned n the path QoS state MIB, the above admssblty test can be done n O() 32 Path wth Mxed Rate- and Delay-based Schedulers We now consder the general case where the path P for a new flow ν conssts of both rate-based and delay-based schedulers In ths case, we need to determne whether a rate-delay parameter par hr ν ;d ν can be found for the new flow ν for t to be admtted Because of the nter-dependence of the reserved rate r ν and the delay parameter d ν n the end-to-end delay bound (4) as well as the more complex schedulablty condton (5) for the delay-based schedulers, the admssblty test for ths case s less straghtforward In ths secton, we present an effcent admsson control algorthm usng the path-orented approach that mnmzes the bandwdth allocaton along the path of the new flow: f the new flow ν s admssble, ths admsson control algorthm fnds a feasble rate-delay parameter par hr ν ;d ν such that r ν s the mnmal feasble rate In other words, no other rate-delay parameter par hr ν0 ;d ν0 such that r ν0 < r ν s feasble Let q be the number of rate-based schedulers and h q the number of delay-based schedulers along path P For smplcty of exposton, we assume that the rate-based schedulers S along path P employ C 6S VC (or any smlar) schedulng algorthm whose schedulablty condton s P j2f rj» C, whereas the delay-based schedulers S employ the VT- EDF schedulng algorthm, whose schedulablty condton s gven n (5) As before, let (ff ν ;ρ ν ;P ν ;L ν;max ) be the traffc profle of the new flow ν, and D ν;req ts end-to-end delay requrement In order for the new flow ν to be admtted along the path P wth a rate-delay parameter par hr ν ;d ν, ts end-to-end delay requrement D ν;req must be satsfed, namely, and ρ ν» r ν» P ν D ν;req Ton ν P ν r ν r ν +(q+) Lν;max r ν +(h q)d ν +Dtot: P (7) Furthermore, the schedulablty condton at each scheduler S must not be volated Let Cres P be the mnmal resdual bandwdth along P, e, Cres P = mn 2P C S res Then from the schedulablty condtons for the rate- and delay-based schedulers, we must have r ν» Cres P In addton, for every delay-based scheduler S along P, let hr k ;d k be the ratedelay parameter par of flow k, where k 2F Then for each 76

7 0 t ν = h q Dtot P + T on ν ] Let m Λ such that d mλ <t ν» d mλ 2 for m = m Λ ;m Λ ;::: ;2; 3 R m fea ψ [rfea m;l fea ] 4 5 R m del ψ [rdel m;l del ] f (R m fea Rm del == ;) 6 7 f (R m fea == ;jjrm del == ;jjrm;r fea <rm;l del ) break wth d ν = d m 8 else /*R m fea R m del 6= ;*/ 9 f (rfea m;l del ) 0 r ν ψ rdel m;l ψ t ν Ξν rν break wth d ν 2 f (d ν >t ν ) no feasble value found 3 else return d ν nterval shrnkng at each teraton nterval movng towards left at each teraton R [ del ] [ fea ] m,l R m r del r r r d 0 = 0 d 2 dm- dm d dm *- dm* dm m m,r m,l m,r del fea fea d ν t ν (constant) rate d = M+ delay Fgure 4: Path-orented admssblty test for mxed rate/delay-based schedulers Fgure 5: The relatonshp between feasble range R m fea and delay constrant ranger m del X k 2F, S 2P such that d k d ν,wemust have [r j (d k dj )+Lj;max ] fj2f:d j»dk g +[r ν (d k dν )+L ν;max ]» Cd k : (8) Hence, n order for hr ν ;d ν to be a feasble rate-delay parameter par for the new flow ν, we must have that r ν 2 [ρ ν ; mnfp ν ;C P resg] and that r ν and d ν must satsfy (7) and (8) In the followng, we present an effcent algorthm to dentfy the feasble rate-delay parameter par hr ν ;d ν for a new flow ν Frst, we ntroduce some notaton Defne t ν = h q (Dν;req Dtot P +Ton), ν Ξ ν = h q [T onp ν ν +(q +)L ν;max ] From (7) we see that the followng condton holds, d ν» t ν Ξν r ν : (9) For any flow k traversng a delay-based scheduler S such that d k d ν, defne S k = C d k X fj2f:d j»dk g [r j (d k d j )+L j;max ]: Let F del be the unon of the sets of the flows at all the delaybased schedulers along the path P of flow ν, e, F del = [fj 2F : S s delay-basedg Suppose there are a total of M dstnctve delay parameters assocated wth the flows n F del Let these dstnctve M delay parameters be denoted by d ;d 2 ;::: ;d M, where 0» d < d 2 < < d M For m =; 2;::: ;M,defne S m = mnfs k : k 2F and d k = d m ; S s delay-basedg: Note that S k denotes the mnmum resdual servce over any tme nterval of length d k at scheduler S Therefore S m represents the mnmal resdual servce among all the delay-based schedulers at tme d m Hence we refer to S m as the mnmal resdual servce ofpath P at tme d m Defne d 0 = 0 and d M+ = Then f the new flow ν s admssble, there must exst a rate-delay parameter par hr ν ;d ν,whered ν 2 [d m ;d m ) for some m =; 2;::: ;M+ From (9), t s clear that 0» d ν» t ν Let m Λ be such that d mλ <t ν» d mλ Clearly, [d mλ ;d mλ ) s the rghtmost delay nterval that may contan a feasble d ν For m = m Λ ;m Λ ;::: ;2;, defne R m fea = [r m;l where fea ;rm;r fea ], r fea m;l Ξ = ν maxf t ν d m ;ρν g; rfea m;r Ξ = ν mnf t ν d m ;Pν ;Cres P g: (0) Smlarly, defne R m del =[r m;l r m;r del = mn ρ del ;rm;r del mn Ξν +L ν;max m»k<m Λf t ν d k ], where g; mn k m Λ S k Ξ ν L ν;max d k t ν r del m;l = max S k Ξ ν L ν;max m»k<m Λf d k t ν g () R m fea and R m del have the followng mportant monotoncty propertes, as llustrated n Fgure 5 When we movefrom the current delay nterval [d m ;d m ) to the next delay nterval to the left [d m 2 ;d m ), the correspondng feasble rate range R m fea determned by (0) shfts to the left In contrast, the correspondng feasble rate range R m del determned by thedelay constrants () shrnks: the left edge rdel m;l ncreases whle the rght edgerdel m;r decreases (For an explanaton why these propertes hold, see [2]) Based on ths observaton, we obtan the followng theorem, whch states whether a feasble rate-delay par hr ν ;d ν exsts such that d ν 2 [d m ;d m ] Theorem If R m fea R m del s empty, then no feasble rate-delay pars hr ν ;d ν exst such that d ν 2 [d m ;d m ] Furthermore, f R m fea s empty, or R m del s empty, or r m;r rdel m;l fea <, then no ntervals to the left contan a feasble soluton ether More precsely, no feasble rate-delay pars hr ν ;d ν exst such that d ν 2 [0;d m ) If R m fea R m del s not empty, then a feasble rate-delay par hr ν ;d ν exsts such that d ν 2 [d m ;d m ] Furthermore, f rfea m;l <rm;l del, then rν = rdel m;l s the smallest rate such that there exsts some d ν 0 for whch hr ν ;d ν safeasble ratedelay par In other words, any rate-delay par hr ν ;d ν where r ν <rdel m;l s not feasble Based on Theorem, the admsson control algorthm s presented (n pseudo-code) n Fgure 4 Note that the tme complexty of the algorthm s O(M), and n general, we have ff ; 77

8 M»jF del j» P S s delay-based jf j Hence the complexty of the algorthm hnges only on the number of dstnctvedelay parameters supported by the schedulers along the path of the new flow Ths reducton n complexty can be sgnfcant f many flows have the same delay requrements Ths s partcularly the case when we consder class-based admsson control wth flow aggregaton where a number of fxed delay classes are pre-defned Clearly ths reducton n complexty s acheved because our admsson control algorthm consders all the admssblty constrants along a path smultaneously Ths s not possble usng the conventonal hop-by-hop reservaton set-up approach (eg, as employed n the IETF IntServ model wth RSVP) 4 CLASS-BASED GUARANTEED SERVICES AND ADMISSION CONTROL WITH DY- NAMIC FLOW AGGREGATION In ths secton we address the problem of admsson control for class-based guaranteed servces under the proposed BB archtecture The class-based guaranteed delay servce model s schematcally shown n Fgure 6 A new user flow wll be placed n one of the delay servce classes f t can be admtted nto the network All flows n the same delay servce class that traverse the same path wll be aggregated nto a sngle macroflow Ths macroflow s shaped usng an aggregate reserved rate at the edge condtoner, and s guaranteed wth an end-to-end delay bound determned by the servce class We refer to the ndvdual user flows consttutng a macroflow as the mcroflows Akey ssue n the desgn of admsson control for ths classbased servce model s the problem of dynamc flow aggregaton The dynamcs comes from the fact that mcroflows may jon or leave amacroflow at any tme Hence the aggregate traffc profle for the macroflow may change dynamcally, and as a result, the reserved rate for the macroflow may need to be adjusted accordngly Dynamc flow aggregaton can cause certan undesrable effect on the end-to-end delay experenced by the macroflow, whch we wll descrbe shortly n Secton 4 Here we note that the exstng work on traffc aggregaton (n partcular, n the context of guaranteed servces, see, eg [6, 9]) has mplctly assumed statc flow aggregaton: a macroflow s an aggregaton of n fxed mcroflows, wth no new mcroflows jonng or exstng consttuent mcroflows leavng n the duraton of the macroflow As far as we areaware, ths problem of dynamc flow aggregaton has not been dentfed nor addressed before n the lterature 4 Impact of Dynamc Flow Aggregaton on End-to-End Delay In ths secton we llustrate the potental negatve mpact of dynamc flow aggregaton on end-to-end delay guarantee provsonng: f proper care s not taken, mcroflow arrvals or departures n an aggregate macroflow can cause potental delay bound volaton We use an example to llustrate ths problem of dynamc flow aggregaton Frst let us ntroduce some notaton and assumptons Consder a macroflow ff whch currently conssts of n mcroflows Let (ff j ;ρ j ;P j ;L j;max ) be the traffc profle of the mcroflow j,» j» n For smplcty, we wll use a dual-token bucket regulator, (ff ff ;ρ ff ;P ff ;L ff;max ), as the aggregate traffc profle for the macroflow ff Hence we have ff ff = P n j= ff j, ρ ff = P n j= ρj, P ff = P n j= P j,andl ff;max = P n j= Lj;max P Note that L ff;max n = j= Lj;max,asapacket of the maxmum sze may arrve from each of the n mcroflow atthe same tme In contrast, snce only one packet from the macroflow ff may leave the edge condtoner at any gven tme, the maxmum burst" the macroflow may carry nto the network core s max n j= L j;max Let P denote the path of the macroflow ff, and L P;max denote the maxmum packet sze permssble n a macroflow along P Then L P;max max n j= L j;max Wthout loss of generalty, we assume that L P;max s fxed Suppose that we treat the macroflow ff as statc, e, wth no mcroflows jonng or leavng at any tme Let hr ff ;d ff be the rate-delay parameter par reserved for the macroflow For smplcty, assume that path P conssts of only ratebased schedulers (h of them n total) Then from the endto-end delay formula (4), the end-to-end delay experenced by packets from the macroflow ff s bounded by d ff e2e = d ff edge +dff core = Ton ff P ff r ff r ff + Lff;max r ff +h LP;max r ff +Dtot P (2) P where Dtot P = 2P (Ψ + ß) Assume that a new mcroflow ν jons the exstng macroflow ff at tme t Λ Let (ff ν ;ρ ν ;P ν ;L ν;max ) be the traffc profle of the new mcroflow Denote the new" macroflow after the mcroflow ν has been aggregated by ff 0, and let (ff ff0 ;ρ ff0 ;P ff0 ;L ff0 ;max ) be ts traffc profle Suppose that the reserved rate for the new" macroflow ncreases from r ff to r ff0 at tme t Λ We frst show that the packets from the new" macroflow may experence a worst-case delay at the edge condtoner that s larger than d ff0 edge = Ton ff0 P ff0 r ff0 + Lff0 ;max Ths r ff0 r can happen, for example, n the scenaro shown n ff0 Fgure 7 In ths scenaro, Ton ff Ton, ν and thus Ton ν» Ton ff0» T ff We assume that all the consttuent mcroflows of the exstng macroflow ff start at the same tme (e, tme 0) and are greedy: they dump the maxmum allowed burst nto the network at any tme t, e, A ff (0;t) = E ff (t) = mnfp ff t + L ff;max ;ρ ff t + ff ff g The new mcroflow ν jons the exstng macroflow ff at tme t Λ = Ton ff Ton, ν and t s also greedy: at any tme t t Λ, A ν (t Λ ;t) = E ν (t t Λ ) = mnfp ν (t t Λ )+L ν;max ;ρ ν (t t Λ )+ff ν g Then t s not dffcult to see that at tme t = Ton, ff the total amount of traffc that s queued at the edge condtoner s gven by Q(t) =(P ff r ff )T ff on+(p ν +r ff r ff0 )T ν on+l ff0 ;max : Hence the delay experenced by a packet arrvng the edge condtoner at tme t = Ton ff wll be at least Q(t)=r ff0, whch can be shown to be larger than d ff0 edge n general Ths larger delay scausedby the fact that at the tme a new mcroflow s aggregated nto an exstng macroflow flow, the buffer at the edge condtoner may not be empty The old" packets queued there can cause the new" packets to experence addtonal delay that s no longer bounded by d ff0 edge We now consder the delay experenced by packets from the new" macroflow ff 0 nsde the network core Despte the 78

9 Arrval process ρ α+ ν mcroflows class macroflows scheduler scheduler α+ ν,max L P α+ ν α P Q statc Q dynamc r α+ ν class K Edge Condtoners Core Router Core Router L α,max r α One dynamc flow arrval pattern Statc flow aggregaton Q dynamc > Q statc 0 α T on ν T on ν Ton α T on Tme Fgure 6: Class-based guaranteed servces: dynamc flow aggregaton along a path Fgure 7: An example llustratng the edge delay bound volaton after rate change n dynamc flow aggregaton fact that packets from the new" macroflow ff 0 are servced wth a hgher reserved rate r ff0 ( r ff ), some of these packets may experence a worst-case delay n the network core that s bounded by d ff core = hl P;max =r ff + Dtot, P not by d ff0 core = hl P;max =r ff0 +Dtot P Intutvely, ths can happen because the packets from the new" macroflow may catch up wth the last packets from the old" macroflow Consderng both the delay at the edge condtoner and that n the network core, we see that packets of the new" macroflow may experence an end-to-end delay that s no longer bounded by the endto-end delay formula (2) A smlar stuaton may occur when a consttuent mcroflow leaves an exstng macroflow, f we mmedately decrease the reserved rate r ff to a new lower reserved rate r ff0 (see [2] for detals) Before we leave ths secton, we would lke to comment that the problem of dynamc flow aggregaton s not unque to the vrtual tme reference system used n ths paper The same problem exsts n a more general context For example, dynamc flow aggregaton wll have the same effect on a network of WFQ schedulers, the reference system used n the IntServ model Ths s because the stuaton happenng at the edge condtoner descrbed above wll also apply to a WFQ scheduler It appears that ths problem mght be more dffcult to address usng the conventonal hop-by-hop reservaton set-up approach In the followng we show how ths problem can be solved usng relatvely smple mechansms under the proposed BB archtecture 42 End-to-End Delay Bounds under Dynamc Flow Aggregaton In ths secton we present new mechansms to effectvely crcumvent the problems caused by dynamc flow aggregaton The basc objectve of our approach s to enable the bandwdth broker to make admsson control decsons at any gven tme, usng only the traffc profle and reserved rate of the macroflow at that tme In other words, we do not want the bandwdth broker to mantan an elaborate hstory record of the mcroflow arrval and departure events of a macroflow 42 Contngency Bandwdth and Edge Delay Bound We ntroduce the noton of contngency bandwdth to elmnate the lngerng delay effect of the backlog queued n the edge condtoner at the tme a mcroflow s aggregated nto or de-aggregated from an exstng macroflow It works as follows: Suppose at tme t Λ a mcroflow ν jons or leaves an exstng macroflow ff Besdes the reserved rate r ff beng adjusted to a new reserved rate r ff0 at t Λ, acontngency bandwdth r ν s also temporarly allocated to the resultng new" macroflow ff 0 for a contngency perod off ν tme unts The contngency bandwdth r ν and contngency perod f ν s chosen n such a manner that the maxmum delay n the edge condtoner experenced by any packets from the new" macroflow ff 0 after tme t Λ s bounded above by d new edge» maxfd old edge;d ff0 edgeg: (3) where d old edge denotes the maxmum edge delay bound on the old" macroflow (e, before t Λ ) and d ff0 edge = Ton(P ff0 ff0 r ff0 )=r ff0 + L ff0 ;max =r ff0 The followng two theorems state the suffcent condtons on r ν and f ν so that (3) holds (see [2] for ther proofs) Theorem 2 (Mcroflow Jon) Suppose at tme t Λ a new mcroflow ν wth the traffc profle (ff ν ;ρ ν ;P ν ;L ν;max ) jons an exstng macroflow ff Let r ν = r ff0 r ff and Q(t Λ ) be the sze of the backlog ntheedge condtoner at tme t Λ Then (3) holds f r ν P ν r ν and f ν Q(tΛ ) r ν : (4) Theorem 3 (Mcroflow Leave) Suppose at tme t Λ a mcroflow ν wth the traffc profle (ff ν ;ρ ν ;P ν ;L ν;max ) leaves an exstng macroflow ff Let r ν = r ff r ff0 and Q(t Λ ) be the sze of the backlog ntheedge condtoner at tme t Λ Then (3) holds f r ν r ν and f ν Q(tΛ ) r ν : (5) 79

10 To compute the contngency perod f ν precsely, weneedto know the backlog Q(t Λ ) n the edge condtoner at tme t Λ Snce at tme t Λ the maxmum delay at the edge condtoner s bounded by d old edge, wehave Q(t Λ )» d old edger(t Λ ) = d old edge(r ff + r ff (t Λ )) (6) where r(t Λ )stotal bandwdth allocated to the macroflow at tme t Λ,whch ncludes the reserved rate r ff and the total contngency bandwdth r ff (t Λ ) allocated to the macroflow ff at tme t Λ Gven ths upper bound on Q(t Λ ), the BB can determne an upper bound ^f ν on the contngency perod f ν as follows: ^f ν = d old r ff + r ff (t Λ ) edge r ν : (7) Hence after ^f ν, the BB can de-allocate the contngency bandwdth r ν at tme t Λ +^f ν We refer to ths method of determnng contngency perod f ν as the (theoretcal) contngency perod boundng approach In general ths scheme can be qute conservatve A more practcal approach s to have the edge condtoner to feedback the actual contngency perod to the BB based on the current buffer occupancy of the macroflow at the edge condtoner We refer to ths scheme as the contngency feedback method Note also that whenever the buffer n the edge condtoner becomes empty, the lngerng delay effect of the old macroflow has gone away Hence, the maxmum delay experenced by any packets of the macroflow ff s bounded by d ff edge, whch s solely determned by thecurrent aggregate traffc profle of the macroflow ff Therefore, the edge condtoner can send a message to the BB to reset all of the contngency bandwdth allocated to the macroflow ff (e, settng r ff = 0) before a contngency perod expres 422 Extenson to VTRS and Core Delay Bound As dscussed n Secton 4, packets of the new macroflow may catch up wth the packets of the old macroflow nsde the network core and hence experence addtonal queueng delay In ths secton we llustrate how the vrtual tme reference system can be extended to accommodate flow aggregaton wth dynamc rate changes Based on ths extenson, we present a modfed core-delay bound for flow aggregaton Consder an exstng macroflow ff whchtraverses the path P where there are q rate-based schedulers and h q delay-based schedulers In order to smplfy the dervaton of the delay experenced by packets nsde a network core, we mpose an assumpton: the delay parameter d ff assocated wth a macroflow ff s fxed, no matter whether there are mcroflow arrvals or departures n the macroflow Suppose that at tme f Λ, the reserved rate of the macroflow ff s adjusted at the edge shaper from r to r 0 Let p kλ be the last packet that leaves the edge condtoner before the rate change at f Λ, and p kλ + be the frst packet that leaves the edge condtoner after the rate change at f Λ Then the packets are shaped as follows For k < k Λ, ^a k+ ^a k L k+ =r, and for k k Λ,^a k+ ^a k L k+ =r 0 Furthermore, we need to ensure that the vrtual tme adjustment term that s carred nsde packet headers s properly calculated at the network edge Due to space lmtaton, we omt the detals here, and refer the nterested reader to [2] The followng theorem states a modfed delay bound nsde the network core for packets of a macroflow Theorem 4 Let packets of one macroflow shaped at the network edge as follows For k<k Λ, ^a k+ ^a k L k+ =r, and for k k Λ, ^a k+ ^a k L k+ =r 0 Moreover, let the vrtual tme adjustment term be properly defned as n [2] Then the vrtual spacng and realty check propertes hold for the macroflow after the rate change at f Λ Furthermore, the delay experenced by these packets n the network core s bounded by the followng modfed core delay formula: ^f h k ^a k» ρ L P;max q max ; ff+(h q)d LP;max ff +D r r tot: P (8) 0 43 Admsson Control wth Dynamc Flow Aggregaton We now llustrate how to perform admsson control and resource reservaton wth dynamc flow aggregaton, based on the results obtaned n Secton 42 and Secton 422 Consder a macroflow ff Let D ff;req be ts end-to-end delay requrement, whch we assume s fxed throughout the entre duraton of the macroflow Whenever a mcroflow jons or leaves the macroflow ff, we need to ensure that ts end-to-end delay requrement s stll satsfed At a gven tme, let r ff be the reserved rate of macroflow ff excludng the contngency bandwdth allocated Let P be the path of macroflow ff, and let C P res be the mnmal resdual bandwdth along path P We consder the cases of mcroflow jon and leave separately below Mcroflow Jon Consder a new mcroflow ν wantng to jon the exstng macroflow ff at tme t Λ If the new mcroflow can be admtted, we need to determne, for the resultng new" macroflow ff 0, a new reserved rate r ff0 r ff as well as r ν amount of new contngency bandwdth for a contngency perod of f ν From Theorem 2, wthout loss of generalty, wechoose r ν = P ν r ff0 + r ff Hence n order to be able to admt the new mcroflow ν nto the exstng macroflow ff, we must have P ν» Cres P If ths condton s satsfed, then we need to fnd the mnmal new reserved rate r ff0 so that the end-to-end delay requrement D ff;req can be satsfed for the resultng macroflow ff 0 Note that after contngency perod, the edge queueng delay for any packets of the class s determned by the new class traffc profle and the reserved rate, therefore, d ff0 e2e = d ff0 edge +maxfd ff core;d ff0 coreg»d ff;req : (9) Snce r ff0 r ff, L P;max =r ff0» L P;max =r ff Hence d ff core» d ff0 core The constrant (9) s reduced to d ff0 edge» D ff;req d ff core Hence the new mcroflow can be admtted f the new reserved rate r ff0 can be accommodated along path P (e, f ρ ν» r ff0 r ff» P ν» Cres) P If the mcroflow can be admtted, r ff + P ν s allocated to the macroflow durng the contngency perod (e, from t Λ to t Λ + f ν ), and after t Λ + f ν, only r ff0 wll be allocated for macroflow ff 0 80

PAPER Providing Scalable Support for Multiple QoS Guarantees: Architecture and Mechanisms

PAPER Providing Scalable Support for Multiple QoS Guarantees: Architecture and Mechanisms 2830 IEICE TRANS. COMMUN., VOL.E84 B, NO.10 OCTOBER 2001 PAPER Provdng Scalable Support for Multple QoS Guarantees: Archtecture and Mechansms Ywe Thomas HOU, Regular Member, Zhenha DUAN, Zh-L ZHANG, Nonmembers,

More information

Goals and Approach Type of Resources Allocation Models Shared Non-shared Not in this Lecture In this Lecture

Goals and Approach Type of Resources Allocation Models Shared Non-shared Not in this Lecture In this Lecture Goals and Approach CS 194: Dstrbuted Systems Resource Allocaton Goal: acheve predcable performances Three steps: 1) Estmate applcaton s resource needs (not n ths lecture) 2) Admsson control 3) Resource

More information

Real-Time Guarantees. Traffic Characteristics. Flow Control

Real-Time Guarantees. Traffic Characteristics. Flow Control Real-Tme Guarantees Requrements on RT communcaton protocols: delay (response s) small jtter small throughput hgh error detecton at recever (and sender) small error detecton latency no thrashng under peak

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

More information

Avoiding congestion through dynamic load control

Avoiding congestion through dynamic load control Avodng congeston through dynamc load control Vasl Hnatyshn, Adarshpal S. Seth Department of Computer and Informaton Scences, Unversty of Delaware, Newark, DE 976 ABSTRACT The current best effort approach

More information

State of the Art in Differentiated

State of the Art in Differentiated Outlne Dfferentated Servces on the Internet Explct Allocaton of Best Effort Packet Delvery Servce, D. Clark and W. Fang A Two bt Dfferentated Servces Archtecture for the Internet, K. Nchols, V. Jacobson,

More information

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 A mathematcal programmng approach to the analyss, desgn and

More information

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

Simulation Based Analysis of FAST TCP using OMNET++

Simulation Based Analysis of FAST TCP using OMNET++ Smulaton Based Analyss of FAST TCP usng OMNET++ Umar ul Hassan 04030038@lums.edu.pk Md Term Report CS678 Topcs n Internet Research Sprng, 2006 Introducton Internet traffc s doublng roughly every 3 months

More information

Efficient Distributed File System (EDFS)

Efficient Distributed File System (EDFS) Effcent Dstrbuted Fle System (EDFS) (Sem-Centralzed) Debessay(Debsh) Fesehaye, Rahul Malk & Klara Naherstedt Unversty of Illnos-Urbana Champagn Contents Problem Statement, Related Work, EDFS Desgn Rate

More information

Advances in Internet Quality of Service

Advances in Internet Quality of Service HTTP://DSCWWW.EPFL.CH/EN/PUBLICATIONS/DOCUMENTS/TR00 049.PDF 1 Advances n Internet Qualty of Servce Vctor Frou, Jean-Yves Le Boudec, Don Towsley, and Zh-L Zhang Abstract We descrbe recent advances n theores

More information

CS 268: Lecture 8 Router Support for Congestion Control

CS 268: Lecture 8 Router Support for Congestion Control CS 268: Lecture 8 Router Support for Congeston Control Ion Stoca Computer Scence Dvson Department of Electrcal Engneerng and Computer Scences Unversty of Calforna, Berkeley Berkeley, CA 9472-1776 Router

More information

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

More information

Load-Balanced Anycast Routing

Load-Balanced Anycast Routing Load-Balanced Anycast Routng Chng-Yu Ln, Jung-Hua Lo, and Sy-Yen Kuo Department of Electrcal Engneerng atonal Tawan Unversty, Tape, Tawan sykuo@cc.ee.ntu.edu.tw Abstract For fault-tolerance and load-balance

More information

Enhancing Class-Based Service Architectures with Adaptive Rate Allocation and Dropping Mechanisms

Enhancing Class-Based Service Architectures with Adaptive Rate Allocation and Dropping Mechanisms Enhancng Class-Based Servce Archtectures wth Adaptve Rate Allocaton and Droppng Mechansms Ncolas Chrstn, Member, IEEE, Jörg Lebeherr, Senor Member, IEEE, and Tarek Abdelzaher, Member, IEEE Abstract Class-based

More information

AADL : about scheduling analysis

AADL : about scheduling analysis AADL : about schedulng analyss Schedulng analyss, what s t? Embedded real-tme crtcal systems have temporal constrants to meet (e.g. deadlne). Many systems are bult wth operatng systems provdng multtaskng

More information

Abstract. Introduction. B. Integrated Services and RSVP. A. Classical Admission Control

Abstract. Introduction. B. Integrated Services and RSVP. A. Classical Admission Control A SURVEY ON ADMISSION-CONTROL SCHEMES AND SCHEDULING ALGORITHMS Masaru Ouda, Murray State Unversty Abstract There has been a sustaned nterest among researchers and networ operators n provdng qualty of

More information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

Analysis of Collaborative Distributed Admission Control in x Networks

Analysis of Collaborative Distributed Admission Control in x Networks 1 Analyss of Collaboratve Dstrbuted Admsson Control n 82.11x Networks Thnh Nguyen, Member, IEEE, Ken Nguyen, Member, IEEE, Lnha He, Member, IEEE, Abstract Wth the recent surge of wreless home networks,

More information

Pricing Network Resources for Adaptive Applications in a Differentiated Services Network

Pricing Network Resources for Adaptive Applications in a Differentiated Services Network IEEE INFOCOM Prcng Network Resources for Adaptve Applcatons n a Dfferentated Servces Network Xn Wang and Hennng Schulzrnne Columba Unversty Emal: {xnwang, schulzrnne}@cs.columba.edu Abstract The Dfferentated

More information

Internet Traffic Managers

Internet Traffic Managers Internet Traffc Managers Ibrahm Matta matta@cs.bu.edu www.cs.bu.edu/faculty/matta Computer Scence Department Boston Unversty Boston, MA 225 Jont work wth members of the WING group: Azer Bestavros, John

More information

THE ability to provide end-to-end guaranteed services

THE ability to provide end-to-end guaranteed services IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 15, NO. 2, FEBRUARY 2004 167 A Core Stateless Bandwidth Broker Architecture for Scalable Support of Guaranteed Services Zhenhai Duan, Zhi-Li

More information

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION 24 CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION The present chapter proposes an IPSO approach for multprocessor task schedulng problem wth two classfcatons, namely, statc ndependent tasks and

More information

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory Background EECS. Operatng System Fundamentals No. Vrtual Memory Prof. Hu Jang Department of Electrcal Engneerng and Computer Scence, York Unversty Memory-management methods normally requres the entre process

More information

Channel 0. Channel 1 Channel 2. Channel 3 Channel 4. Channel 5 Channel 6 Channel 7

Channel 0. Channel 1 Channel 2. Channel 3 Channel 4. Channel 5 Channel 6 Channel 7 Optmzed Regonal Cachng for On-Demand Data Delvery Derek L. Eager Mchael C. Ferrs Mary K. Vernon Unversty of Saskatchewan Unversty of Wsconsn Madson Saskatoon, SK Canada S7N 5A9 Madson, WI 5376 eager@cs.usask.ca

More information

Routing in Degree-constrained FSO Mesh Networks

Routing in Degree-constrained FSO Mesh Networks Internatonal Journal of Hybrd Informaton Technology Vol., No., Aprl, 009 Routng n Degree-constraned FSO Mesh Networks Zpng Hu, Pramode Verma, and James Sluss Jr. School of Electrcal & Computer Engneerng

More information

A fair buffer allocation scheme

A fair buffer allocation scheme A far buffer allocaton scheme Juha Henanen and Kalev Klkk Telecom Fnland P.O. Box 228, SF-330 Tampere, Fnland E-mal: juha.henanen@tele.f Abstract An approprate servce for data traffc n ATM networks requres

More information

Chapter 6 Programmng the fnte element method Inow turn to the man subject of ths book: The mplementaton of the fnte element algorthm n computer programs. In order to make my dscusson as straghtforward

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

Scheduling and queue management. DigiComm II

Scheduling and queue management. DigiComm II Schedulng and queue management Tradtonal queung behavour n routers Data transfer: datagrams: ndvdual packets no recognton of flows connectonless: no sgnallng Forwardng: based on per-datagram forwardng

More information

Solutions for Real-Time Communication over Best-Effort Networks

Solutions for Real-Time Communication over Best-Effort Networks Solutons for Real-Tme Communcaton over Best-Effort Networks Anca Hangan, Ramona Marfevc, Gheorghe Sebestyen Techncal Unversty of Cluj-Napoca, Computer Scence Department {Anca.Hangan, Ramona.Marfevc, Gheorghe.Sebestyen}@cs.utcluj.ro

More information

A Hybrid Genetic Algorithm for Routing Optimization in IP Networks Utilizing Bandwidth and Delay Metrics

A Hybrid Genetic Algorithm for Routing Optimization in IP Networks Utilizing Bandwidth and Delay Metrics A Hybrd Genetc Algorthm for Routng Optmzaton n IP Networks Utlzng Bandwdth and Delay Metrcs Anton Redl Insttute of Communcaton Networks, Munch Unversty of Technology, Arcsstr. 21, 80290 Munch, Germany

More information

The Codesign Challenge

The Codesign Challenge ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.

More information

Statistical Admission Control Using Delay Distribution Measurements

Statistical Admission Control Using Delay Distribution Measurements Statstcal Admsson Control Usng Delay Dstrbuton Measurements KARTIK GOPALAN State Unversty of New York at Bnghamton LAN HUANG IBM Almaden Research Center GANG PENG and TZI-CKER CHIUEH Stony Brook Unversty

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

More information

TECHNICAL REPORT AN OPTIMAL DISTRIBUTED PROTOCOL FOR FAST CONVERGENCE TO MAXMIN RATE ALLOCATION. Jordi Ros and Wei K Tsai

TECHNICAL REPORT AN OPTIMAL DISTRIBUTED PROTOCOL FOR FAST CONVERGENCE TO MAXMIN RATE ALLOCATION. Jordi Ros and Wei K Tsai TECHNICAL REPORT AN OPTIMAL DISTRIUTED PROTOCOL FOR FAST CONVERGENCE TO MAXMIN RATE ALLOCATION Jord Ros and We K Tsa Department of Electrcal and Computer Engneerng Unversty of Calforna, Irvne 1 AN OPTIMAL

More information

DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT

DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT Bran J. Wolf, Joseph L. Hammond, and Harlan B. Russell Dept. of Electrcal and Computer Engneerng, Clemson Unversty,

More information

Virtual Machine Migration based on Trust Measurement of Computer Node

Virtual Machine Migration based on Trust Measurement of Computer Node Appled Mechancs and Materals Onlne: 2014-04-04 ISSN: 1662-7482, Vols. 536-537, pp 678-682 do:10.4028/www.scentfc.net/amm.536-537.678 2014 Trans Tech Publcatons, Swtzerland Vrtual Machne Mgraton based on

More information

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

More information

ARTICLE IN PRESS. Signal Processing: Image Communication

ARTICLE IN PRESS. Signal Processing: Image Communication Sgnal Processng: Image Communcaton 23 (2008) 754 768 Contents lsts avalable at ScenceDrect Sgnal Processng: Image Communcaton journal homepage: www.elsever.com/locate/mage Dstrbuted meda rate allocaton

More information

Technical Report. i-game: An Implicit GTS Allocation Mechanism in IEEE for Time- Sensitive Wireless Sensor Networks

Technical Report. i-game: An Implicit GTS Allocation Mechanism in IEEE for Time- Sensitive Wireless Sensor Networks www.hurray.sep.pp.pt Techncal Report -GAME: An Implct GTS Allocaton Mechansm n IEEE 802.15.4 for Tme- Senstve Wreless Sensor etworks Ans Koubaa Máro Alves Eduardo Tovar TR-060706 Verson: 1.0 Date: Jul

More information

On the Fairness-Efficiency Tradeoff for Packet Processing with Multiple Resources

On the Fairness-Efficiency Tradeoff for Packet Processing with Multiple Resources On the Farness-Effcency Tradeoff for Packet Processng wth Multple Resources We Wang, Chen Feng, Baochun L, and Ben Lang Department of Electrcal and Computer Engneerng, Unversty of Toronto {wewang, cfeng,

More information

Hierarchical clustering for gene expression data analysis

Hierarchical clustering for gene expression data analysis Herarchcal clusterng for gene expresson data analyss Gorgo Valentn e-mal: valentn@ds.unm.t Clusterng of Mcroarray Data. Clusterng of gene expresson profles (rows) => dscovery of co-regulated and functonally

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Response-Time Guarantees in ATM Networks

Response-Time Guarantees in ATM Networks Response-Tme Guarantees n ATM Networks Andreas Ermedahl Hans Hansson Mkael Sjödn Department of Computer Systems Uppsala Unversty Sweden E-mal: febbe,hansh,mcg@docs.uu.se Abstract We present a method for

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

More information

Verification by testing

Verification by testing Real-Tme Systems Specfcaton Implementaton System models Executon-tme analyss Verfcaton Verfcaton by testng Dad? How do they know how much weght a brdge can handle? They drve bgger and bgger trucks over

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors

We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors We are IntechOpen, the world s leadng publsher of Open Access books Bult by scentsts, for scentsts 3,500 108,000 1.7 M Open access books avalable Internatonal authors and edtors Downloads Our authors are

More information

Dynamic Bandwidth Provisioning with Fairness and Revenue Considerations for Broadband Wireless Communication

Dynamic Bandwidth Provisioning with Fairness and Revenue Considerations for Broadband Wireless Communication Ths full text paper was peer revewed at the drecton of IEEE Communcatons Socety subject matter experts for publcaton n the ICC 008 proceedngs. Dynamc Bandwdth Provsonng wth Farness and Revenue Consderatons

More information

Problem Set 3 Solutions

Problem Set 3 Solutions Introducton to Algorthms October 4, 2002 Massachusetts Insttute of Technology 6046J/18410J Professors Erk Demane and Shaf Goldwasser Handout 14 Problem Set 3 Solutons (Exercses were not to be turned n,

More information

Needed Information to do Allocation

Needed Information to do Allocation Complexty n the Database Allocaton Desgn Must tae relatonshp between fragments nto account Cost of ntegrty enforcements Constrants on response-tme, storage, and processng capablty Needed Informaton to

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

On Achieving Fairness in the Joint Allocation of Buffer and Bandwidth Resources: Principles and Algorithms

On Achieving Fairness in the Joint Allocation of Buffer and Bandwidth Resources: Principles and Algorithms On Achevng Farness n the Jont Allocaton of Buffer and Bandwdth Resources: Prncples and Algorthms Yunka Zhou and Harsh Sethu (correspondng author) Abstract Farness n network traffc management can mprove

More information

An Investigation into Server Parameter Selection for Hierarchical Fixed Priority Pre-emptive Systems

An Investigation into Server Parameter Selection for Hierarchical Fixed Priority Pre-emptive Systems An Investgaton nto Server Parameter Selecton for Herarchcal Fxed Prorty Pre-emptve Systems R.I. Davs and A. Burns Real-Tme Systems Research Group, Department of omputer Scence, Unversty of York, YO10 5DD,

More information

Maintaining temporal validity of real-time data on non-continuously executing resources

Maintaining temporal validity of real-time data on non-continuously executing resources Mantanng temporal valdty of real-tme data on non-contnuously executng resources Tan Ba, Hong Lu and Juan Yang Hunan Insttute of Scence and Technology, College of Computer Scence, 44, Yueyang, Chna Wuhan

More information

Conditional Speculative Decimal Addition*

Conditional Speculative Decimal Addition* Condtonal Speculatve Decmal Addton Alvaro Vazquez and Elsardo Antelo Dep. of Electronc and Computer Engneerng Unv. of Santago de Compostela, Span Ths work was supported n part by Xunta de Galca under grant

More information

Priority-Based Scheduling Algorithm for Downlink Traffics in IEEE Networks

Priority-Based Scheduling Algorithm for Downlink Traffics in IEEE Networks Prorty-Based Schedulng Algorthm for Downlnk Traffcs n IEEE 80.6 Networks Ja-Mng Lang, Jen-Jee Chen, You-Chun Wang, Yu-Chee Tseng, and Bao-Shuh P. Ln Department of Computer Scence Natonal Chao-Tung Unversty,

More information

Efficient Load-Balanced IP Routing Scheme Based on Shortest Paths in Hose Model. Eiji Oki May 28, 2009 The University of Electro-Communications

Efficient Load-Balanced IP Routing Scheme Based on Shortest Paths in Hose Model. Eiji Oki May 28, 2009 The University of Electro-Communications Effcent Loa-Balance IP Routng Scheme Base on Shortest Paths n Hose Moel E Ok May 28, 2009 The Unversty of Electro-Communcatons Ok Lab. Semnar, May 28, 2009 1 Outlne Backgroun on IP routng IP routng strategy

More information

Lecture 7 Real Time Task Scheduling. Forrest Brewer

Lecture 7 Real Time Task Scheduling. Forrest Brewer Lecture 7 Real Tme Task Schedulng Forrest Brewer Real Tme ANSI defnes real tme as A Real tme process s a process whch delvers the results of processng n a gven tme span A data may requre processng at a

More information

Evaluation of an Enhanced Scheme for High-level Nested Network Mobility

Evaluation of an Enhanced Scheme for High-level Nested Network Mobility IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.15 No.10, October 2015 1 Evaluaton of an Enhanced Scheme for Hgh-level Nested Network Moblty Mohammed Babker Al Mohammed, Asha Hassan.

More information

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed

More information

RAP. Speed/RAP/CODA. Real-time Systems. Modeling the sensor networks. Real-time Systems. Modeling the sensor networks. Real-time systems:

RAP. Speed/RAP/CODA. Real-time Systems. Modeling the sensor networks. Real-time Systems. Modeling the sensor networks. Real-time systems: Speed/RAP/CODA Presented by Octav Chpara Real-tme Systems Many wreless sensor network applcatons requre real-tme support Survellance and trackng Border patrol Fre fghtng Real-tme systems: Hard real-tme:

More information

5 The Primal-Dual Method

5 The Primal-Dual Method 5 The Prmal-Dual Method Orgnally desgned as a method for solvng lnear programs, where t reduces weghted optmzaton problems to smpler combnatoral ones, the prmal-dual method (PDM) has receved much attenton

More information

Module Management Tool in Software Development Organizations

Module Management Tool in Software Development Organizations Journal of Computer Scence (5): 8-, 7 ISSN 59-66 7 Scence Publcatons Management Tool n Software Development Organzatons Ahmad A. Al-Rababah and Mohammad A. Al-Rababah Faculty of IT, Al-Ahlyyah Amman Unversty,

More information

Enhanced Signaling Scheme with Admission Control in the Hybrid Optical Wireless (HOW) Networks

Enhanced Signaling Scheme with Admission Control in the Hybrid Optical Wireless (HOW) Networks Enhanced Sgnalng Scheme wth Admsson Control n the Hybrd Optcal Wreless (HOW) Networks Yng Yan, Hao Yu, Henrk Wessng, and Lars Dttmann Department of Photoncs Techncal Unversty of Denmark Lyngby, Denmark

More information

ATM Switch. Traffic Shaper. ATM Switch. Traffic Shaper. ATM Switch. Terminal Equipment. Terminal Equipment UNI UNI

ATM Switch. Traffic Shaper. ATM Switch. Traffic Shaper. ATM Switch. Terminal Equipment. Terminal Equipment UNI UNI Mult-rate Trac Shapng and End-to-End Performance Guarantees n ATM Debanjan Saha Department of Computer Scence Unversty of Maryland College Park, MD 20742 E-mal: debanjan@cs.umd.edu Sart Mukherjee y Dept.

More information

Video Proxy System for a Large-scale VOD System (DINA)

Video Proxy System for a Large-scale VOD System (DINA) Vdeo Proxy System for a Large-scale VOD System (DINA) KWUN-CHUNG CHAN #, KWOK-WAI CHEUNG *# #Department of Informaton Engneerng *Centre of Innovaton and Technology The Chnese Unversty of Hong Kong SHATIN,

More information

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints TPL-ware Dsplacement-drven Detaled Placement Refnement wth Colorng Constrants Tao Ln Iowa State Unversty tln@astate.edu Chrs Chu Iowa State Unversty cnchu@astate.edu BSTRCT To mnmze the effect of process

More information

Delay Variation Optimized Traffic Allocation Based on Network Calculus for Multi-path Routing in Wireless Mesh Networks

Delay Variation Optimized Traffic Allocation Based on Network Calculus for Multi-path Routing in Wireless Mesh Networks Appl. Math. Inf. Sc. 7, No. 2L, 467-474 2013) 467 Appled Mathematcs & Informaton Scences An Internatonal Journal http://dx.do.org/10.12785/ams/072l13 Delay Varaton Optmzed Traffc Allocaton Based on Network

More information

A New Token Allocation Algorithm for TCP Traffic in Diffserv Network

A New Token Allocation Algorithm for TCP Traffic in Diffserv Network A New Token Allocaton Algorthm for TCP Traffc n Dffserv Network A New Token Allocaton Algorthm for TCP Traffc n Dffserv Network S. Sudha and N. Ammasagounden Natonal Insttute of Technology, Truchrappall,

More information

Real-Time Systems. Real-Time Systems. Verification by testing. Verification by testing

Real-Time Systems. Real-Time Systems. Verification by testing. Verification by testing EDA222/DIT161 Real-Tme Systems, Chalmers/GU, 2014/2015 Lecture #8 Real-Tme Systems Real-Tme Systems Lecture #8 Specfcaton Professor Jan Jonsson Implementaton System models Executon-tme analyss Department

More information

Use of Genetic Algorithms in Efficient Scheduling for Multi Service Classes

Use of Genetic Algorithms in Efficient Scheduling for Multi Service Classes Use of Genetc Algorthms n Effcent Schedulng for Mult Servce Classes Shyamale Thlakawardana and Rahm Tafazoll Centre for Communcatons Systems Research (CCSR), Unversty of Surrey, Guldford, GU27XH, UK Abstract

More information

A Sub-Critical Deficit Round-Robin Scheduler

A Sub-Critical Deficit Round-Robin Scheduler A Sub-Crtcal Defct ound-obn Scheduler Anton Kos, Sašo Tomažč Unversty of Ljubljana, Faculty of Electrcal Engneerng, Ljubljana, Slovena E-mal: anton.kos@fe.un-lj.s Abstract - A scheduler s an essental element

More information

Online Packet Scheduling with Hard Deadlines in Wired Networks

Online Packet Scheduling with Hard Deadlines in Wired Networks Onlne Packet Schedulng wth Hard Deadlnen Wred Networks Zhouja Mao, C. Emre Koksal, Ness B. Shroff E-mal: maoz@ece.osu.edu, koksal@ece.osu.edu, shroff@ece.osu.edu Abstract The problem of onlne job or packet

More information

Adaptive Resource Allocation Control with On-Line Search for Fair QoS Level

Adaptive Resource Allocation Control with On-Line Search for Fair QoS Level Adaptve Resource Allocaton Control wth On-Lne Search for Far QoS Level Fumko Harada, Toshmtsu Usho, Graduate School of Engneerng Scence Osaka Unversty {harada@hopf, usho@}sysesosaka-uacjp Yukkazu akamoto

More information

Fast Retransmission of Real-Time Traffic in HIPERLAN/2 Systems

Fast Retransmission of Real-Time Traffic in HIPERLAN/2 Systems Fast Retransmsson of Real-Tme Traffc n HIPERLAN/ Systems José A Afonso and Joaqum E Neves Department of Industral Electroncs Unversty of Mnho, Campus de Azurém 4800-058 Gumarães, Portugal {joseafonso,

More information

Intelligent Traffic Conditioners for Assured Forwarding Based Differentiated Services Networks 1

Intelligent Traffic Conditioners for Assured Forwarding Based Differentiated Services Networks 1 Intellgent Traffc Condtoners for Assured Forwardng Based Dfferentated Servces Networks B. Nandy, N. Seddgh, P. Peda, J. Ethrdge Nortel Networks, Ottawa, Canada Emal:{bnandy, nseddgh, ppeda, jethrdg}@nortelnetworks.com

More information

arxiv: v3 [cs.ds] 7 Feb 2017

arxiv: v3 [cs.ds] 7 Feb 2017 : A Two-stage Sketch for Data Streams Tong Yang 1, Lngtong Lu 2, Ybo Yan 1, Muhammad Shahzad 3, Yulong Shen 2 Xaomng L 1, Bn Cu 1, Gaogang Xe 4 1 Pekng Unversty, Chna. 2 Xdan Unversty, Chna. 3 North Carolna

More information

Distributed Resource Scheduling in Grid Computing Using Fuzzy Approach

Distributed Resource Scheduling in Grid Computing Using Fuzzy Approach Dstrbuted Resource Schedulng n Grd Computng Usng Fuzzy Approach Shahram Amn, Mohammad Ahmad Computer Engneerng Department Islamc Azad Unversty branch Mahallat, Iran Islamc Azad Unversty branch khomen,

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

3. CR parameters and Multi-Objective Fitness Function

3. CR parameters and Multi-Objective Fitness Function 3 CR parameters and Mult-objectve Ftness Functon 41 3. CR parameters and Mult-Objectve Ftness Functon 3.1. Introducton Cogntve rados dynamcally confgure the wreless communcaton system, whch takes beneft

More information

Harfoush, Bestavros, and Byers, Robust Identfcaton of Shared Losses Usng End-to-End Uncast Probes 2 Introducton One of the defnng prncples of the netw

Harfoush, Bestavros, and Byers, Robust Identfcaton of Shared Losses Usng End-to-End Uncast Probes 2 Introducton One of the defnng prncples of the netw Robust Identfcaton of Shared Losses Usng End-to-End Uncast Probes Λ Khaled Harfoush Azer Bestavros John Byers harfoush@cs.bu.edu best@cs.bu.edu byers@cs.bu.edu Computer Scence Department Boston Unversty

More information

A Saturation Binary Neural Network for Crossbar Switching Problem

A Saturation Binary Neural Network for Crossbar Switching Problem A Saturaton Bnary Neural Network for Crossbar Swtchng Problem Cu Zhang 1, L-Qng Zhao 2, and Rong-Long Wang 2 1 Department of Autocontrol, Laonng Insttute of Scence and Technology, Benx, Chna bxlkyzhangcu@163.com

More information

WIRELESS communication technology has gained widespread

WIRELESS communication technology has gained widespread 616 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 4, NO. 6, NOVEMBER/DECEMBER 2005 Dstrbuted Far Schedulng n a Wreless LAN Ntn Vadya, Senor Member, IEEE, Anurag Dugar, Seema Gupta, and Paramvr Bahl, Senor

More information

y and the total sum of

y and the total sum of Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton

More information

Advanced Computer Networks

Advanced Computer Networks Char of Network Archtectures and Servces Department of Informatcs Techncal Unversty of Munch Note: Durng the attendance check a stcker contanng a unque QR code wll be put on ths exam. Ths QR code contans

More information

Resource and Virtual Function Status Monitoring in Network Function Virtualization Environment

Resource and Virtual Function Status Monitoring in Network Function Virtualization Environment Journal of Physcs: Conference Seres PAPER OPEN ACCESS Resource and Vrtual Functon Status Montorng n Network Functon Vrtualzaton Envronment To cte ths artcle: MS Ha et al 2018 J. Phys.: Conf. Ser. 1087

More information

A SIMULATION ANALYSIS OF AGGREGATION STRATEGIES IN A WF 2 Q+ SCHEDULERS NETWORK

A SIMULATION ANALYSIS OF AGGREGATION STRATEGIES IN A WF 2 Q+ SCHEDULERS NETWORK A SIMULATION ANALYSIS OF AGGREGATION STRATEGIES IN A WF 2 Q+ SCHEDULERS NETWORK R. G. Garroppo, S. Gordano, S. Nccoln, F. Russo {r.garroppo, s.gordano, s.nccoln, f.russo}@et.unp.t Department of Informaton

More information

Efficient Content Distribution in Wireless P2P Networks

Efficient Content Distribution in Wireless P2P Networks Effcent Content Dstrbuton n Wreless P2P Networs Qong Sun, Vctor O. K. L, and Ka-Cheong Leung Department of Electrcal and Electronc Engneerng The Unversty of Hong Kong Pofulam Road, Hong Kong, Chna {oansun,

More information

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements Module 3: Element Propertes Lecture : Lagrange and Serendpty Elements 5 In last lecture note, the nterpolaton functons are derved on the bass of assumed polynomal from Pascal s trangle for the fled varable.

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms Course Introducton Course Topcs Exams, abs, Proects A quc loo at a few algorthms 1 Advanced Data Structures and Algorthms Descrpton: We are gong to dscuss algorthm complexty analyss, algorthm desgn technques

More information

GSLM Operations Research II Fall 13/14

GSLM Operations Research II Fall 13/14 GSLM 58 Operatons Research II Fall /4 6. Separable Programmng Consder a general NLP mn f(x) s.t. g j (x) b j j =. m. Defnton 6.. The NLP s a separable program f ts objectve functon and all constrants are

More information

Sequential search. Building Java Programs Chapter 13. Sequential search. Sequential search

Sequential search. Building Java Programs Chapter 13. Sequential search. Sequential search Sequental search Buldng Java Programs Chapter 13 Searchng and Sortng sequental search: Locates a target value n an array/lst by examnng each element from start to fnsh. How many elements wll t need to

More information

Efficient Broadcast Disks Program Construction in Asymmetric Communication Environments

Efficient Broadcast Disks Program Construction in Asymmetric Communication Environments Effcent Broadcast Dsks Program Constructon n Asymmetrc Communcaton Envronments Eleftheros Takas, Stefanos Ougaroglou, Petros copoltds Department of Informatcs, Arstotle Unversty of Thessalonk Box 888,

More information

SAO: A Stream Index for Answering Linear Optimization Queries

SAO: A Stream Index for Answering Linear Optimization Queries SAO: A Stream Index for Answerng near Optmzaton Queres Gang uo Kun-ung Wu Phlp S. Yu IBM T.J. Watson Research Center {luog, klwu, psyu}@us.bm.com Abstract near optmzaton queres retreve the top-k tuples

More information