Loop-Free Multipath Routing Using Generalized Diffusing Computations

Size: px
Start display at page:

Download "Loop-Free Multipath Routing Using Generalized Diffusing Computations"

Transcription

1 1 Loop-Free Multpath Routng Usng Generalzed Dffusng Computatons Wllam T. Zaumen J.J. Garca-Luna-Aceves Sun Mcrosystems, Inc. Computer Engneerng Department 901 San Antono Avenue School of Engneerng Palo Alto, CA Unversty of Calforna Santa Cruz, CA Abstract A new dstrbuted algorthm for the dynamc computaton of multple loop-free paths from source to destnaton n a computer network or nternet are presented, valdated, and analyzed. Accordng to ths algorthms, whch s called DASM (Dffusng Algorthm for Shortest Multpath), each router mantans a set of entres for each destnaton n ts routng table, and each such entry conssts of a set of tuples specfyng the next router and dstance n a loop-free path to the destnaton. DASM guarantees nstantaneous loop freedom of multpath routng tables by means of a generalzaton of Djkstra and Scholten s dffusng computatons. Wth generalzed dffusng computatons, a node n a drected acyclc graph (DAG) defned for a gven destnaton has multple next nodes n the DAG and s able to modfy the DAG wthout creatng a drected loop. DASM s shown to be loop-free at every nstant, and ts average performance s analyzed by smulaton and compared aganst an deal lnk-state algorthm and the Dffusng Update Algorthm (DUAL). 1. Introducton The routng protocols used n today s computer networks and nternetworks are based on shortest-path algorthms that are usually classfed as dstance-vector algorthms and lnk-state or topologybroadcast algorthms. In a dstance-vector algorthm, a router knows the length of the shortest path from each neghbor router to every network destnaton, and uses ths nformaton to compute the shortest path and next router n the path to each destnaton. A router sends update messages to ts neghbors and only to them; each update message contans a vector of one or more entres, each of whch specfes, as a mnmum, the dstance to a gven destnaton. Some dstance-vector algorthms and protocols also specfy the second to last hop n the shortest path [7] [9] or the entre path to a destnaton [12]. In a lnk-state algorthm, each router broadcasts messages contanng the state of each of the router s adjacent lnks to every other router n the network; routers use ths nformaton to compute shortest paths to all network destnatons. Recently, Garca-Luna-Aceves and Behrens ntroduced a hybrd type of routng algorthm, called lnk-vector algorthms [6], n whch a router communcates to ts neghbors the costs of those lnks that belong to ts preferred paths to reach destnatons. Data packets can be routed usng ether ncremental routng or source routng. Wth ncremental routng, also called destnatonbased or hop-by-hop routng, each router n the path to the destnaton makes a decson of where to route the packet next. In contrast, wth source routng, the source router specfes the entre path to the destnaton n the header of each data packet, or a connecton s establshed so that only the connecton-establshment packet has to specfy the entre path, whle the rest of the packets need only specfy the dentfer of the connecton. Usng ncremental routng to route packets along multple loop-free paths can be more desrable than usng source routng, because the routers forwardng a packet can react more quckly to changes n local network condtons than the source router can. Furthermore, source routng lmts the number of dfferent paths that can be used to those known by the source, whch means that only a small number of possble paths can be used. In contrast, f multple loop-free paths are obtaned usng ncremental routng, packets can flow through paths that the source need not know. It s desrable, of course, to avod out-of-order packet delvery, and ths can be acheved wth ncremental routng by arrangng that packets assocated wth partcular end ponts (both source and destnaton) follow the same path. Varous technques can be used to do ths; the choce of technques depends on the hardware and software capabltes of the router, but can be done n ways that do not mpact the algorthm used to compute paths to destnatons, and as such are outsde the scope of ths paper. The provson of multple paths n exstng nternet routng protocols s very lmted. For example, OSPF [15] allows a router to choose more than one path to the same destnaton only when multple paths of mnmum cost exst. IGRP [1] allows a router to forward packets through paths whose length s less than the product of the shortest-path length tmes a varance factor specfed by the network admnstrator. If the varance s too small, ths approach becomes one of routng through one or more shortest paths; f the varance s too large, long-term routng loops can occur even when all routng tables are correct. The lnk-state algorthm reported n [2] and [3] attempts to provde multple loop-free paths by requrng a router to forward packets to a destnaton through neghbors who have reported a smaller dstance than the router forwardng the packet. None of these three approaches prevents routng loops when routng tables are nconsstent. To prevent routng loops due to nconsstent routng tables, the algorthm n [3] requres every router to have consstent topology nformaton; however, requrng that the topology tables of all routers be consstent s mpractcal n large nternets. Because of the routng loops that can occur wth ncremental routng based on any shortest-path algorthm, Gardner, et al. [8] argue that source routng be used. Ths paper demonstrates that ncremental routng over multple loop-free paths s possble, even when routng tables are nconsstent. It extends prevous results on loop-free routng along shortest paths by ntroducng the concept of a shortest multpath, whch s defned as a drected acyclc graph defned by the successor entres of the routng tables of the routers n all the paths from a source to a destnaton that are guaranteed to be loop-free at a gven nstant. A shortest multpath to a destnaton contans the shortest path, but may contan longer paths, as shown n Fgure 1, where dark arrows show the shortest path between routers (and ther successors) and destnaton j. Grayarrows denote part of a shortest multpath that does not le along the shortest path from that pont n the shortest multpath. The shortest multpath from router to destnaton j, as shown n Fgure 1,

2 j Fg. 1. Shortest Multpath Routes shortest multpath shortest path contans addtonal paths wth lengths longer than the shortest path. In ths paper, we ntroduce the Dffusng Algorthm for Shortest Multpath (DASM) to compute shortest multpaths dstrbutedly. DASM s based on a generalzaton of the basc scheme frst proposed by Djkstra and Scholten for the dssemnaton of dstrbuted operatons [4], [13]. Ths generalzaton, whch we call Generalzed Dffusng Computatons, permts a router to synchronze the updatng of the routng-table entry for a gven destnaton whle mantanng multple successors n loop-free paths to the destnaton. Although several approaches for shortest-path routng wth nstantaneous loop-freedom have been proposed based on Djkstra and Scholten s dffusng computatons [10] [5] or smlar apporaches [14], [16], all of these algorthms guarantee nstantaneous loop freedom only f the sngle successor chosen by each router accordng to the algorthm s the one used for packet forwardng. In contrast, DASM mantans nstantaneous loop-freedom through multple successorsfor every destnaton at each router; the only nformaton exchanged among neghbor routers consst of vectors of shortest dstances to destnatons. Synchronzng the update actvty of routers elmnates loopng but can ncur consderable addtonal overhead. To reduce ths overhead, DASM uses a feasblty condton to determne when a router can update ts routng table wthout synchronzng wth other routers. Ths condton depends on the router s own dstance and the dstances reported by ts neghbors. DASM operates wth arbtrary transmsson or processng delays, assumes arbtrary postve lnk costs, and provdes correct routng tables wthn a fnte tme after the occurrence of an arbtrary sequence of lnk-cost or topologcal changes. Secton 2 presents the network model and notaton assumed to descrbe DASM. Secton 3 descrbes the operaton of DASM and shows how t supportsloop-free area routng. Secton 4 proves that DASM s loop-free at every nstant and that t converges to correct routng tables wthn a fnte tme after a sequence of lnk-cost or topology changes. Sectons 5 and 6 analyze DASM s complexty and average performance; t s shown that DASM provdes consderable mprovements over DUAL [5] and the deal lnk-state algorthm, whch consttutes an upper bound on the performance of exstng routng protocols based on topology broadcast (e.g., OSPF). 2. Network Model and Notaton An nternet s modeled as an undrected connected graph n whch each lnk has two lengths or costs assocated wth t one for each drecton and n whch any lnk of the graph exsts n both drectons at any one tme. Each router has a unque dentfer, and lnk costs can vary n tme but are always postve. The smallest dstance between two routers s defned as the sum of the lnk costs n any path of least cost or shortest path between them. In addton, DASM requres an underlyng protocol, called a neghbor-toneghbor protocol for passng messages to neghbors relably and for detectng changes n lnk status. The requrements for ths protocol are: Every router detects wthn a fnte tme the exstence of a new neghbor or the loss of connectvty wth a neghbor. All packets transmtted over an operatonal lnk are receved correctly and n the proper sequence wthn a fnte tme. All messages, changes n the cost of a lnk, lnk falures, and new-neghbor notfcatons are processed one at a tme wthn a fnte tme and n the order n whch they occur. Throughout ths paper, the followng notaton s used: G(V; E): a connected network of arbtrary topology n whch DASM s used wth V denotng the set of all routers and E denotng the set of all lnks. N : the set of routers connected through a lnk wth router ; a router n that set s sad to be a neghbor of router. k: a neghbor of router. (; k): the lnk between routers and k n V. lk: the cost of the lnk to neghbor k; f the lnk to neghbor k fals, the cost s assumed to be 1. j: a destnaton router n V. SG j: the drected graph consstng of those routers that can reach router j. Each lnk n SG j conssts of a drected lnk from a router to a lnk n that router s successor set. 3. The Dstrbuted Algorthm for Shortest Multpath 3.1 Prncples of Operaton In any routng algorthm, the aggregate of each router s routngtable entry mantaned for a gven destnaton j defnes a drected graph rooted at j. When the algorthm converges to correct values, ths drected graph s acyclc. The objectve of a loop-free routng algorthm s to ensure that the routng tables never contan loops,.e., that the routng-table entres of routers defne a drected acyclc graph (DAG) for each destnaton at every nstant. The basc objectve n DASM s to permt routers to mantan at all tmes a DAG for each destnaton that has much more connectvty than the drected trees obtaned wth pror loop-free routng algorthms [5], [7], [10], [14]. To accomplsh ths, each router has a successor set for each destnaton, rather than a sngle successor. A message sent by router s an ordered trplet of the form [MessageType, j, RD j], where MessageType can be update, query or reply, where j s a destnaton, and where RD j s the cost reported n the message. DASM tself does not depend on the use of sequence number or tmers drectly. It follows that, n DASM, a router knows only the dstances to destnatons reported by each neghbor, and communcates the same nformaton to ts neghbors. Requrng routers to synchronze ther update actvty every tme any one of them must change ts routng table ncurs excessve communcaton overhead. Accordngly, DASM uses generalzed dffusng computatons to ensure loop freedom only when a suffcent condton for loop-freedom s not met after an nput event s receved, and behaves much lke the Dstrbuted Bellman-Ford (DBF) algorthm when such a condton s met. Accordngly, each router that uses DASM can be n one of two states: actve and passve. Routers begn executon n the passve state and need to know only about ther own exstence; they may becomes actve at varous tmes by sendng queres to all ts neghbors. When the router receves a reply from each neghbor, t may ether become passve or stay actve. The tme nterval between sendng a query and recevng the last reply to that query s called an actve phase, and multple actve phases for a destnaton can occur successvely, but may not overlap. A flag rjk s mantaned to ensure ths behavor: rjk s true f router has sent a query to router k for destnaton j but has not yet receved a reply and false otherwse. As wth DUAL [5], DASM cannot send a query durng an actve phase, so that once a router sends a query to ts neghbors for

3 destnaton j, the same router cannot send more queres for destnaton j, untl all reples are receved. When router sends a query for destnaton j, router sets r jk = true for all k 2 N,andsets r jk = false when a reply from router k s receved. Router s not allowed to send another query for destnaton j untl r jk = false for all k 2 N (the tme at whch ths frst occurs after sendng a query marks the end of an actve phase). Ths prevents actve phases from overlappng, whch smplfes the state that each router needs to mantan for each destnaton. The rest of ths secton descrbes the condtons used n DASM to determne when to nvoke generalzed dffusng computatons, how generalzed dffusng computatons work n DASM n statc topologes, how topology changes are handled, and why DASM supports not only multple paths to each destnaton n a flat topology, but also loop-free multpath area routng. 3.2 Suffcent Condtons for Loop-Free Routng DASM avods routng-table loops by forcng routers to choose as ther next hops to destnatons only among those neghbor routers that satsfy a destnaton-based orderng constrant that s vald for all routers n the network. Smply put, for a gven destnaton, a router a s gven a label such that t can use a neghbor router b as a successor to the destnaton f and only f b s label s strctly smaller than a s own label. The followng descrbes how DASM accomplshes ths makng use of the followng varables: FD j: s the feasble dstance at router for destnaton j. Sj: s the set of neghbors of router that are used as successors of router for destnaton j. Ths setrepresents the neghbors to use along all loop-free paths mantaned by the routngalgorthm. Djk: the dstance reported by neghbor k n an update, query, or reply. Djk: an upper bound on neghbor k s feasble dstance. RD j: the cost for destnaton j that wll be used n messages sent to neghbors. Let 2 V be an arbtrary router n graph G(V; E). For each destnaton j, and every router k 2 N, DASM requres that router mantans a label FD j representng a feasble dstance and a label Djk that by desgn satsfes FD k j Djk at all tmes (ths s proven n Theorem 1). Because FD k j represents a value avalable at router k (whch s a neghbor of router ), and gven that Djk and FD j represent values avalable at router, the condton FD k j Djk mples that router has an upper bound on each of ts neghbor s feasble dstances for destnaton j, and loop-free paths are easly obtaned by requrng that the next router along a path have a feasble dstance smaller than the current router s feasble dstance. The followng defnton specfes ths orderng constrant as mposed n DASM: Defnton 1: Loop-Free Routng Condton (LRC). Router can choose any neghbor router n the set Sj = fk 2 N j Djk < FD jg. Clearly, for any router k n Sj, FD k j < FD j. Then, provded that all labels are properly updated, loop-free multpath routes are obtaned at each nstant of tme by smply choosng, at each router any router n Sj as the next hop. Furthermore, f Sj contans routers other than ones along the shortest path, there are more avalable paths to a destnaton than what s possble wth shortest-path routng alone. In order to mantan the condton FD k j < FD j at all tmes, DASM behaves dfferently wthn a passve phase than wthn an actve phase. Whle passve (for destnaton j), a router s feasble dstance must be a decreasng (as opposed to monotoncally ncreasng) functon of tme, and ts feasble dstance must be no larger than ts actual dstance to the destnaton. Thus, f the dstance that wll be reported to neghbors falls below the feasble dstance, the feasble dstance must be decreased. Router can explot ths behavor as descrbed below to mantan an upper bound on router k s feasble dstance, thus ensurng loop-free routng whle routers are passve. It s also mportant, however, for the neghbor provdng the shortest path to be n router s successor set, and f ths s not true, router must ncrease ts feasble dstance. Ths s done n two steps. Frst a query contanng the desred dstance s sent, startng an actve phase. Ths query n effect asks permsson from router s neghbors for router to rase ts feasble dstance. When a reply has been receved from each neghbor, endng the actve phase (see Secton 3.5 for how lnk falures are handled), router can safely rase ts feasble dstance to the lowest value reported n any message snce the query. Between recevng a query and sendng a reply, an addtonal varable s used to track the mnmum reported dstance snce the query was receved, and ths becomes the new upper bound once the reply s sent. Ths s descrbed n detal below, and n subsequent sectons, addtonal condtons on the behavor of a router durng an actve phases are ntroduced. These condtons are needed for ensurng convergence and termnaton. To proceed, several addtonal defntons are needed: ~D jk: an upper bound on the feasble dstance of neghbor k after a query from the neghbor s processed and before the next reply s sent. QS j: the set of routers for whch a query has been receved, by router, but for whch a reply has not been sent. The followng summarzes the way n whch the values of Djk and FD j are updated to make LRC a vald orderng constrant. Intally, Sj = ;, rjk = false, and the values of the remanng varables are nfnte, except when = j, n whch case FD j j = RD j j =0. At router 6= j, FD j RD j,andfd j ncreases only at the end of an actve phase, at whch pont can rase FD j to the mnmum value of RD j that occurred durng the actve phase. When router k sends a message to router, the value n the message wll be the current value of RD k j. When router receves ths message, t sets Djk to that value, and updates Djk so that t contans the mnmum of Djk and the prevous value of Djk. If the message was a query, router also sets D ~ jk to the new value of Djk, and subsequently sets D ~ jk to the mnmum of Djk and the prevous value of D ~ jk upon recevng each addtonal message from router k untl router sends router k a reply. When the reply s sent, Djk s set to D ~ jk. Whether a reply has been sent or not s determned by usng the set QS j a neghborng router m s added to ths set when a query for destnaton j s receved from router m, and router m s removed from QS j when a reply for destnaton j s sent by router to router m. Ths s llustrated n Fgure 2 (although ~ D jk and QS j are not shown). It s worth pontng out that when router s passve, Djk = D ~ jk. As we wll show n Secton 4.1, the above mples that FD k j Djk. Therefore, because the successor set Sj satsfes Sj = fk 2 N j Djk < FD jg, then loop-free routng follows because a loop would mply that FD j < FD j, whch s not possble. AlthoughLRC leads to loop-free paths at every nstant, the paths obtaned by smply applyng LRC need not be useful addtonal condtons are needed, whch we dscuss subsequently.

4 FD k j RD j k D * jk Q node k s actve actve phase 1 R Q actve phase 2 tme R Q = Query R = Reply propagaton delay Actve Perod Fg. 2. Feasblty Condton and Reported Dstances 3.3 Condtons for Local Computatons A router s sad to use a local computaton for destnaton j when t updates ts routng-table entry for that destnaton wthout requrng any nternodal synchronzaton. In such a case, we say that router s n the passve state. In passve state, a router sends only updates or reples to queres, but t does not send ts own queres. Because nternodal synchronzaton requres addtonal communcaton overhead, t s desrable for routers to execute local computatons as much as possble, but wthout creatng loops. Because DASM s desgned to provde shortest multpaths, routers should be allowed to update ther successor sets (.e., whch neghbors they can use to reach destnatons) wthout synchronzng wth other routers, as long as those successorsets contan neghbors that provde shortest paths to destnatons. Otherwse, the lengths of the loop-free paths obtaned usng LRC could ncrease wthout bound, even f they are loop-free. Hence, a router remans passve as long as LRC yelds paths that nclude shortest paths. The followng descrbes how DASM accomplshes ths usng the followng varables: D j mn : the cost of a path from router to destnaton j along the shortest path. D j mn = mnfd jk + lk j k 2 N g, assumng that Djk has been updated f router s respondng to a message. Zj: the set of neghbors of router that le along a shortest path to destnaton j, and s computed after D j mn s updated. SjP1 : the value the successor set of router would have after an actve-to-passve transton or f router remans passve. SjP1 = fk 2 N j D ~ jk < mn(fd j;d j mn )g, andscomputed after D ~ jk, FD j,andd j mn are updated. In terms of the above varables, DASM allows a router to execute local computatons for destnaton j as long as Zj \ SjP1 6= ;. If router s passve when a message has been receved for destnaton j or a lnk-cost change has occurred, and Djk, Djk, QS j, Sj, D j mn, Z j,and D ~ jk havebeenupdated, thenrouter wll reman passve (and therefore execute local computatons) f Zj \ SjP1 6= ;. If ths condton s true, then router updates several varable: RD j D j mn, S j SjP1,and FD j mn(fd j;d j mn ). Router then sends updates to all ts neghbors f RD j has changed. If Zj \ SjP1 = ;, then router must become actve. In ths case, f Sj = ;, thenrd j 1. Otherwse a value must be chosen. Ths value can be ether 1 or some value n fdjk + lk j k 2 N g,butfqs j 6= ;, then the value must be at least as large as mnfdjk + lk j k 2 QS jg. 3.4 Internodal Synchronzaton A router becomes actve when LRC does not yeld any shortest path. At that pont, the router must synchronze wth ts neghbors, before t can brng new neghbors to ts successor set to reflect the changes caused by the nput event that makes the router update ts routng table. Internodal synchronzaton n DASM s based on generalzed dffusng computatons, whch dffer from the dffusng computatons used n DUAL n that a router can change ts query successor set wthn a dffusng computaton and can reply to the members of ts query successor set and become a root of a dffusng computaton (DUAL by contrast requres ths set contan a sngle member for the duraton of a dffusng computaton unless the lnk to that member fals). Secton 4.3 provdes condtons under whch ths can be done. The key dea s to explot loop-freedom and to requre that a router never empty ts successor set unless (a) the router becomes passve wth nfnte feasble dstance or (b) a topology change has made the successor unreachable. Wth loop-free paths, a path must ether reach the destnaton, or reach a router wth an empty successor set. Furthermore, except n response to a lnk falure, a router can prevent a neghborng router from leavng the router s successor set by not replyng to a query untl router wll ether have some other router n ts successor set (ths may requre that router send a query wth a suffcently hgh feasble dstance so another successor can be obtaned) or wll become passve wth nfnte feasble dstance (n whch case all the neghbors wll have reported an nfnte dstance, so a passve router wth nfnte feasble dstance has an empty successor set and the defnton of a successor set ensures that such a router s not n any other router s successor set). The requrement that a router never empty ts successor set unless (a) or (b) hold mples that, f router has no successors and s also n the successor set of a neghborng router, then router must be actve and n fact only a lnk falure can cause router to become actve whle stll n the successor set of another node. Hence, as long as an actve router eventually becomes passve (condtons that ensure ths are gven below), then n a stable topology, all routers wth no successors wll eventually become passve routers that cannot be n the successor set of any other routers. Once ths happens, all paths wll lead to the destnaton, and all routers for whch the destnaton s not reachable wll have empty successor sets. We now proceed by descrbng the nternodal sychronzaton mechansms n DASM n detal. When a router receves a query from a destnaton that s not n ts successor set, the router must send a reply to ensure that the lveness property gven n Secton 4.2 holds. Otherwse, when the local computaton condton does not hold at router for destnaton j, router begns or jons a generalzed dffusng computaton by sendng queres to ts neghbors usng the new value chosen for RD j n the queres. Durng ths actve phase, FD j s not ncreased. At the end of the actve phase, FD j wll ncrease to RD j f router starts another actve phase, and wll assume a value no larger than RD j f router becomes passve. The behavor of router at the end of an actve phase (where rjk = false for all k) s determned by a varable SjP2 defned as follows: SjP2 : At the end of an actve phase, S jp2 s the successor set that router would have for destnaton j f router becomes passve for destnaton j. SjP2 = fk 2 N j D ~ jk < mn(d j mn ; RD j)g. If SjP2 \ Z j 6= ;, a neghbor n SjP2 wll provde the shortest path to destnaton j, and router can become passve, settng Sj SjP2, RD j D j mn,andfd j mn(rd j;d j mn ). In becomng passve, router wll also send reples to all neghbors n QS j,and

5 wll send updates as needed to announce changes n RD j.ifsjp2 \ Zj = ; and RD j = 1, then router can also become passve. In ths case, D j mn s also 1, and router wll become passve wth FD j = 1. If SjP2 \ Z j = ; and RD j 6= 1, router must start a new actve phase, and to do ths, t must rase ts feasble dstance. The method used ensures that n a stable topology, router can become passve after a fnte number of actve phases. Frst, router sets FD j RD j, and then sets Sj fk 2 N j Djk < FD jg. If Sj = ;, thenrd j 1;elsef QS j 6= ;, thenrd j mnfdjk + l k j k 2 QS j g,otherwserd j mnfdjk + lk j k 2 Sjg. Because a neghbor can only rase ts feasble dstance when the neghbor receves a reply to a query, and because a router does not have to reply mmedately to a query from any router n ts successor set, a router can keep a neghbor n ts successor set long enough for the router to rase ts feasble dstance suffcently that the router becomes passve. Fnally, f, durng an actve phase, an event results n Zj \ SjP1 6= ;^RD j D j mn at router for destnaton j, then router may send reples to all ts neghbors n QS j and updates as requred to all other neghbors. For performance reasons, ths s done only when QS j \fk 2 N j D ~ jk = 1g = ;. Under these condtons, router would be able to become passve f reples to all ts queres had arrved. Upstream neghbors wll, of course, behave smlarly. Under the approprate crcumstances, the root of the dffusng computaton wll then move upstream shortenng the tme needed for convergence. 3.5 Handlng of Topology Changes Lnk falures are treated as mplct reples to outstandng queres sent to a neghbor reachable over the lnk. Thus, f a falure of lnk (; k) s detected at router, router sets rjk false for each destnaton j. Because no path may traverse a lnk that has faled, router also sets Djk 1, ~D jk 1,andDjk 1.DASM then behaves just as t would for any other event. 3.6 Loop-Free Area Routng Suppose each area s represented by a separate entry n the routng tables, and that a destnaton can belong to more than one area. Whle one could add routng-table entres to represent such ntersectons, ths would ncrease table szes and the length or number of routng-table updates that would have to be sent. We can avod the need for these addtonal table entres as follows. Consder a graph G(V; E) and let J = fj 1;j 2;:::;j mg beaset of m destnatons such that J V.LetFD J = mnffd jn j jn 2 Jg. Furthermore, let SJ = fx j8(j n 2 J) :(D jnx < FD J )g We defne the successor graph SG J of graph G(V;E) for a set of destnatons J at tme t to be a drected graph wth the same routers as those n G(V;E), but for whch a drected edge from router 2 V to router k 2 V exsts f and only f k 2 SJ. Loop-free routng follows mmedately, as s shown n Theorem 3. Thus, to route to a router v that s n more than one area, one can set J to be the destnatons representng the areas that contan v, and use S J n SJ tme. to determne a successor. No matter what successor s used, the path wll always be loop free at each nstant of 4. Correctness of DASM The correctness proof for DASM requres provng that (1) DASM s loop free at every nstant of tme, (2) DASM ensures that a router wll not wat ndefntely before recevng reples to queres, (3) n a stable topology DASM converges n C j (the routers wth no paths to destnaton j), and (4) n a stable topology, DASM converges n C j (the routers for whch a path to destnaton j exsts). For purposes of the proof, varables n DASM are modeled as functons of tme, wth an equvalent defnton of Djk that avods the need to use the varable D ~ jk. 4.1 Loop Freedom For a router, a destnaton j, and neghbor k of router, FD j(t) s the feasble dstance at router, RD j(t) s the dstance that router wll report n an update, Mjk s the set of tmes at whch router processes a message from router k for destnaton j, andxmt k (t) s the tme at whch a message processed by router at tme t was sent by router k. Djk(t) s defned as follows: Defnton 2: Let G(V;E) be a graph, let 2 V be a router, let j 2 V be a destnaton, and let k 2 N be a neghbor of router. Then Djk(t) s a functon satsfyng the followng condtons: When lnk (; k) s not operatonal or has just recovered, or when router ntalzes tself at tme t, D (t) =1. Suppose router sends a reply to router k at tme t r. Let t 1 be the last tme before or at t r at whch ether a query from router k was processed, lnk (; k) recovered, or router ntalzed tself. Then Djk(t r) = mnfrd k j (xmt k ()) j 2 M jk \ [t 1; t r]g. Otherwse, Djk(t) = mn(frd k (xmt k ()) j 2 M jk \ [t 2; t]g[fd jk(t 2)g), wheret 2 s the last tme before t when router ether sent router k a reply, detected a lnk recovery for lnk (; k), or ntalzed tself. Lemma 1: Let G(V; E) be a graph n whch each router runs DASM. If router 2 V ntalzes tself or receves the last reply to a query for destnaton j 2 V at tme t, all undelvered messages (f any) sent by router for destnaton j wll contan a dstance greater than or equal to RD j(t, ), the value mmedately before router processes the last reply. Proof: When a router ntalzes tself, there are no undelvered messages. The lemma holds n ths case. DASM requres that all messages be delvered n FIFO order, and that all messages that are undelvered before a lnk falure are lost. Thus, all messages that router orgnated before sendng a query wll ether have been lost or delvered. For some destnaton, the reply dstance of a router must be a decreasng functon of tme over the nterval between sendng a query for that destnaton and recevng the last reply to that query. Accordngly, all messages from router that are undelvered when the last reply to that query s receved must contan a value greater than or equal to RD j(t, ). Lemma 2: Let G(V; E) be a graph n whch each router runs DASM, and let j 2 V be a destnaton. At tme t, when a router 2 V ntalzes tself or receves ts last reply f t had become actve, any neghbor k 2 N that can use router as a feasble successorfor some destnaton j wll have ether set Dj k (t) to some value such that RD j(t, ) Dj k (t). Proof: At t =0, all routers ntalze themselves and set the dstance to ther neghbors to 1. The lemma s consequently true at tme t =0. At any pont at whch a router has become actve, or at whch t wll stay actve after recevng reples from all ts neghbors, the router wll send queres to all ts neghbors when orgnatng a dffusng computaton, and wll send queres to all ts neghbors. Suppose that, at tme t 2, router receves ts last reply to a query for destnaton j sent at tme t 1. Let RD j(t + 1 ) be the value of the reply dstance for router that wll be sent n the query generated at tme t 1,andletRD j(t, ) be the value of the reply dstance for router just before t receves ts last reply. Because jk

6 RD j s a decreasng functon n (t 1;t), RD j(t + 1 ) RD j(t, ). Because router s actve for all t 2 (t 1;t), and accordng to DASM s operaton, router can only send messages for destnaton j contanng a dstance RD j() at some tme 2 (t 1;t) f RD j() mnfrd j(t 0 ) j t 0 2 (t 1;t)g. Suppose that router has sent a query to router k. Unless the lnk from router fals n (t 1;t), router k wll have processed that query and router wll have receved ts reply sometme at or before t. When router k sends ts reply at tme t r, t wll have receved a query contanng the value RD j(t + 1 ) and wll have set Dk j (t r) to the mnmum value or RD j receved n the query or a subsequent update. Because RD j s a decreasng functon of tme n (t 1;t), and because router k receved the query transmtted from router at tme t 1, RD j(t, ) s less than or equal to the value n the last message receved by router k before tme t. Thus, RD j(t, ) Dj k (t). If the lnk has faled at some tme t f durng (t 1;t 2), router k wll set Dj k (t f )=1. If a message from s subsequently receved durng (t f;t), router k can set Dj k (t) to some value such that Dj k (t) 2fRD j() j 2 (t f ;t)g. Ths s also true f the lnk fals multple tmes. If the lnk had faled at or before t 1 and recovered (or was ntally establshed) after t 1, then router k may or may not have receved a message (whch must have been generated after t 1) from router for destnaton j. If router k has not receved ths message, then Dj k (t) =1 n whch case RD j(t, ) Dj k (t)g, otherwse Dj k (t 2) 2fRD j(t 0 ) j t 0 2 (t + 1;t, 2 )g, n whch case RD j(t, ) Dj k (t) because RD j s a decreasng functon of tme n (t 1;t 2). Lemma 3: Let G(V; E) be a graph n whch each router runs DASM. Let and k be routers n V,andj be a destnaton. Furthermore, let k 2 N. Suppose that, at tme t a, FD j(t a) Dj k (t a), all messages that are undelvered (but not lost) at tme t a and that are generated by router before t a contan a dstance at least as large as FD j(t a), andthatfd j s a decreasng functon of t for t 2 [t a;t b]. ThenFD j(t) Dj k (t) for all t 2 [t a;t b]. Proof: If the lnk from router to router k fals at some tme t f 2 [t a;t b], wheret f s the earlest tme n ths nterval at whch such a lnk falure occurs, then router k wll set Dj k (t f ) to nfnty, and reset the value only when a new message for destnaton j s receved from router. The lemma thus holds when Dj k (t f )=1. In the nterval [t f ;t b], the only messages receved by router k wll have been generated by router after t f because of the operaton of the neghbor-to-neghbor protocol. Thus, there are no messages are undelvered at tme t f that wll be receved after t f, and by assumpton, Fj s a decreasng functon of t n [t f;t b], and all the condtons the lemma requres are therefore true over the nterval [t f ;t b] f these condtons are true for [t a;t b]. It s therefore suffcent to prove that the lemma s true only n the case where there are no falures of lnk (; k) n the nterval [t a;t b] By assumpton, all messages generated by router before t a and receved by router k contan dstances larger than FD j(t a), and FD j(t) FD j(t a). The lemma therefore holds when any of these messages are delvered to router k, and f no other messages or no messages at all are delvered to router k from router durng [t a;t b], the lemma s true. The remanng case s the one where at least one message generated after tme t a s receved before tme t 2 [t a;t b]/ For any message generated at tme t 0 2 [t a;t b] for destnaton j, DASM requres that RD j(t 0 ) FD j(t 0 ), and because FD j s a decreasng functon of tme n [t a;t b], t follows that FD j(t) mnfrd j() j 2 [t a;t]g for t 2 [t a;t b]. Smlarly, because Dj k (t a) FD j(t a), because all messages generated before t a and receved after t a contan dstances larger than FD j(t a), and because the defnton of Dj k requres t to contan the mnmum dstance seen n a message after ether a router ntalzed tself (at whch pont the value s nfnty), the lnk recovered (at whch pont the value s nfnty), or a query was receved for whch a reply has been sent, t follows that Dj k (t) mn(ffd j(t a)g [frd j(tau) j 2 [t a;t]g for t 2 [t a;t b]. Because FD j(t) mnfrd j() j 2 [t a;t]g, t s obvously true that FD j(t) mn(ffd j(t a)g [frd j() j 2 [t a;t]g). Consequently, FD j(t) Dj k (t) for t 2 [t a;t b].. Theorem 1: Let G be a graph n whch each router runs DASM. Let, j and k be routers n G,letj be a destnaton, and let k 2 N. Then FD j(t) Dj k (t) s true for all t 0. Proof: Consder the tmes n the nterval [0;t] at whch router ether ntalzes tself, looses all ts neghbors, or receves the last reply to a query. These tmes form a sequence L = ft 0 = 0;t 1;t 2;:::;t n <tg, wheret 0 s the tme at whch router ntalzes tself, and t 1;t 2;:::;t n are the tmes at whch router receves the last reply from some query. DASM requres that the feasble dstance of router always decrease except at the tmes n L. When a router frst ntalzes tself, the theorem s true because all ts neghbors wll set ther dstance-table entry for that router to 1 for destnaton j untl they receve a message for destnaton j contanng a lower dstance. Other than at tme t =0, DASM allows a router a router to ncrease ts feasble dstance only when t receves reples from all ts neghbors, or when t has lost all neghbors (n whch case ts neghbors wll set ther dstance for that router to 1). Thus, n the nterval [0;t 1) the Lemma 3 apples and FD j(t) Dj k (t) for all t 2 [0;t 1). The proof proceeds by nducton. Consder any tme t m 2 L and assume the theorem holds up to that tme. Accordng to Lemma 1, at any tme t 2 [t m;t m+1], any undelvered message for destnaton j that router sent before t m wll contan a dstance greater than or equal to RD j(t, m). By Lemma 2, for any neghbor k of router, Djm(t k m) wll be larger than RD j(t, m). When router receves ts last reply at tme t m, t wll change ts feasble dstance to FD j(t m)=rd j(t, m). Because DASM requres that at any router, FD j(t) RD j(t), t follows that FD j(t m) Dj k (t m). Because FD j(t) s a decreasng functon n the nterval [t m;t m+1), Lemma 3 mples that FD j(t) Dj k (t) for all t 2 [t m;t m+1). We can therefore proceed by nducton up to tme t. Theorem 2: Let G(V; E) be a graph n whch each router runs DASM. Let P j(t) be a path through G for destnaton j at tme t for whch each 2 P j(t) (other than the last router) chooses a router s as the next router n P j(t) such that s 2 Sj(t). ThenP j(t) s loop free. Proof: By defnton, each router s 2 Sj satsfes Djs(t) < FDj(t). By Theorem 1, for any router 2 P j(t), the next router s n P j(t) wll satsfy FDj s (t) Djs. By assumpton, Djs(t) < FD j(t); accordngly, t follows that FDj s (t) < FD j(t). Thus, each subsequent router along the path has a lower value for ts feasble dstance. If P j(t) contaned a loop, then for any router p 2 P j(t) that was part of such a loop, t would follow that FD p j (t) <FDp j (t). Thus, by contradcton, the path must be loop free. The followng theorem shows that DASM also provdes loopfree area routng: Theorem 3: Let G(V; E) be a graph n whch each router runs DASM. Let SG J(t) be the successor graph for graph G(V; E) and for destnaton set J at tme t. ThenSG J(t) s loop free. Proof: By defnton, for each node 2 SG J(t), every neghbor k 2 N \ SG J(t) satsfes k 2 SJ,whereSJ (t) = fx j 8(j n 2 J) : (Djnx(t) < FD J (t))g. For k 2 N,let

7 Jk (t) =fj n 2 J j Djnk(t) < FD J (t)g. Clearly, for each j n 2 Jk (t), D jnk (t) < FD J (t). By Theorem 2, FD k jn (t) D jnk (t), and thus FD k jn (t) FD J (t). By the defnton of FD k J (t), FD k J (t) FD k jn (t), and therefore FDk J (t) < FD J (t) for any k 2 SJ (t). Accordngly, f there s a loop L J(t) n SG J(t) at tme t, then for some p 2 L J(t), FD p J (t) < FDp J (t), and by contradcton, SG J(t) s loop free. 4.2 Lveness Theorem 4: Let G(V; E) be a graph n whch each router runs DASM. If router becomes actve, t wll receve a reply from each neghbor (wth lnk falures generatng mplct reples) n a fnte tme. The formal proof s essentally the same as the proof for DUAL [5], and can be descrbed nformally as follows: The frst requrement for showng that generalzed dffusng computatons termnate s to show that a router that sends a query wll always receve a reply to that query and wll eventually become passve. To ensure ths, several requrements are necessary: a router must reply mmedately to a query from any router that s not n ts successor set, a router may not rase RD j durng an actve phase (excludng ts start), and f a new actve phase begns mmedately after a prevous actve phase, then the router must rase ts reported dstance to mn(fdjk(t)+l k(t) j k 2 QS j (t)g [ f1g, and s requred to become passve f possble (n whch case the router must send reples to any neghbors stll n QS j (t)). Furthermore, f QS j (t) 6= ;, router should not decrease RD j durng an actve phase to a value below mn(fdjk(t) +lk(t) j k 2 QS j (t)g unless router reples to all routers n QS j (t). As a result, router wll ether have sent all the routers n QS j a reply durng an actve phase, or wll be able to become passve at the end of the actve phase, as long as router tself receves reples to all the queres t has sent. Thus, a router wll reply to a query unless some other router does not reply to t s query, whch cannot happen: f t dd, one can generate a sequence of routers, followng the successor graph upstream, untl one reach a router that had not repled, but s not n the successor set of any router. Because a router that s not n the successor set of any other router must send reples, t follows by contradcton that all dffusng computatons wll termnate. 4.3 Convergence n C j The proof for convergence s C j makes use of the followng sets: Defnton 3: W 1j(t)=fx 2 V j x 6= j ^ S x j (t) =;^9y :(y 2 V ^ x 2 S y j )g W 2j(t)=fx 2 V j x 6= j ^ S x j (t) =;^actve x j (t)g A 1j(t)=fx 2 V j x 6= j ^ S x j (t) 6= ;g A 2j(t)=fx 2 V j x 6= j ^ S x j (t) =;^passve x j (t)g and W j(t) = W 1j(t) [ W 2j(t), where actve x j (t) s true f and only f router x s actve for destnaton j at tme t, and passve x j (t) s true f and only f router x s passve for destnaton j at tme t. These sets are shown n Fgure 3. As proven below, DASM ensures that routers move between these sets only as shown n Fgure 4, and that the set W j(t) eventually becomes empty once the topology s stable. W j(t) contans all the routers other than the destnaton at whch a path (obtaned by followng successors)termnates, so once W j(t) s empty, all paths must be ones that go to the destnaton, and all routers that cannot reach the destnaton must have empty successor sets. j C j W j A 2j Fg. 3. The Successor Graph and Sets Wj, A 1j,andA2j Lnk change only A 1j W j A 1j C j A 2j Fg. 4. Transtons between Sets Wj, A 1j,andA2j The operaton of DASM s desgned to ensure that the followng condtons are true: R1 If an actve phase starts at tme t and Sj(t) = ;, then RD j(t) = 1. R2 Suppose an actve phase at router starts at tme t q and ends at tme t, andthatrd j() =1 for all 2 [t q;t). Then router must become passve at tme t. R3 When router s passve at tme t, andf 6= j and Sj(t) = ;,thenfd j(t) = 1 and RD j(t) = 1. R4 Suppose router s actve at tme t and that the actve phase at tme t began at tme t 1 <t.letmj be the tmes at whch router generated a message for destnaton j. Then for some 2 [t 1;t] \ Mj, RD j(t)=rd j() where RD() =1 f Sj() =;. R5 If Sj(t, ) 6= ; and router does not process a lnk event at tme t,thensj(t) 6= ; unless router becomes passve at tme t or stays passve at tme t. R6 If a router s passve at tme t, and Sj(t, ) =;,then router must stay passve at tme t. R7 f FD j(t) = 1, then router cannot be n the successor set of any other router. Lemma 4: Let G(V; E) be a graph for whch each router runs DASM and let j 2 V be a destnaton. Then W j(t) [ A 1j(t) [ A 2j(t) [fjg = V. Proof: Routers are ether actve or passve, but not both. The lemma follows by drect computaton of W j(t)[a 1j(t)[A 2j(t)[ fjg. Lemma 5: Let G(V; E) be a graph for whch each router runs DASM and let j 2 V be a destnaton. If W j(t) = ;, then A 1j(t) C j(t) and C j(t) A 2j(t). Proof: If A 1j(t)\ C j(t) 6= ;, then, because SG j(t) s acyclc by Theorem 2, startng at some router n A 1j(t) \ C j(t) and followng successors must lead to a router n C j(t) wth no successors. Such a router must be n W 1j(t),andW 1j(t) 2 W j(t). Therefore, f W j(t) =;, thena 1j(t) \ C j(t) =;. By Lemma 4, C j(t) = [W j(t)\ C j(t)][[a 1j(t)\ C j(t)][[a 2j(t)\ C j(t)]; therefore,, t follows that, f W j(t) =;,thenc j(t) =A 2j(t) \ C j(t). Accordngly, f W j(t) =;, thena 1j(t) C j(t) and C j(t) A 2j(t).

8 Lemma 6: Let G(V; E) be a graph for whch each router runs DASM and let j 2 V be a destnaton. Then all routers n W j(t) are actve (.e., W j(t) =W 2j(t)). Proof: Suppose router 2 V s passve at tme t and that 2 W 1j(t). By defnton of W 1j, Sj(t) =; and by R3, FD j(t) = 1. Thus, by R7, router cannot be n the successor set of any router. Because a router n W 1j(t) must be n the successor set of some other router by Defnton 3, the assumpton that a passve router s n W 1j(t) s false. By defnton, all routers n W 2j(t) are actve, W 1j(t) W 2j(t); therefore, W j(t) =W 2j(t). Accordngly, by Defnton 3, all routers n W j(t) are actve. Lemma 7: Let G(V; E) be a graph for whch each router runs DASM and let j 2 V be a destnaton. Then A 1j(t), A 2j(t), W j(t),andfjg are dsjont sets that cover V. Proof: By Defnton 3, fjg s dsjont from A 1j(t), A 2j(t), and W j(t). Agan by Defnton 3, both W j(t) and A 2j(t) are subsets of fx 2 V j x 6= j ^ Sj x (t) = ;g, and by defnton, A 1j(t) = fx 2 V j x 6= j ^ Sj x (t) 6= ;g. Clearly, A 1j(t) \ W j(t) =; and A 1j(t) \ A 2j(t) =;. By defnton, all routers n A 2j(t) are passve and by Lemma 6, all routers n W j(t) are actve. Thus, A 2j(t) \ W j(t) =;. anda 1j(t), A 2j(t), W j(t), and fjg are consequently dsjont sets. By Lemma 4, A 1j(t), A 2j(t), W j(t),andfjg cover V. Lemma 8: Let G(V; E) be a graph for whch each router runs DASM, let 2 V be a router, and let j 2 V be a destnaton. If 2 W j(t), then at some tme t 0 >t, 2 A 1j(t 0 ) [ A 2j(t 0 ). Proof: Suppose 2 W j() for all t. By Lemma 6, router s actve at tme t and by Theorem 4, must termnate the actve phase that exsts at tme t at some tme t 1 >t. By assumpton, 2 W j(), and accordng to Lemma 6 all routers n W j(t 1) are actve; therefore, t follows that router must begn another actve phase at tme t 1. Because Sj(t 1) = ; f 2 W j(t), then by R1, RD j(t 1)=1. By Theorem 4, ths second actve phase must termnate at some tme t 2. If 2 W j() for 2 [t 1;t 2), then Sj() =; for 2 [t 1;t 2), and by R4, RD j() =1 for 2 [t 1;t 2). Thus, mnfrd j() j 2 [t 1; t 2)g = 1. By R2, router must become passve at tme t 2, and hence, 62 W j(t 2), contradctng the assumpton that 2 W j() for all t. Gven that j 62 W j(t),then 6= j. Thus by Lemma 7, at some tme t 0 >t, 2 A 1j(t 0 ) [ A 2j(t 0 ). Lemma 9: Let G(V; E) be a graph for whch each router runs DASM, let 2 V be a router, and let j 2 V be a destnaton. Suppose that no lnk events occur after tme t c. If t > t c, 2 A 1j(t, ),then 62 W j(t). Proof: By Lemma 7, 62 W j(t, ) and by Defnton 3, Sj(t, ) 6= ;. By Lemma 6, f 2 W j(t), router must ether be actve at tme t, or change from passve to actve at tme t. Because there are no lnk events after tme t c,byr5,sj(t) 6= ; for router to be actve at tme t. Hence 62 W j(t). Lemma 10: Let G(V; E) be a graph for whch each router runs DASM, let 2 V be a router, and let j 2 V be a destnaton. Suppose that no lnk events occur after tme t c.if2a 2j(t, ) for t, >t c,then 62 W j(t). Proof: By Defnton 3, router s passve at tme t, and Sj(t, ) 6= ;. By R6, router must reman passve at tme t, and therefore 62 W j(t) because accordng to Lemma 6, all routers n W j(t) are actve. Theorem 5: Let G(V; E) be a graph for whch each router runs DASM, and let j 2 V be a destnaton. If no lnk events occur after tme t c, then there must exst a tme t f >t c such that for all t t f, W j(t) =; and C j(t) A 2j(t). Proof: By Lemmas 7 and 8, any router 2 W j(t c) must satsfy 62 W j(t 0 ) and 2 A 1j(t 0 ) [ A 2j(t 0 ) for some tme t 0 > t c. By Lemmas 9, 10, the assumpton that there are no lnk events after t c, and the condton 2 A 1j(t 0 ) [ A 2j (t 0 ), t follows that 62 W j(t) for all t>t 0. Because W j(t c) can contan only a fnte number of routers, there must exst a tme t f such that W j(t) =; for all t t f. Because W j(t) =; for all t t f, Lemma 8 mples that C j(t) A 2j(t) for t t f. Theorem 5 shows that eventually for destnaton j,gven a stable network, every router n C j wll be n the set A 2j(t) for t>t f,and the defnton of A 2j(t) requres that members of ths set be passve wth an empty successor set. The operaton of DASM also ensures that routers n ths state have ther dstance to j set to nfnty, thus ensurng convergence n the C j wth correct dstances. Termnaton follows because DASM does not send messages (other than reples to queres) when passve unless dstances change. 4.4 Convergence n C j The proof for convergence n C j s omtted; t s smlar to the proof for DUAL [5] and follows by an nductve argument. The dea behnd the proof s to show that, wth a stable topology, all routers n C j eventually have obtaned dstances from each neghbor at least as large as the dstances along the shortest paths through those neghbors. The proof then shows that the router whose shortest-path dstance to the destnaton s smallest (ths must be a neghbor of the destnaton) wll become passve wth the correct dstance, and only send reples to queres. 5. Complexty of DASM DASM s complexty s measured by the number of steps, called tme complexty or TC, and the number of messages,called communcaton complexty or CC, requred by the algorthms after a sngle change n the cost or status of a lnk. To calculate DASM s complexty, t s necessary to assume that the algorthms under study behave synchronously, so that every router n the network executes a step of the algorthm smultaneously at fxed ponts n tme. At each step, a router receves and processes all nput events orgnated durng the precedng step, and, f needed, creates an update message after processng each nput event; all these update messages are transmtted n the same step. The frst step occurs when at least one router detects a topologcal change and ssues update messages to ts neghbors. Durng the last step, at least one router receves and processes updates from ts neghbors, after whch all routng tables are correct and routers stop transmttng updatesuntl a new topologcal change takes place. DASM has the same tme and communcaton complexty as DUAL. After a sngle lnk falure or lnk-cost ncrease, DASM has TC = O(x) [5] [10], where x s the number of routers affected by the routng-table perturbaton. To verfy ths, we note that n the worst case all routers upstream of a destnaton router j n SG j must partcpate n a dffusng computaton for router j, whch corresponds to the operaton of DUAL. DASM s communcaton complexty s the same as DUAL s,.e., CC = O(6D x), whered s the maxmum degree of a router; ths s verfed n [5]. After a sngle lnk addton or lnk-cost reducton, DASM has TC = O(d) and CC = O(E), wheree s the number of lnks n the network and d s the network dameter (.e., the length of the longest shortest path n hops between any two network routers). To verfy ths, note that any router that receves a query reportng a dstance decrease must be able to fnd a feasble successor; accordngly, a router that sends a query after ts dstance decreases must receve mmedate reples from all ts neghbors, wthout those neghbors havng to forward the query. On the other hand, routers as far as d hops from the router detectng the change may requre to send an update after recevng an update from a neghbor on ther successor set, and all routers may have to send an update as a result of the update they

MPATH: A Loop-free Multipath Routing Algorithm

MPATH: A Loop-free Multipath Routing Algorithm 1 MPATH: A Loop-free Multpath Routng Algorthm SRINIVAS VUTUKURY vutukury@cse.ucsc.edu Computer Scences Department Unversty of Calforna Santa Cruz, CA 9564 J.J. GARCIA-LUNA-ACEVES jj@cse.ucsc.edu Computer

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

A Simple Approximation to Minimum-Delay Routing

A Simple Approximation to Minimum-Delay Routing A Smple Approxmaton to Mnmum-Delay Routng Srnvas Vutukury Computer Scence Department J.J. Garca-Luna-Aceves Computer Engneerng Department Baskn School of Engneerng Unversty of Calforna, Santa Cruz, CA

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

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

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

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

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

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

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

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

Transaction-Consistent Global Checkpoints in a Distributed Database System

Transaction-Consistent Global Checkpoints in a Distributed Database System Proceedngs of the World Congress on Engneerng 2008 Vol I Transacton-Consstent Global Checkponts n a Dstrbuted Database System Jang Wu, D. Manvannan and Bhavan Thurasngham Abstract Checkpontng and rollback

More information

CMPS 10 Introduction to Computer Science Lecture Notes

CMPS 10 Introduction to Computer Science Lecture Notes CPS 0 Introducton to Computer Scence Lecture Notes Chapter : Algorthm Desgn How should we present algorthms? Natural languages lke Englsh, Spansh, or French whch are rch n nterpretaton and meanng are not

More information

Optimal Fault-Tolerant Routing in Hypercubes Using Extended Safety Vectors

Optimal Fault-Tolerant Routing in Hypercubes Using Extended Safety Vectors Optmal Fault-Tolerant Routng n Hypercubes Usng Extended Safety Vectors Je Wu Department of Computer Scence and Engneerng Florda Atlantc Unversty Boca Raton, FL 3343 Feng Gao, Zhongcheng L, and Ynghua Mn

More information

Making Name-Based Content Routing More Efficient than Link-State Routing

Making Name-Based Content Routing More Efficient than Link-State Routing Makng Name-Based Content Routng More Effcent than Lnk-State Routng Ehsan Hemmat and J.J. Garca-Luna-Aceves, Comuter Engneerng Deartment, UC Santa Cruz, Santa Cruz, CA 95064 PARC, Palo Alto, CA 94304 {

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

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

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

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

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

PERFORMANCE ANALYSIS OF ROUTING ALGORITHMS OF RD-C/TDMA PACKET RADIO NETWORKS UNDER DYNAMIC RANDOM TOPOLOGY1

PERFORMANCE ANALYSIS OF ROUTING ALGORITHMS OF RD-C/TDMA PACKET RADIO NETWORKS UNDER DYNAMIC RANDOM TOPOLOGY1 PERFORMANCE ANALYSIS OF ROUTING ALGORITHMS OF 1- RD-C/TDMA PACKET RADIO NETWORKS UNDER DYNAMIC RANDOM TOPOLOGY1 A Thess Presented to The Faculty of the College of Engneerng and Technology Oho Unversty

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

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

Constructing Minimum Connected Dominating Set: Algorithmic approach

Constructing Minimum Connected Dominating Set: Algorithmic approach Constructng Mnmum Connected Domnatng Set: Algorthmc approach G.N. Puroht and Usha Sharma Centre for Mathematcal Scences, Banasthal Unversty, Rajasthan 304022 usha.sharma94@yahoo.com Abstract: Connected

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

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

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

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

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

Math Homotopy Theory Additional notes

Math Homotopy Theory Additional notes Math 527 - Homotopy Theory Addtonal notes Martn Frankland February 4, 2013 The category Top s not Cartesan closed. problem. In these notes, we explan how to remedy that 1 Compactly generated spaces Ths

More information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

More information

CHAPTER 2 DECOMPOSITION OF GRAPHS

CHAPTER 2 DECOMPOSITION OF GRAPHS CHAPTER DECOMPOSITION OF GRAPHS. INTRODUCTION A graph H s called a Supersubdvson of a graph G f H s obtaned from G by replacng every edge uv of G by a bpartte graph,m (m may vary for each edge by dentfyng

More information

Greedy Technique - Definition

Greedy Technique - Definition Greedy Technque Greedy Technque - Defnton The greedy method s a general algorthm desgn paradgm, bult on the follong elements: confguratons: dfferent choces, collectons, or values to fnd objectve functon:

More information

Improving Low Density Parity Check Codes Over the Erasure Channel. The Nelder Mead Downhill Simplex Method. Scott Stransky

Improving Low Density Parity Check Codes Over the Erasure Channel. The Nelder Mead Downhill Simplex Method. Scott Stransky Improvng Low Densty Party Check Codes Over the Erasure Channel The Nelder Mead Downhll Smplex Method Scott Stransky Programmng n conjuncton wth: Bors Cukalovc 18.413 Fnal Project Sprng 2004 Page 1 Abstract

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

On Maximizing the Lifetime of Delay-Sensitive Wireless Sensor Networks with Anycast

On Maximizing the Lifetime of Delay-Sensitive Wireless Sensor Networks with Anycast On Maxmzng the Lfetme of Delay-Senstve Wreless Sensor Networks wth Anycast Joohwan Km, Xaojun Ln, Ness B. Shroff, and Prasun Snha School of Electrcal and Computer Engneerng, Purdue Unversty Departments

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

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

CSE 326: Data Structures Quicksort Comparison Sorting Bound

CSE 326: Data Structures Quicksort Comparison Sorting Bound CSE 326: Data Structures Qucksort Comparson Sortng Bound Steve Setz Wnter 2009 Qucksort Qucksort uses a dvde and conquer strategy, but does not requre the O(N) extra space that MergeSort does. Here s the

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

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

Notes on Organizing Java Code: Packages, Visibility, and Scope

Notes on Organizing Java Code: Packages, Visibility, and Scope Notes on Organzng Java Code: Packages, Vsblty, and Scope CS 112 Wayne Snyder Java programmng n large measure s a process of defnng enttes (.e., packages, classes, methods, or felds) by name and then usng

More information

Sorting Review. Sorting. Comparison Sorting. CSE 680 Prof. Roger Crawfis. Assumptions

Sorting Review. Sorting. Comparison Sorting. CSE 680 Prof. Roger Crawfis. Assumptions Sortng Revew Introducton to Algorthms Qucksort CSE 680 Prof. Roger Crawfs Inserton Sort T(n) = Θ(n 2 ) In-place Merge Sort T(n) = Θ(n lg(n)) Not n-place Selecton Sort (from homework) T(n) = Θ(n 2 ) In-place

More information

Online Policies for Opportunistic Virtual MISO Routing in Wireless Ad Hoc Networks

Online Policies for Opportunistic Virtual MISO Routing in Wireless Ad Hoc Networks 12 IEEE Wreless Communcatons and Networkng Conference: Moble and Wreless Networks Onlne Polces for Opportunstc Vrtual MISO Routng n Wreless Ad Hoc Networks Crstano Tapparello, Stefano Tomasn and Mchele

More information

Brave New World Pseudocode Reference

Brave New World Pseudocode Reference Brave New World Pseudocode Reference Pseudocode s a way to descrbe how to accomplsh tasks usng basc steps lke those a computer mght perform. In ths week s lab, you'll see how a form of pseudocode can be

More information

such that is accepted of states in , where Finite Automata Lecture 2-1: Regular Languages be an FA. A string is the transition function,

such that is accepted of states in , where Finite Automata Lecture 2-1: Regular Languages be an FA. A string is the transition function, * Lecture - Regular Languages S Lecture - Fnte Automata where A fnte automaton s a -tuple s a fnte set called the states s a fnte set called the alphabet s the transton functon s the ntal state s the set

More information

CSE 326: Data Structures Quicksort Comparison Sorting Bound

CSE 326: Data Structures Quicksort Comparison Sorting Bound CSE 326: Data Structures Qucksort Comparson Sortng Bound Bran Curless Sprng 2008 Announcements (5/14/08) Homework due at begnnng of class on Frday. Secton tomorrow: Graded homeworks returned More dscusson

More information

Overview. Basic Setup [9] Motivation and Tasks. Modularization 2008/2/20 IMPROVED COVERAGE CONTROL USING ONLY LOCAL INFORMATION

Overview. Basic Setup [9] Motivation and Tasks. Modularization 2008/2/20 IMPROVED COVERAGE CONTROL USING ONLY LOCAL INFORMATION Overvew 2 IMPROVED COVERAGE CONTROL USING ONLY LOCAL INFORMATION Introducton Mult- Smulator MASIM Theoretcal Work and Smulaton Results Concluson Jay Wagenpfel, Adran Trachte Motvaton and Tasks Basc Setup

More information

Non-Split Restrained Dominating Set of an Interval Graph Using an Algorithm

Non-Split Restrained Dominating Set of an Interval Graph Using an Algorithm Internatonal Journal of Advancements n Research & Technology, Volume, Issue, July- ISS - on-splt Restraned Domnatng Set of an Interval Graph Usng an Algorthm ABSTRACT Dr.A.Sudhakaraah *, E. Gnana Deepka,

More information

Space-Optimal, Wait-Free Real-Time Synchronization

Space-Optimal, Wait-Free Real-Time Synchronization 1 Space-Optmal, Wat-Free Real-Tme Synchronzaton Hyeonjoong Cho, Bnoy Ravndran ECE Dept., Vrgna Tech Blacksburg, VA 24061, USA {hjcho,bnoy}@vt.edu E. Douglas Jensen The MITRE Corporaton Bedford, MA 01730,

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

b * -Open Sets in Bispaces

b * -Open Sets in Bispaces Internatonal Journal of Mathematcs and Statstcs Inventon (IJMSI) E-ISSN: 2321 4767 P-ISSN: 2321-4759 wwwjmsorg Volume 4 Issue 6 August 2016 PP- 39-43 b * -Open Sets n Bspaces Amar Kumar Banerjee 1 and

More information

Fibre-Optic AWG-based Real-Time Networks

Fibre-Optic AWG-based Real-Time Networks Fbre-Optc AWG-based Real-Tme Networks Krstna Kunert, Annette Böhm, Magnus Jonsson, School of Informaton Scence, Computer and Electrcal Engneerng, Halmstad Unversty {Magnus.Jonsson, Krstna.Kunert}@de.hh.se

More information

Related-Mode Attacks on CTR Encryption Mode

Related-Mode Attacks on CTR Encryption Mode Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory

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

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

with `ook-ahead for Broadcast WDM Networks TR May 14, 1996 Abstract

with `ook-ahead for Broadcast WDM Networks TR May 14, 1996 Abstract HPeR-`: A Hgh Performance Reservaton Protocol wth `ook-ahead for Broadcast WDM Networks Vjay Svaraman George N. Rouskas TR-96-06 May 14, 1996 Abstract We consder the problem of coordnatng access to the

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

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

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

Life Tables (Times) Summary. Sample StatFolio: lifetable times.sgp

Life Tables (Times) Summary. Sample StatFolio: lifetable times.sgp Lfe Tables (Tmes) Summary... 1 Data Input... 2 Analyss Summary... 3 Survval Functon... 5 Log Survval Functon... 6 Cumulatve Hazard Functon... 7 Percentles... 7 Group Comparsons... 8 Summary The Lfe Tables

More information

Report on On-line Graph Coloring

Report on On-line Graph Coloring 2003 Fall Semester Comp 670K Onlne Algorthm Report on LO Yuet Me (00086365) cndylo@ust.hk Abstract Onlne algorthm deals wth data that has no future nformaton. Lots of examples demonstrate that onlne algorthm

More information

Information Sciences

Information Sciences Informaton Scences 79 (9) 369 367 ontents lsts avalable at ScenceDrect Informaton Scences ournal homepage: www.elsever.com/locate/ns Necessary and suffcent condtons for transacton-consstent global checkponts

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

Integrating External and Internal Clock Synchronization *

Integrating External and Internal Clock Synchronization * appeared n: Real-Tme Systems, Volume 12, Number 2, 123 171 (1997) Integratng External and Internal Clock Synchronzaton * ork performed @ Department of Computer Scence & Engneerng Unversty of Calforna,

More information

On the Exact Analysis of Bluetooth Scheduling Algorithms

On the Exact Analysis of Bluetooth Scheduling Algorithms On the Exact Analyss of Bluetooth Schedulng Algorth Gl Zussman Dept. of Electrcal Engneerng Technon IIT Hafa 3000, Israel glz@tx.technon.ac.l Ur Yechal Dept. of Statstcs and Operatons Research School of

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

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

On Selfishness, Local Information, and Network Optimality: A Topology Control Example

On Selfishness, Local Information, and Network Optimality: A Topology Control Example On Selfshness, Local Informaton, and Network Optmalty: A Topology Control Example Ramakant S. Komal, Allen B. MacKenze, and Petr Mähönen Department of Wreless Networks, RWTH Aachen Unversty, 52072 Aachen

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

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

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

Summarizing Data using Bottom-k Sketches

Summarizing Data using Bottom-k Sketches Summarzng Data usng Bottom-k Sketches Edth Cohen AT&T Labs Research 8 Park Avenue Florham Park, NJ 7932, USA edth@research.att.com Ham Kaplan School of Computer Scence Tel Avv Unversty Tel Avv, Israel

More information

AMath 483/583 Lecture 21 May 13, Notes: Notes: Jacobi iteration. Notes: Jacobi with OpenMP coarse grain

AMath 483/583 Lecture 21 May 13, Notes: Notes: Jacobi iteration. Notes: Jacobi with OpenMP coarse grain AMath 483/583 Lecture 21 May 13, 2011 Today: OpenMP and MPI versons of Jacob teraton Gauss-Sedel and SOR teratve methods Next week: More MPI Debuggng and totalvew GPU computng Read: Class notes and references

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Introduction. Leslie Lamports Time, Clocks & the Ordering of Events in a Distributed System. Overview. Introduction Concepts: Time

Introduction. Leslie Lamports Time, Clocks & the Ordering of Events in a Distributed System. Overview. Introduction Concepts: Time Lesle Laports e, locks & the Orderng of Events n a Dstrbuted Syste Joseph Sprng Departent of oputer Scence Dstrbuted Systes and Securty Overvew Introducton he artal Orderng Logcal locks Orderng the Events

More information

Private Information Retrieval (PIR)

Private Information Retrieval (PIR) 2 Levente Buttyán Problem formulaton Alce wants to obtan nformaton from a database, but she does not want the database to learn whch nformaton she wanted e.g., Alce s an nvestor queryng a stock-market

More information

Harvard University CS 101 Fall 2005, Shimon Schocken. Assembler. Elements of Computing Systems 1 Assembler (Ch. 6)

Harvard University CS 101 Fall 2005, Shimon Schocken. Assembler. Elements of Computing Systems 1 Assembler (Ch. 6) Harvard Unversty CS 101 Fall 2005, Shmon Schocken Assembler Elements of Computng Systems 1 Assembler (Ch. 6) Why care about assemblers? Because Assemblers employ some nfty trcks Assemblers are the frst

More information

An efficient iterative source routing algorithm

An efficient iterative source routing algorithm An effcent teratve source routng algorthm Gang Cheng Ye Tan Nrwan Ansar Advanced Networng Lab Department of Electrcal Computer Engneerng New Jersey Insttute of Technology Newar NJ 7 {gc yt Ansar}@ntedu

More information

All-Pairs Shortest Paths. Approximate All-Pairs shortest paths Approximate distance oracles Spanners and Emulators. Uri Zwick Tel Aviv University

All-Pairs Shortest Paths. Approximate All-Pairs shortest paths Approximate distance oracles Spanners and Emulators. Uri Zwick Tel Aviv University Approxmate All-Pars shortest paths Approxmate dstance oracles Spanners and Emulators Ur Zwck Tel Avv Unversty Summer School on Shortest Paths (PATH05 DIKU, Unversty of Copenhagen All-Pars Shortest Paths

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

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

On Correctness of Nonserializable Executions

On Correctness of Nonserializable Executions Journal of Computer and System Scences 56, 688 (1998) Artcle No. SS971536 On Correctness of Nonseralzable Executons Rajeev Rastog,* Sharad Mehrotra, - Yur Bretbart, [,1 Henry F. Korth,* and Av Slberschatz*

More information

Future Generation Computer Systems

Future Generation Computer Systems Future Generaton Computer Systems 29 (2013) 1631 1644 Contents lsts avalable at ScVerse ScenceDrect Future Generaton Computer Systems journal homepage: www.elsever.com/locate/fgcs Gosspng for resource

More information

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following.

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following. Complex Numbers The last topc n ths secton s not really related to most of what we ve done n ths chapter, although t s somewhat related to the radcals secton as we wll see. We also won t need the materal

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

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

More on the Linear k-arboricity of Regular Graphs R. E. L. Aldred Department of Mathematics and Statistics University of Otago P.O. Box 56, Dunedin Ne

More on the Linear k-arboricity of Regular Graphs R. E. L. Aldred Department of Mathematics and Statistics University of Otago P.O. Box 56, Dunedin Ne More on the Lnear k-arborcty of Regular Graphs R E L Aldred Department of Mathematcs and Statstcs Unversty of Otago PO Box 56, Dunedn New Zealand Ncholas C Wormald Department of Mathematcs Unversty of

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

VRT012 User s guide V0.1. Address: Žirmūnų g. 27, Vilnius LT-09105, Phone: (370-5) , Fax: (370-5) ,

VRT012 User s guide V0.1. Address: Žirmūnų g. 27, Vilnius LT-09105, Phone: (370-5) , Fax: (370-5) , VRT012 User s gude V0.1 Thank you for purchasng our product. We hope ths user-frendly devce wll be helpful n realsng your deas and brngng comfort to your lfe. Please take few mnutes to read ths manual

More information

UNIT 2 : INEQUALITIES AND CONVEX SETS

UNIT 2 : INEQUALITIES AND CONVEX SETS UNT 2 : NEQUALTES AND CONVEX SETS ' Structure 2. ntroducton Objectves, nequaltes and ther Graphs Convex Sets and ther Geometry Noton of Convex Sets Extreme Ponts of Convex Set Hyper Planes and Half Spaces

More information

Distributed Topology Control for Power Efficient Operation in Multihop Wireless Ad Hoc Networks

Distributed Topology Control for Power Efficient Operation in Multihop Wireless Ad Hoc Networks Dstrbuted Topology Control for Power Effcent Operaton n Multhop Wreless Ad Hoc Networks Roger Wattenhofer L L Paramvr Bahl Y-Mn Wang Mcrosoft Research CS Dept. Cornell Unversty Mcrosoft Research Mcrosoft

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

Secure Index Coding: Existence and Construction

Secure Index Coding: Existence and Construction Secure Index Codng: Exstence and Constructon Lawrence Ong 1, Badr N. Vellamb 2, Phee Lep Yeoh 3, Jörg Klewer 2, and Jnhong Yuan 4 1 The Unversty of Newcastle, Australa; 2 New Jersey Insttute of Technology,

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

Ramsey numbers of cubes versus cliques

Ramsey numbers of cubes versus cliques Ramsey numbers of cubes versus clques Davd Conlon Jacob Fox Choongbum Lee Benny Sudakov Abstract The cube graph Q n s the skeleton of the n-dmensonal cube. It s an n-regular graph on 2 n vertces. The Ramsey

More information

MULTIHOP wireless networks are a paradigm in wireless

MULTIHOP wireless networks are a paradigm in wireless 400 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 1, JANUARY 2018 Toward Optmal Dstrbuted Node Schedulng n a Multhop Wreless Network Through Local Votng Dmtros J. Vergados, Member, IEEE, Natala

More information

Optimizing Energy-Latency Trade-off in Sensor Networks with Controlled Mobility

Optimizing Energy-Latency Trade-off in Sensor Networks with Controlled Mobility Optmzng Energy-Latency Trade-off n Sensor Networks wth Controlled Moblty Ryo Sughara Rajesh K. Gupta Computer Scence and Engneerng Department Unversty of Calforna, San Dego La Jolla, Calforna 9293 Emal:

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