广义线性反演在网络层析成像中的应用-通信与信息系统专业论文.docxVIP

AbstractAbstract Abstract Abstract With the rapid developme毗of山e In啪％thc stn】ctlll．e of the network h舔 changed thoroughly．It is necessary t0 get into the inner plq船r哆of the speci丘c network to dcsi萨，∞n_缸oI and m弛age it successfllHy．Tra伍c Man奴and link一1evel packet 10ss 船te arc the impo姗t p缸锄etcrs t0 indentify the perf曲mance ofⅡle network．Bm it is hard to me硒ure tIlem d由ecny雒the mtemct is becolIling massive，distributed aIId he咖gencous． Ne懈ork T0mo鲫hy is a妒渤砌ng techIlolog)r to i11fcr n他ne脚0rI【parame觚 through iIldirect me硒u∞Ⅱ屺nt．MLE is widely used in netwofk tomo鲫hy field t0 硒surc thc acclIracy of tIlc result，but it bfings in the problem of s讪mty and us a_bilit)r wllich is thc main botneneck to push the ne柳ork tomo乒aphy estilIlation iIlto practical use．Thc s协ility aIld tlle usdbmy of the ne脚ork tomo掣aphy t葩honology is dicusscd hef；e．Wb introduce龇GU alg鲥thm into network tomo鲫hy field to estimatc oD 仃a筋c锄d nnk．1cvd packet loss rate．S0me improvemet arc made iII GU tO makc me algorithmmo心stable锄dmorepractical， 0D昀伍c esti脚tion is one of tlle impon锄t aspects of ne哪ork tomography．We 朋叩h弧i∞on the existing alg讲inl玎勰孤d propose a new algo血tlIn whkh is based on 仃a伍c covaIi锄ce．1Ks novel alg嘶thm is pmved t0 be e硒enct aIld fast iⅡlpl锄ented with GU她d the slipe—window scbeme．We use GU to estiInate ttle 1M in large scak ne啪ork．Di船rcnt constrains bascd on inmal面0r锄d cons仃ains based on panial me硒ureI咖t arc discussed．A new method based on statc predictiOn and another method in which llistory mean is combined witll graVity Inodel is propoScd．Under the implementati伽of tbe di圩brent constmins and incoIpor曲甜with panical me雒u化ment， we get a讲圮aling resl】】t，wllich is better mat general伊aVity metllod． LiIll(·level packet 10ss rate is孤iInponaIlt pa姗嘲of ne哪ork pemmnaIlce．h is valu a_ble in m锄agement of t}le network．GLI is in旬roduced into this field．A new method is proposcd t0 estimate lillk_level packet loss ratc under general SPT while exiting metbod can ollly solve this problem under constallt biIlar