ComPASSsolved
solved 时间:2021-01-17 阅读:()
eScholarshipprovidesopenaccess,scholarlypublishingservicestotheUniversityofCaliforniaanddeliversadynamicresearchplatformtoscholarsworldwide.
LawrenceBerkeleyNationalLaboratoryLawrenceBerkeleyNationalLaboratoryTitle:UpdateonElectron-CloudSimulationsUsingthePackageWARP-POSINSTAuthor:Penn,G.
PublicationDate:10-09-2009PublicationInfo:LawrenceBerkeleyNationalLaboratoryPermalink:http://escholarship.
org/uc/item/0vp975ckUpdateonElectron-CloudSimulationsUsingthePackageWARP-POSINSTJ.
-L.
Vay,C.
M.
Celata,M.
A.
Furman,G.
Penn,M.
Venturini,LBNL,Berkeley,USAD.
P.
Grote,LLNL,Livermore,USA;K.
G.
Sonnad,U.
ofKarlsruhe,GermanyINTRODUCTIONAtPAC05[1]andPAC07[2],wepresentedthepackageWARP-POSINSTforthemodelingoftheeffectofelec-troncloudsonhigh-energybeams.
Wepresentherethelatestdevelopmentsinthepackage.
Threenewmodesofoperationswereimplemented:1)abuild-upmodewhere,similarlytoPOSINST(LBNL)orECLOUD(CERN),thebuild-upofelectroncloudsdrivenbyalegislatedbunchtrainismodeledinoneregionofanaccelerator;2)aquasi-staticmodewhere,similarlytoHEADTAIL(CERN)orQuickPIC(USC/UCLA),thefrozenbeamapproximationisusedtosplitthemodelingofthebeamandtheelec-tronsintotwocomponentsevolvingontheirrespectivetimescales;and3)aLorentzboostedmodewherethesim-ulationisperformedinamovingframewherethespaceandtimescalesrelatedtothebeamandelectrondynamicsfallinthesamerange.
Theimplementationofmodes(1)and(2)wasprimarymotivatedbytheneedforbenchmark-ingwithothercodes,whiletheimplementationofmode(3)fulllsthedrivetowardfullyself-consistentsimulationsofe-cloudeffectsonthebeamincludingthebuild-upphase.
BUILD-UPMODEFigure1:Sketchofthebuild-upmode.
Thedynamicsofelectronsisfollowedforathin(2-D)orthick(3-D)slicelocatedatagivenlocationinthelattice,undertheinuenceofalegislatedparticlebeampassingthroughtheslice.
Inordertofacilitatedirectcomparisonwithbuild-upcodeslikePOSINST[4,5,6,7],ECLOUD(CERN)orCloudland(SLAC),abuild-upmodeclasswasimple-mentedinWarp.
Inthismode,thedynamicsofelectronsisfollowedforathin(2-D)orthick(3-D)slicelocatedatagivenlocationinthelattice,undertheinuenceofalegis-latedparticlebeampassingthroughtheslice(Fig.
1).
RunsWorksupportedbytheUS-DOEunderContractDE-AC02-05CH11231,theUS-LHCLARP,andtheUS-DOESciDACprogramComPASS.
ThisworkusedresourcesofNERSC,supportedbytheUS-DOEunderContractDE-AC02-05CH11231.
jlvay@lbl.
govwereperformedwithWarpandPOSINSTfortheevolu-tionofanelectroncloudsliceinthemiddleofadipole.
TheaverageelectrondensityhistoryisgiveninFig.
2foraPOSINSTrunandthreeWarprunsin:(a)2-D,(b)3-Dwith4cellslongitudinallyandalengthof0.
2σz,and(c)3-Dwith16cellslongitudinallyandalengthof0.
8σz,whereσzisthebeamRMSlength.
Forthe3-Druns,pe-riodicboundaryconditionswereappliedlongitudinallyforeldsandparticles.
SnapshotsofcoloredelectrondensityplotsandverticalphasespacearegiveninFig.
3,takenatt=130ns.
TheseresultsdemonstrateaverygooddegreeofagreementforelectroncloudbuildsimulationsbetweenPOSINST,Warpin2-D,andWarpin3-D.
Figure2:AverageelectrondensityversustimefromPOSINSTandWarpinbuild-upmodesimulations.
QUASISTATICMODEWehaveimplementedaquasistatic[8]modeinWarp.
Inthismode,a2-Dslabofelectronmacroparticlesissteppedbackward(withsmalltimesteps)throughthebeameld(seeFig.
4).
The2-Delectronelds(solvedateachstep)arestackedina3-Darray,thatisusedtogiveakicktothebeam.
Finally,thebeamparticlesarepushedforward(withlargertimesteps)tothenextstationofelectrons,us-ingeithermapsoraLeap-Frogpusher.
Therstimple-mentationwasforacceleratorlatticestreatedinthesmoothapproximation.
Amoredetailedlatticedescriptionwasimplementedlater(seebelow).
Thismodeallowsfordi-rectcomparisonwiththequasistaticcodesHEADTAIL[9],QuickPIC[10],PEHTS[11]orCMAD[12].
Theparal-lelizationismono-dimensional(alongs)usingpipelining,similarlytoQuickPIC(seeFig.
5).
Wehavesimulatedane-clouddriveninstabilityinanLHC-likeringwithWarpinaquasistaticmode,andHEADTAIL.
Weusedthepa-rametersfromtable1inadrift(Fig.
6)andinadipole(Fig.
7).
SomeoftheparameterswerepurposelychosentoFigure3:Snapshotsofelectrondensityandverticalphasespacefrombuild-upsimulationsusing(left)POSINST,(middle)Warpin2-D,(right)Warpin3-D.
beunphysicallylarge,soastomagnifytheireffects.
Thetwocodespredictsimilaremittancegrowthunderthevar-iousconditions,withexcellentqualitativeagreementandgoodtoverygoodquantitativeagreement.
Wetentativelyattributethequantitativediscrepanciestodifferencesinim-plementationsincluding:adaptiveversusxedgridsizes,differenteldsolversandparticlepushers,differenteldinterpolationproceduresnearinternalconductors,slightlydifferentvaluesofphysicalconstants,etc.
Table1:Parametersusedforsimulationsofe-clouddriveninstabilitystudiesintheLHC.
circumferenceC26.
659kmbeamenergyEb450GeVbunchpopulationNb1.
1*1011rmsbunchlengthσz0.
13mrmsbeamsizesσx,y0.
884,0.
884mmbetafunctionsβx,y66.
,71.
54mbetatrontunesQx,y64.
28,59.
31chromaticitiesQx,y1000.
,1000.
synchrotrontuneν0.
59momentumcompactionfactorα0.
347*103rmsmomentumspreadδrms4.
68*102BOOSTEDFRAMEAPPROACHItwasshownin[13]thatitwaspossibletoperformsim-ulationsofelectron-driveninstabilitiesfromrstprinciples(e.
g.
usingstandardParticle-In-Cellmethods),atmuchre-ducedcomputingcostbyperformingthecalculationinaFigure4:Sketchofthequasistaticmode.
A2-Dslabofelectronmacroparticlesissteppedbackward(withsmalltimesteps)throughthebeameld.
The2-Delectronelds(solvedateachstep)arestackedina3-Darray,thatisusedtogiveakicktothebeam.
Finally,thebeamparticlesarepushedforward(withlargertimesteps)tothenextstationofelectrons.
Figure5:Sketchoftheparalleldecompositionforthequa-sistaticmode.
Thebeamisdistributedamongnslices,thatareuniformlyspreadamongNprocessors.
Usingapipeliningalgorithm,slicesonagivenprocessorarepushedfromonestationtothenext,withoutwaitingfortheslicesofthepreviousprocessorstoreachthesamestation.
suitableLorentzboostedframe.
Numericaldevelopmentsthatwereneededhavebeenimplemented,includinganewparticlepusherandeldsolver,andaredescribedin[14].
Specialhandlingofinputsandoutputsbetweentheboostedframeandthelaboratoryframearedescribedin[15].
TwoWarpcalculationsofanelectronclouddriveninstabilityFigure6:FractionalemittancegrowthfromWarp(red)andHEADTAIL(black)simulationsofane-clouddriveninsta-bilityindriftsofanLHC-likeringforanelectronback-grounddensityof1014m3for(top)ν=α=δrms=Qx=Qy=0,(middle)Qx=Qy=0,(bottom)parame-tersfromtable1.
showedverygoodagreement[14]betweenafullPICcal-culationinaboostedframeandacalculationusingthequa-sistaticmode,forsimilarcomputationalcost.
FURTHERDEVELOPMENTSWehaverecentlyaddedthecapabilitytouselinearmapstopushparticlesinacceleratorlattices,withinthequa-sistaticmodeandthefullPICmodeinaLorentzboostedframe.
GoodquantitativeagreementwasobtainedbetweenWarpusingthequasistaticmodeandCMAD[12].
SimilarcalculationswiththefullPICmethodinaboostedframeareinprogress.
Figure7:FractionalverticalemittancegrowthfromWarpandHEADTAILsimulationsindipolesofanLHC-likeringforthreeassumedinitialelectrondensities.
ACKNOWLEDGMENTSWethankG.
Rumoloforprovidingthesourceandin-valuablesupportforusingthecodeHEADTAIL.
REFERENCES[1]JLVayetal,ParticleAcceleratorConference,Knoxville,TN(2005),papersROPB006andFPAP016[2]M.
A.
Furmanetal,ParticleAcceleratorConference,Albu-querque,NM(2007),paperTUXAB03[3]D.
P.
Grote,A.
Friedman,J.
-L.
Vay.
I.
Haber,AIPConf.
Proc.
749(2005)55.
[4]M.
A.
FurmanandG.
R.
Lambertson,LBNL-41123/CBPNote-246,PEP-IIAPNoteAP97.
27(Nov.
25,1997).
Proc.
Intl.
WorkshoponMultibunchInstabilitiesinFutureElectronandPositronAccelerators"MBI-97"(KEK,15-18July1997;Y.
H.
Chin,ed.
),KEKProceedings97-17,Dec.
1997,p.
170.
[5]M.
A.
FurmanandM.
T.
F.
Pivi,LBNL-49771/CBPNote-415(Nov.
6,2002).
PRST-AB5,124404(2003),http://prst-ab.
aps.
org/pdf/PRSTAB/v5/i12/e124404.
[6]M.
A.
FurmanandM.
T.
F.
Pivi,LBNL-52807/SLAC-PUB-9912(June2,2003).
[7]M.
A.
Furman,LBNL-41482/CBPNote247/LHCProjectReport180(May20,1998).
[8]P.
Sprangle,E.
Esarey,andA.
Ting,Phys.
Rev.
Letters64,2011-2014(1990).
[9]G.
RumoloandF.
Zimmermann,PRST-AB5121002(2002).
[10]C.
Huang,V.
K.
Decyk,C.
Ren,M.
Zhou,W.
Lu,W.
B.
Mori,J.
H.
Cooley,T.
M.
Antonsen,Jr.
andT.
Katsouleas,J.
ofCom-put.
Phys.
217,658-679(2006).
[11]K.
Ohmi,SingleBunchElectronCloudInstabilityforaRoundBeam(Memo),19.
Nov.
2002.
[12]M.
Pivi,TheseproceedingsWE1PBI01.
[13]J.
-L.
Vay,Phys.
Rev.
Lett.
,98(2007)130405.
[14]J.
-L.
Vay,Phys.
Plas.
,15(2008)056701.
[15]J.
-L.
Vayetal,TheseproceedingsTU1PBI04.
LawrenceBerkeleyNationalLaboratoryLawrenceBerkeleyNationalLaboratoryTitle:UpdateonElectron-CloudSimulationsUsingthePackageWARP-POSINSTAuthor:Penn,G.
PublicationDate:10-09-2009PublicationInfo:LawrenceBerkeleyNationalLaboratoryPermalink:http://escholarship.
org/uc/item/0vp975ckUpdateonElectron-CloudSimulationsUsingthePackageWARP-POSINSTJ.
-L.
Vay,C.
M.
Celata,M.
A.
Furman,G.
Penn,M.
Venturini,LBNL,Berkeley,USAD.
P.
Grote,LLNL,Livermore,USA;K.
G.
Sonnad,U.
ofKarlsruhe,GermanyINTRODUCTIONAtPAC05[1]andPAC07[2],wepresentedthepackageWARP-POSINSTforthemodelingoftheeffectofelec-troncloudsonhigh-energybeams.
Wepresentherethelatestdevelopmentsinthepackage.
Threenewmodesofoperationswereimplemented:1)abuild-upmodewhere,similarlytoPOSINST(LBNL)orECLOUD(CERN),thebuild-upofelectroncloudsdrivenbyalegislatedbunchtrainismodeledinoneregionofanaccelerator;2)aquasi-staticmodewhere,similarlytoHEADTAIL(CERN)orQuickPIC(USC/UCLA),thefrozenbeamapproximationisusedtosplitthemodelingofthebeamandtheelec-tronsintotwocomponentsevolvingontheirrespectivetimescales;and3)aLorentzboostedmodewherethesim-ulationisperformedinamovingframewherethespaceandtimescalesrelatedtothebeamandelectrondynamicsfallinthesamerange.
Theimplementationofmodes(1)and(2)wasprimarymotivatedbytheneedforbenchmark-ingwithothercodes,whiletheimplementationofmode(3)fulllsthedrivetowardfullyself-consistentsimulationsofe-cloudeffectsonthebeamincludingthebuild-upphase.
BUILD-UPMODEFigure1:Sketchofthebuild-upmode.
Thedynamicsofelectronsisfollowedforathin(2-D)orthick(3-D)slicelocatedatagivenlocationinthelattice,undertheinuenceofalegislatedparticlebeampassingthroughtheslice.
Inordertofacilitatedirectcomparisonwithbuild-upcodeslikePOSINST[4,5,6,7],ECLOUD(CERN)orCloudland(SLAC),abuild-upmodeclasswasimple-mentedinWarp.
Inthismode,thedynamicsofelectronsisfollowedforathin(2-D)orthick(3-D)slicelocatedatagivenlocationinthelattice,undertheinuenceofalegis-latedparticlebeampassingthroughtheslice(Fig.
1).
RunsWorksupportedbytheUS-DOEunderContractDE-AC02-05CH11231,theUS-LHCLARP,andtheUS-DOESciDACprogramComPASS.
ThisworkusedresourcesofNERSC,supportedbytheUS-DOEunderContractDE-AC02-05CH11231.
jlvay@lbl.
govwereperformedwithWarpandPOSINSTfortheevolu-tionofanelectroncloudsliceinthemiddleofadipole.
TheaverageelectrondensityhistoryisgiveninFig.
2foraPOSINSTrunandthreeWarprunsin:(a)2-D,(b)3-Dwith4cellslongitudinallyandalengthof0.
2σz,and(c)3-Dwith16cellslongitudinallyandalengthof0.
8σz,whereσzisthebeamRMSlength.
Forthe3-Druns,pe-riodicboundaryconditionswereappliedlongitudinallyforeldsandparticles.
SnapshotsofcoloredelectrondensityplotsandverticalphasespacearegiveninFig.
3,takenatt=130ns.
TheseresultsdemonstrateaverygooddegreeofagreementforelectroncloudbuildsimulationsbetweenPOSINST,Warpin2-D,andWarpin3-D.
Figure2:AverageelectrondensityversustimefromPOSINSTandWarpinbuild-upmodesimulations.
QUASISTATICMODEWehaveimplementedaquasistatic[8]modeinWarp.
Inthismode,a2-Dslabofelectronmacroparticlesissteppedbackward(withsmalltimesteps)throughthebeameld(seeFig.
4).
The2-Delectronelds(solvedateachstep)arestackedina3-Darray,thatisusedtogiveakicktothebeam.
Finally,thebeamparticlesarepushedforward(withlargertimesteps)tothenextstationofelectrons,us-ingeithermapsoraLeap-Frogpusher.
Therstimple-mentationwasforacceleratorlatticestreatedinthesmoothapproximation.
Amoredetailedlatticedescriptionwasimplementedlater(seebelow).
Thismodeallowsfordi-rectcomparisonwiththequasistaticcodesHEADTAIL[9],QuickPIC[10],PEHTS[11]orCMAD[12].
Theparal-lelizationismono-dimensional(alongs)usingpipelining,similarlytoQuickPIC(seeFig.
5).
Wehavesimulatedane-clouddriveninstabilityinanLHC-likeringwithWarpinaquasistaticmode,andHEADTAIL.
Weusedthepa-rametersfromtable1inadrift(Fig.
6)andinadipole(Fig.
7).
SomeoftheparameterswerepurposelychosentoFigure3:Snapshotsofelectrondensityandverticalphasespacefrombuild-upsimulationsusing(left)POSINST,(middle)Warpin2-D,(right)Warpin3-D.
beunphysicallylarge,soastomagnifytheireffects.
Thetwocodespredictsimilaremittancegrowthunderthevar-iousconditions,withexcellentqualitativeagreementandgoodtoverygoodquantitativeagreement.
Wetentativelyattributethequantitativediscrepanciestodifferencesinim-plementationsincluding:adaptiveversusxedgridsizes,differenteldsolversandparticlepushers,differenteldinterpolationproceduresnearinternalconductors,slightlydifferentvaluesofphysicalconstants,etc.
Table1:Parametersusedforsimulationsofe-clouddriveninstabilitystudiesintheLHC.
circumferenceC26.
659kmbeamenergyEb450GeVbunchpopulationNb1.
1*1011rmsbunchlengthσz0.
13mrmsbeamsizesσx,y0.
884,0.
884mmbetafunctionsβx,y66.
,71.
54mbetatrontunesQx,y64.
28,59.
31chromaticitiesQx,y1000.
,1000.
synchrotrontuneν0.
59momentumcompactionfactorα0.
347*103rmsmomentumspreadδrms4.
68*102BOOSTEDFRAMEAPPROACHItwasshownin[13]thatitwaspossibletoperformsim-ulationsofelectron-driveninstabilitiesfromrstprinciples(e.
g.
usingstandardParticle-In-Cellmethods),atmuchre-ducedcomputingcostbyperformingthecalculationinaFigure4:Sketchofthequasistaticmode.
A2-Dslabofelectronmacroparticlesissteppedbackward(withsmalltimesteps)throughthebeameld.
The2-Delectronelds(solvedateachstep)arestackedina3-Darray,thatisusedtogiveakicktothebeam.
Finally,thebeamparticlesarepushedforward(withlargertimesteps)tothenextstationofelectrons.
Figure5:Sketchoftheparalleldecompositionforthequa-sistaticmode.
Thebeamisdistributedamongnslices,thatareuniformlyspreadamongNprocessors.
Usingapipeliningalgorithm,slicesonagivenprocessorarepushedfromonestationtothenext,withoutwaitingfortheslicesofthepreviousprocessorstoreachthesamestation.
suitableLorentzboostedframe.
Numericaldevelopmentsthatwereneededhavebeenimplemented,includinganewparticlepusherandeldsolver,andaredescribedin[14].
Specialhandlingofinputsandoutputsbetweentheboostedframeandthelaboratoryframearedescribedin[15].
TwoWarpcalculationsofanelectronclouddriveninstabilityFigure6:FractionalemittancegrowthfromWarp(red)andHEADTAIL(black)simulationsofane-clouddriveninsta-bilityindriftsofanLHC-likeringforanelectronback-grounddensityof1014m3for(top)ν=α=δrms=Qx=Qy=0,(middle)Qx=Qy=0,(bottom)parame-tersfromtable1.
showedverygoodagreement[14]betweenafullPICcal-culationinaboostedframeandacalculationusingthequa-sistaticmode,forsimilarcomputationalcost.
FURTHERDEVELOPMENTSWehaverecentlyaddedthecapabilitytouselinearmapstopushparticlesinacceleratorlattices,withinthequa-sistaticmodeandthefullPICmodeinaLorentzboostedframe.
GoodquantitativeagreementwasobtainedbetweenWarpusingthequasistaticmodeandCMAD[12].
SimilarcalculationswiththefullPICmethodinaboostedframeareinprogress.
Figure7:FractionalverticalemittancegrowthfromWarpandHEADTAILsimulationsindipolesofanLHC-likeringforthreeassumedinitialelectrondensities.
ACKNOWLEDGMENTSWethankG.
Rumoloforprovidingthesourceandin-valuablesupportforusingthecodeHEADTAIL.
REFERENCES[1]JLVayetal,ParticleAcceleratorConference,Knoxville,TN(2005),papersROPB006andFPAP016[2]M.
A.
Furmanetal,ParticleAcceleratorConference,Albu-querque,NM(2007),paperTUXAB03[3]D.
P.
Grote,A.
Friedman,J.
-L.
Vay.
I.
Haber,AIPConf.
Proc.
749(2005)55.
[4]M.
A.
FurmanandG.
R.
Lambertson,LBNL-41123/CBPNote-246,PEP-IIAPNoteAP97.
27(Nov.
25,1997).
Proc.
Intl.
WorkshoponMultibunchInstabilitiesinFutureElectronandPositronAccelerators"MBI-97"(KEK,15-18July1997;Y.
H.
Chin,ed.
),KEKProceedings97-17,Dec.
1997,p.
170.
[5]M.
A.
FurmanandM.
T.
F.
Pivi,LBNL-49771/CBPNote-415(Nov.
6,2002).
PRST-AB5,124404(2003),http://prst-ab.
aps.
org/pdf/PRSTAB/v5/i12/e124404.
[6]M.
A.
FurmanandM.
T.
F.
Pivi,LBNL-52807/SLAC-PUB-9912(June2,2003).
[7]M.
A.
Furman,LBNL-41482/CBPNote247/LHCProjectReport180(May20,1998).
[8]P.
Sprangle,E.
Esarey,andA.
Ting,Phys.
Rev.
Letters64,2011-2014(1990).
[9]G.
RumoloandF.
Zimmermann,PRST-AB5121002(2002).
[10]C.
Huang,V.
K.
Decyk,C.
Ren,M.
Zhou,W.
Lu,W.
B.
Mori,J.
H.
Cooley,T.
M.
Antonsen,Jr.
andT.
Katsouleas,J.
ofCom-put.
Phys.
217,658-679(2006).
[11]K.
Ohmi,SingleBunchElectronCloudInstabilityforaRoundBeam(Memo),19.
Nov.
2002.
[12]M.
Pivi,TheseproceedingsWE1PBI01.
[13]J.
-L.
Vay,Phys.
Rev.
Lett.
,98(2007)130405.
[14]J.
-L.
Vay,Phys.
Plas.
,15(2008)056701.
[15]J.
-L.
Vayetal,TheseproceedingsTU1PBI04.
- ComPASSsolved相关文档
- Spainsolved
- earlysolved
- denedsolved
- situationsolved
- picksolved
- producedsolved
老薛主机VPS年付345元，活动进行时。
老薛主机，虽然是第一次分享这个商家的信息，但是这个商家实际上也有存在有一些年头。看到商家有在进行夏季促销，比如我们很多网友可能有需要的香港VPS主机季度及以上可以半价优惠，如果有在选择不同主机商的香港机房的可以看看老薛主机商家的香港VPS。如果没有记错的话，早年这个商家是主营个人网站虚拟主机业务的，还算不错在异常激烈的市场中生存到现在，应该算是在众多商家中早期积累到一定的用户群的，主打小众个人网站...
Sparkedhost（$8/月）美国迈阿密AMD Ryzen高性能VPS;免费100G高防
sparkedhost怎么样?sparkedhost主机。Sparkedhost于2017年7月注册在美国康涅狄格州，2018年收购了ClynexHost，2019年8月从Taltum Solutions SL收购了The Beast Hosting，同年10月从Reilly Bauer收购了OptNode Hosting。sparkedhost当前的业务主要为：为游戏“我的世界”提供服务器、虚拟...
注册做什么96%可以干啥，常用的7个常用的国内国外域名注册服务商_云服务器可以干什么
日前，国内知名主机服务商阿里云与国外资深服务器面板Plesk强强联合，推出 阿里云域名注册与备案、服务器ECS购买与登录使用 前言云服务器（Elastic 只需要确定cpu内存与带宽基本上就可以了，对于新手用户来说，我们在购买阿里云服务申请服务器与域名许多云服务商的云服务器配置是弹性的 三周学会小程序第三讲：服务 不过这个国外服务器有点慢，可以考虑国内的ngrokcc。 ngrokcc...
solved为你推荐
-
服务器空间租用租个服务器 一年多少钱美国主机租用国外服务器提供商有哪几家比较好的。。。域名服务域名服务器是什么？美国vps租用香港VPS：那里有租用香港VPS或者美国的VPS免费域名空间求速度快的 免费空间 带域名的 谢谢大家免备案虚拟空间免备案的虚拟主机空间，买了以后会强制备案不？虚拟空间哪个好国内哪个空间商（虚拟主机）最好美国网站空间美国空间做什么网站好？虚拟主机mysql虚拟主机的数据库有哪些虚拟主机试用购买虚拟主机为什么商家会让你试用和测试？