unidata | home - scanned by camscanner...synoptic meteorology textbook (lackmann). the manual goes...
TRANSCRIPT
Scanned by CamScanner
B.ProjectSummary:
WeproposetoobtainanewLinuxmachine,toshadowandthenreplace(uponfailure)
ourcurrentagingmachine(weather.rsmas.miami.edu),asignificantgeographicalnodein
Unidata’sIDDnetwork.Weatherisalsoalargepublicdatarepository,ashowcasefor
UnidatatechnologiesbothtotheMiamiresearchcommunityandtotheonlinecommunity
atlarge.Thecontinuityofweather.rsmasisimportant:Forinstance,itisadvertizedinmy
YouTubevideosenthusingaboutUnidata’sIDVandTheMapesIDVCollection.That
CollectionnowincludesvisualizationsforournewLabManualsupplementtoapopular
synopticmeteorologytextbook(Lackmann).ThemanualgoesonsalesoonfromAMSpress,
andmanyusersoftheLackmanntextbook.Aweather.rsmasfailureasfallclassesstart
hittingtheserverwouldbeembarrassingandinconvenient,sowewanttoobtainahot
sparethissummer.Wehavehadonescare,andthereisaconcerningupswinginopaque
RAIDcontrollerwarningmessages.
Wewillputtheagingmachine(whichwillberefurbishedatMiamiexpenseuponfirst
failure)togooduse.Forintellectualmerit,itwillserveasabackupofoursubstantial(~50
TB)communitydatarepository,includinglargedatasetofferingsthatshowcaseUnidata
technologiestoclimate,oceanography,field-campaign,andotherpotentialuserbasesat
themarginofUnidata’straditionalrealtimeweathercommunity.Communityservice
functionssuchasInternetDataDistributionwillbesustainedandenhanced,asnew
capacitymayallowlongerrollingarchivesandlargernewLDMfeeds(GFS0.25deg,HRRR,
GOES-R).RedundantcapacitywillalsobeusedasatestplatformforthePI’s(collaborative
withUnidata)EarthcuberesearchprojecttointegrateUnidata’spowerfulIDVwithother
technologies(notebooksandrepositories),tomakeUnidatamorerelevanttostatistical
andclimaticresearch.High-capacityinternaltesting(withgoodwill)ofanysoftwarestack
iscrucialtoavoidspoilingthebroadercommunity’spatienceviaadisappointingrollout.
Thesparecapacitywillfacilitatesuchtesting,usingclassroomsassynchronousloads.
Thesestudents,atthebleedingedgeofJupyterNotebook-IDVintegration,willalsobe
entrainedintotheproductionofpromotionalmaterialssuchasthePI’sYouTubevideos.
C.ProjectDescription
WeproposetoobtainahotspareandreadyreplacementforouragingLinuxserver
(weather.rsmas.miami.edu).Thenewmachinewillbesetupasaquasi-cloneoftheexisting
machine,includingadailysynchronizedcopyofmanyTBofhighvaluedata,whichwas
itselfsetupasamotherlode.ucar.educlone,forrapidfailoverandminimalmaintenance
burden.Unidatastaffwhohelpmanagetheresourceforcommunityaswellaslocalgood
willhavecapableMiamisysadminstaffassistance,whereagoodworkingrelationship
existsalready.After4yearsofservice,theweather.rsmasmachineisgivingoffmoreand
moreRAIDwarnings.Itcouldstillhaveyearsofservicelife,butredundancyisneededto
preventamajorserviceinterruptionuponfailure.Infactwehavehadonescarealready
(twodrivesfailedintheRAID,requiringanervousweekortwotorebuild).Thevirtuesof
suchredundancyforreliability,includingcrossmounteddiskswithadailysynchronized
copyofmanyTBofthehighestvaluedata,willnotbebelaboredhere,buttheyare
significant.
Inthemeantime,beforefailureoftheoldsystem,theoverlappingcomputerandstorage
capacitywillbehighlyvaluedandputtogoodusetowardintellectualmerit,community
dataservices,andpromotionofUnidatatechnologiestobothresearchersinadjacentbut
under-engagedcommunities(climate,oceanography)andtheworldatlarge(through
YouTubevideos,andthehoped-forreliabilityofthePI’snewtextbook-relatedIDV
exercises,intheupcomingLackmann,Mapes,Tylelabmanual,see
https://tinyurl.com/LMTmanual.
SpecialConsiderationsfor2017
ThisproposaladdressesSpecialConsiderationitemsdetailedintheRFP
http://www.unidata.ucar.edu/community/equipaward/RFP2017.html,especiallythese:
• Production of online training materials (code notebooks, video tutorials, online documentation, or similar resources) that can be shared with the Unidata community
• Installation of equipment that provides student access to modern visualization tools such as AWIPS, IDV, or Jupyter notebooks
Inparticular,PIMapeshasasubstantialpresenceinonlineIDVtrainingmaterials.
BesidestheLMTLabManualdescribedabove,hehasproducedmanyIDV-basedresources,
underthenameMapesIDVCollection
(http://www.rsmas.miami.edu/users/bmapes/MapesIDVcollection.html).Theseresources
includewell-receivedintroductoryYouTubevideos,butalsoasubstantialdeeperbodyof
tools.Alibraryofcuratedbundlesisthemostvisible.Myself-updatingIDVpluginincludes
notonlyaFavoritesfolderpointingtothosebundles,butalsoadvancedformulasfor
statistics,filtering,interactivescatterplot,publication-qualityverticalsectionsona
pressurecoordinate,andmanyotherusefuladditionstotheIDV’sbasiccapabilities,aimed
atkeepingitrelevantformodernresearchandnotjustweatherdepiction.ThePIiseager
tokeepuptheeffort,butworriedaboutthefrailtyofitshardwareunderpinnings.Cloud
solutionsdonotseemreadyyettoreplace“baremetal”ofLinuxrackshostedlocally,but
thefrailtyofmycurrentmachine(froma2013Unidataequipmentgrant)iswearingonme.
AnothertrulynovelcontributionbythePI’sgroupistheIDVnotebook(seeexamples
andcodeathttps://github.com/suvarchal/JyIDV).ThisallowsanIDVsessiontobe
operatedandannotatedfromaJupyterNotebookenvironment(http://jupyter.org).This
couplingtransformsanIDVsessionfromapersonalexperienceintoadocumentedand
repeatableexploration.ThePIhascollaboratedwithcurrentandformerUnidatastaff,
includingsomeconsultingpaidatmyownexpenseaswellasanEarthcubegrantwithco-PI
YuanHoatUnidata(themainIDVdeveloper),tohelpfirmupthiscoupling,andbringIDV
stronglyintotheworldofJupyter.Thesetechnologieswillneedexplaining,orreally
showing,andcontinuityoftheplatformfortheirdevelopmentandtestingisessential.
Figure1:IDVNotebook,aninnovationfromthePI’sgroup(https://github.com/suvarchal/JyIDV).It
couplestheIDVtoJupyternotebookstocreateasaved,replicable,literate(human-reader-oriented)
wrapperaroundtheotherwiseephemeralGUIexperience.
Basedpartlyonthisprototype,thePIreceived(collaborativewithUnidata)anNSF-
EarthcuberesearchgranttofurtherintegrateUnidata’sIDVwithnotebooksand
repositories,tobuildasoftwarestackthatismuchmoreusefulthanitspartstoanalysis
challengesintheBigDataage.Specifically,ourgoalistofulfillthevisionofVisualizationfor
AlgorithmDevelopment,theoriginalbasisfortheIDV’sunderpinnings(VisADisinfactthe
correspondingJavalibraryname).Algorithmsareessentialtoextractmeaningfrombig
datasets,butedgecasesoftheirperformanceneedtobeexamined(Visualized)carefully,in
ordertofurthertheirDevelopment.Inthissense,wearefulfillingveryelementalgoalsat
thebaseofUnidata’shistory.Astablecomputingplatformwithsomeavailablecapacity
willallowbothresearchandclassroomtestingofthese“Unidata-plus”softwarestacks.
Classroomtestingisespeciallypowerful:Besidesbeingeducationalforthestudents,it
quicklyexposesjargonandworkaroundsthatexpertshaveceasedtonoticetheyareusing.
Researchusesoftheweather.rsmas.miami.eduRAMADDArespositorywillbesustained
bythesafebackupandenhancedwithanysurpluscapacity.Inadditiontotherealtime
weatherdatastreams,wehostanumberofdatasetsthatutilizeRAMADDA’sservicestoaid
communitygoals.Readersareinvitedtobrowseathttp://tinyurl.com/Miami-RAMADDA-
datasets,buthereisasample:
• Largenestedhurricane“naturerun”simulationdatasets,withassociatedIDVbundles–valuableforshowcasingmultiscalehurricanedynamics,andforcomputingandillustratingvortexdynamicsquantitiesonarealistic3Ddataset.
• Huge“Giga-LES”(2048x2048x128)convectionsimulationoutputs,andcoarse-
grainings(like256x256),allwithassociatedIDVbundlesdisplayingbothsets--valuableforturbulencestudies,allowingstatisticstobesuperposedonthegridpoint-leveldetailstheysummarize.
• DatasetsfromtheDYNAMO-AMIEfieldcampaign,allgatheredtogether,withIDVbundlesthatoverlayandjuxtaposethem.Whilethesedatacanbedownloadedonebyonefrom“catalogs”elsewhereontheweb,dataintegrationissomucheasierwhentheyarehostedonarepositorywithservices.
• EnergyandWaterCycledatasetsarchive,synthesizingmultipleestimatesofdifficulttoobservequantitiesinoneplaceforeasycomparison(incollaborationwiththeNASANEWSprogramanditsresearchcommunity).
• Oceanographicdatasets(comingsoon)
AcommonthemeisthatresearchcommunitiesatthemarginsofUnidata’s
traditionalmeteorologyrootsarebeingeducatedaboutthepowerofthesetools,andsome
ofthelimitationsoftheold“download,folderize,andwritecode”modelofsciencefordata
integration,anincreasinglyimportantpartofresearch.
Personnel,expertiseandinstitutionalsetting
ProfessorMapesisaneducatorandresearcherwithexperienceandskillsincomputing
andscientificprogramming.HeservesontheIDVsteeringcommittee,anduntilrecentlyon
Unidata’sStrategicAdvisoryCommittee.HeisalsoPIonanEarthCubeproject,withUM
oceanographerMohammedIskandaraniandUnidataco-PIYuanHo,designedtokeepthe
strengthsoftheIDVrelevantintheageofPythonandJupyternotebooks,assketched
brieflyabove.
TheDepartmentofAtmosphericScienceconsistsof12facultyandmorethan50
students,undergraduateandgraduate.Thereisnocentralcomputingforthisgroup.The
RosenstielSchoolofMarineandAtmosphericSciences(RSMAS)hasafacultyofabout80,
andtotalresearchcomputerusersinthehundreds.TheRSMASComputingFacility(RCF)
hasUniversity-leadingskillsduetothedata-intensivenatureofourfield.Inaddition,the
UniversitymaintainsaCenterforComputationalScience(CCS).Togetherthesefacilities
andexpertiseandservices(frompowertobandwidthtomaintenance)aremorethan
adequatetosuccessfullydeployandmaintaintheproposedpurchase.
D.BudgetThetotalrequestedfundingforthisproposalis$18,649,acapitalequipmentlineexempt
fromindirectcosts.Thisquotewasbasedontheexistingweather.rsmas.miami.edumachine,
andusesthesamevendor.Again,ourgoalislow-maintenancecontinuationofexisting
services,enablingasafeplatformforresearchandoutreachnovelties,whilecontinuingto
fosterthecoreUnidata-relatedcommunitygoods.
Server-roomrackspace,electricityandcoolingwillbeprovidedbytheSchoolasan
expenseofitsresearchandinstructionalmission.Optionally,theUniversityhasofferedto
hostthemachineatitsUniversity’sCenterforComputationalScience,inacrosstown
facilitywithmajorregional-infrastructuregradeinternetlinesandelectricalandphysical
security.Thisdecisionwillbemadeuponrecommendationforfunding,inconsultation
withexpertsatUnidata,theSchool,andtheUniversity.ThePIcontributestimeandtravel
costs,includingfrequentvisitstoUnidata,andisactive(tothepointofapersonalmission)
intransferringknowledgeaboutUnidatatechnologiesandservicestofaculty,studentsand
otherinterestedmembersoftheUM-RSMAScommunityandbeyond.Thistimeis
consideredaninvestmentinmyteachingandresearch.
E.ProjectMilestonesWeexpectatimelinelikethefollowing:
•May2017–Notificationofawardandpurchaseofquotedmachine.
•June2017–Installequipmentandsoftware,perhapsliterallycloninganimageofexisting
weather.rsmasasastartingpoint.ConsiderbroadeninganddeepeningIDD
subscriptions(LDMfeeds)toutilizeincreasedcapacities.
•July2017:Prepareandtestfailoverschemesforweather.rsmas.miami.eduIP.
•Summer2017:LMTtextbooklabmanualreleased...Inpreparationforautumnsemester
synopticmeteorologycourses,retestallmaterialsforreadinessandpreparebackup
plans,withAlbany(viacollaboratorK.Tyle)asafurtherfailsafe.Prepareandpostmore
YouTubevideosandothertrainingmaterials.
•Autumn2017:Problem-solveanyLMTtextbookmanualissuesastheyarise.
•Ongoing:Utilizeserverforhostingofadditionalopenlyshareddatasourcessuchashigh-
resolutionoceangrids,testingandsharingimprovementstotheMapesIDVCollection,new
capabilities(suchasIDVNotebook),etc.
•April2018–Submitafinalreport,intheformofashortarticle,totheUnidataProgram
Centerthatdescribes:1)Lessonslearnedaboutcomputersetupforfailureproofing,2)
ActivitiesandservicestotheMiamiresearchandbroadercommunity,madepossibleby
theavailabilityofoverlappingcapacity,3)Assessmentsonapproachesthat
worked/failedinordertoimprovetheprogram,and4)Recommendationsandbenefits
ofparticipationintheprogram.