MVCC，支持schema和partitionevolution

支持主流SQL引擎

workloadAISpark態(tài)，提供ACID事

存），深度使用Flink構(gòu)可見(jiàn)，各有側(cè)重，企業(yè)面臨多種Lakehouse 公有 私有云/數(shù)據(jù)中 AnalysisCosmos

ConnecttotheDataMattersAutomateend-to-enddatarightsrequestsandEnrichtechnicalmetadatawithbusinessandoperationalmetadataforfullvisibility.

Technical&Business3rdPartyDataPII&PIDataIP

Automaticallyfind,classify,andmapallofyourdata-everywhere.AutomaticallyclassifymoreofdatainmoreSimplifyandautomatedatalifecyclemanagementfromcollectiontodestruction.HiveMetastore DWCatalog

Model

從元數(shù)據(jù)層面上達(dá)到Datalake/

AINext-GenDataCatalogistheCoreinNewOpenDataUnifiedRESTUnifiedUnifiedFunctionalityUnifiedRESTUnifiedUnifiedCoreobjectConnection

IcebergIcebergREST Tabular Non-tabularGravitinoVirtual

PythonUnifiedTabular

Schema:{name:stringcomment:properties:Schema:{name:stringcomment:properties:map<string,Schema:{name:stringcomment:properties:map<string, Table:name:stringcolumns:Colum[]partitioning:Transform[]distribution:DistributionsortOrder:SortOrder[]indexes:Index[]

Fileset:name:stringstorageLocation:stringtype:TypeUnifiedAccessControlGravitinoOpenlineageREST

JDBC

GravitinoIcebergRESTCatalogIceberg此外，用戶(hù)依然可以通過(guò)GravitinoRESTAPIIcebergcatalog，實(shí)現(xiàn)擴(kuò)展功能，包括權(quán)限、Tag、血緣等。

GravitinoUnifiedRESTAPI

GravitinoIcebergRESTAPIIceberg

GravitinoRESTAPIIRCAPImetadataAPIAPI.獲得擴(kuò)展的功能，如Tag、血推薦使用JDBCOAuthCredentialvending(支持AWSLakehouseHMS為歷史原因或業(yè)務(wù)需求，可能使用不同版本的HMS。這種多版本共存的情況往往導(dǎo)致開(kāi)發(fā)和運(yùn)維團(tuán)隊(duì)GravitinoAPIHiveMetastore，TableIceberg:8kactivetables,4kareforHive:80kactivetables,57karefor

S3Iceberg:Hive:

…

GravitinoIcebergRESTHive JDBC …From“PluggableCatalogs”the“RESTIceberg

IcebergPinterestIRCisrunningonGravitino0.7CustomizedeventlistenerfortableCustomizedrequestrewritefortablegovernance(WIPAllengines(Spark,Trino,Flink)areonboardedonIRCGravitinoClusterAdhoccluster:5Scheduledcluster:5UsageDailyGravitinoeventsisSingleIRCinstanceloadtableQPS<1,durationp95<SingleIRCinstanceupdatetableQPS<1,durationp95<Thesmoothtransitionexperience,thankstoopenIcebergRestCatalogspec,solidGravitinoimplementationandgreatcommunitysupport.TBDSMetaservice（poweredbyGravitino）統(tǒng)一利用TBDSMetaservice的多引擎支持，TBDS的湖倉(cāng)一體得益于TBDSMetaservice對(duì)Iceberg等數(shù)據(jù)湖表格式的良Hive在人工智能時(shí)代，數(shù)據(jù)成為了創(chuàng)新和決策的核心資源。OpenDataCatalog景下扮演著至關(guān)重要的角色，它不僅是數(shù)據(jù)的管理者，更是AI發(fā)展的催化劑。

管理好AI非結(jié)構(gòu)化數(shù)據(jù)海量數(shù)據(jù)“大模型預(yù)訓(xùn)練/實(shí)現(xiàn)DataAIPBHDFS

FilesetFilesetmgmtREST,Java,HCFSKerberos

AccesspolicywriteCloudstoragesupportandcredentialvendingAuditing

IRCupgradetolatestModelmgntMetadataaccessDatalineageJDBCsourcessupportinv0.9(Q2

Lance,FlussUDFTableTablePythonfsspecCentralizedaccess

v0.8（Ja