1、超大规模实时数仓架构挑战与实践解析技术创新 变革未来数据库发展RDBMS SQL+OLTP Data warehouse Data Cube ETL+OLAP GraphVectorTextStructured DataHeterogeneous DataRDBMSNoSQL/NewSQL DB Multi-Model+HTAP Structured DataStructured DataTime SeriesSpatial Data阿里巴巴 数据库OLAP-OLTP On-Line Transaction Processing Short-lived tr(ns()tions(Typi)(l
2、ly Sm(ll d(t()ess footprint(Often time Repetitive oper(tions On-Line nal-tical Processing Lon-runnin-queries Complex join oper(tions Explor(tory queries th(t m(y()ess(l(r-e(mount of d(t(数据库发展 架构数据库发展 核心技术组件数据库发展 更高的应用诉求数据库发展技术挑战整体架构支持任意条件组合的高并发低 延时查询支持大吞吐数据写入兼容MySQL生态支持高可用（在线升级/扩缩容，单点故障应用透明）执行调度Work
3、loadPer Query FairScheduler Per Task WeightScheduler场景混合负载10-20XPerformance Improve内存管理内存调度：统一内存结构，定长块内存池化,Binary Process，弹性分配，支持落盘自研优化器GPU引擎JIT代码生成和IR层优化支持CPU和GPU运行时加速库Snappy解压高速排序高并发显存管理CPU-GPU堆外内存传输SSD至显存直传Compute NodeGPU引擎 性能SQL OptimizationMemory optParameterizationMonitoringTuningAuto-adminDia
4、gnosisRestoreAnomaly DetectionSecure proxyProtectionPatchIdentificationThreat DetectionAlertElasticitySchedulingResource predictionBackup RestoreOperationsMonitoringHA、Disaster RecoveryUpgradeExpansion and contractionWhole stage OptResource managementDatabase Self-DrivingMetadataTask manageData coll
5、ectionSecureReliableCost-effectiveEfficientML Algorithms自动化运维新硬件RDMANVM3DX PointOpen-Channel SSDAPPSFile systemF T LOpen Channel 10 LibraryNVMe DriverOpen Channel FirmwareSSD ControllerGPU/FPGAForrester-ContendersGartner Niche Player业界认可标准测试-TPC-DS Realistic table scaling Real world data content Non-uniform distributions Complex relationships(Fact to Fact)TPC-DS R&port&VLDBVLDB 2019AnalyticDB:Real-time OLAP Database System at Alibaba CloudTPC-DS
