加快卡塔尔从数据中心效率到基础设施和技能的人工智能准备.pdf

《加快卡塔尔从数据中心效率到基础设施和技能的人工智能准备.pdf》由会员分享，可在线阅读，更多相关《加快卡塔尔从数据中心效率到基础设施和技能的人工智能准备.pdf（20页珍藏版）》请在三个皮匠报告上搜索。

1、Accelerating Qatars AI Readiness form Data Centre Efficiency to Infrastructure and SkillsetTable of contentsData Centres Infrastructure0102MEEZA AcademyMEEZA.AI0304Data Centres 01AI is Transforming the Digital LandscapeData Centres Driving Qatars Digital Economy Foundation of Digital Economy:DCs as

2、the backbone of robust&secure digital infrastructureScalable and Reliable:Hosting advanced applications like AI,IoT,and Big DataEconomic Impact:Accelerating business agility and innovation&driving foreign investments MEEZAs Role in the Digital Economy:Supporting Qatar National Vision 2030Data Centre

3、s Land and PowerAI Capacity of All DC PowerNew AI ChallengesNot All Solutions are created EqualWorkload and Use-Case DependencyCourtesy of Infrastructure02Worlds 1stGenerative LLMPre-trained on CPUsAI is split into Training and Inference workloads,each with different implications on data center requ

4、irements.Training vs.InferencingGPUs&Advanced Infrastructure Training WorkloadHeavy GPUsTraining TIME is importantLow Latency and high resiliency not requiredE.g.Chat GPT-4Extreme kW/Rack(70 to 120 kW/rack)Inference WorkloadMix of CPU and GPUPredication TIME is importantDemand for low Latency&high r

5、esiliency E.g.Predictive analytics,Fraud detection.Moderate kW/Rack(40 to 70 kW/rack)Edge AI Hardware Market Size,2018-2028Increasing Demand for GPUs:TrainingLarge Clusters for Large Models10,000 GPUs:Train LLMs1,000 GPUs:Inferencing,ML apps,multiple users100 GPUs:Fast workloadsSlice Usage of GPUs t

6、o optimize ROITime slideMulti-instance GPUVirtual GPUMEEZA Academy03Prepareanewgenerationofexpertssupporting the growing demands in the ITsector and contribute to MEEZAs mission tobecome the preferred managed IT servicesand solutions provid