SAS Viya. Примеры проектов на новой платформе
SAS 9.4 high-level architecture Variuos Data Sources Streaming data SAS Event Stream Processing SAS Micro Analytic Server SAS High-Performance Analytics SAS LASR Analytics SAS Metadata Cluster PostgreSQL SAS Web Infrastructure Platform Data Server SAS Desktop clients (Java,.NET) SAS Mobile clients Data Warehouse SAS Embedded Process Clustered SAS Web Application Servers (Applications and Services) SAS Web clients Analytical Data Warehouse Data Tier SAS Grid Compute Nodes Metadata Tier Middle Tier Client Tier Server Tier
Parallel & Serial, Pub / Sub, Web Services, MQs Source-based Engines Customer Intelligence Analytics In-Stream In-Hadoop In-Database In-Memory Runtime Engine Cloud Analytics Services (CAS) UAA UAA UAA Microservices Data Source Mgmt Folders etc BI GUIs CAS Mgmt Log Analytics GUIs Data Mgmt GUIs Query Gen Env Mgr Model Mgmt Audit Solutions APIs Business Visualization Risk Management! Fraud and Security Intelligence Data Management Platform
SAS Viya 3.4
Today s architecture SAS 9.4 M5 products and UIs Customer-written code Customerwritten code SAS Viya products and UIs Metadata (WIP)-based mid-tier SAS 9.4 M5 SAS 9 bridge to SAS Viya Microservices-based mid-tier SAS Viya MVA runtime (full functionality) LASR and / or HPA runtimes Other runtimes (ESP, In-Database) Viya MVA runtime (minimal functionality) CAS runtime SAS/Connect Server SAS Viya bridge to SAS 9
PURE SAS VIYA USE CASE CHINA SEMICONDUCTOR MANUFACTURING COMPANY Use Case #1 Quality Control Process Efficiency Efficient in-memory processing Reduce time to run weekly QC checks Utilize SAS Visual Analytics to review results Use Case #2 Image Processing and Deep Learning Read in images of wafers Apply new SAS Viya CNN algorithms to identify flaws in wafers Use Case #3 Open Source Interface Use Python as an interface Copyr i g ht 2014, SAS Ins titut e Inc. All rights res er ve d.
Introduce defect image type Purpose : Detect on the wafer with spatial patterns is usually a cube for the identification of equipment problems or process variations. Usually, defect image classification capability for different defect types need over 90%accuracy. The challenge is redetection of small size defect, number of classified defect type. Do defect types classification is key objectives on defect image application. In common, having around 10-15 different types and how to defect unknow using learning process will be other key objectives. Classification which types
Parallel, Web Services, MQs PURE SAS VIYA Viya ARCHITECTURE USE CASE SAS Source-based Engines Analytics In-Hadoop In-Database Cloud Analytics Services (CAS) In-Memory Engine UAA UAA UAA Data Source Mgmt Folders etc. Microservices BI GUIs CAS Mgmt Log Analytics GUIs Data Mgmt GUIs Query Gen Env Mgr Model Mgmt Audit Solutions APIs Business Visualization Data Management Platform Infrastructure KEY BENEFITS 1) Access to Hadoop 2) Fast In-Memory data processing 3) Application of Deep Learning (CNN) algorithms 4) Use Python as the primary interface Copyr i g ht 2016, SAS Ins titut e Inc. All rights res er ve d.
SAS Germany s first opportunity for Viya Customer Profile Germany s largest retail company About 25% market share in Germany 11.500 stores and supermarkets ~ 350 000 employees ~ 50 billion euro sales volume in 2016 Long-term SAS customer SAS DI Server, Enterprise Miner, Data Loader, STAT/ETS/IML Introduced Hadoop to their landscape 2 years ago - SAS Germany s first customer on MapR
SAS Viya Business unit (Database Marketing) got interested in Viya at the end of 2016 Data scientists mainly use programming clients (SAS Display Manager, EG) Little use of SAS Enterprise Miner GUI SAS VDMML seemed to be a good fit in terms of analytical capabilities Everything client need seemed to be available EDEKA want to use some of the new features (Factorization Machines) Some data scientists with Python know-how Meet them where they are
Viya Single Node Recommend to use a separate machine SAS 9.4 M5 on the SAS9 side to leverage seamless integration tools - Data preparation still done on SAS9 side SAS Viya Solution scenarios 20.05.2017, v1.0 (MAK) Option A: SAS 9.4 Plattform & SAS VDMML auf Single -Server (SMP) SAS 9.4 M5 SAS Analytics Server 4 Cores / 32-64 GB RAM SLES 11 SP3 SAS 9.4 M5 Metadaten Server SAS Enterprise Miner Server - Lizenz läuf im Dezember aus SAS Data Surveyor for SAP - SAS Data Integration Server SAS/GRAPH; SAS/STAT SAS/ETS; SAS/IML SAS/Access to Oracle SAS/Access to Teradata SAS/Access to Hadoop SAS Data Loader for Hadoop SAS Bridge to Viya SAS Viya 3.2/3.3 (VDMML) 4 Cores / min. 64-96 GB RAM SLES 11 SP3 SAS Cloud Analytics Services (CAS) - In-Memory Engine CAS Controller - CAS Monitor - CAS Client Services CAS Worker CAS Clients - SAS Studio Web Application - SAS Workspace CAS Client SAS Microservices Visual Data Mining and Machine Learning (VDMML) EP EP EP EP EP EP Datenquelle(n) SAP ERP System Oracle Teradata Shared Storage (optional) SAS Enterprise Miner Projekte SAS Data Marts SAS ABTs Hadoop Cluster MapR Hadoop Platform v5.2 6 Data Nodes SAS Embedded Process Authentifizierung Host Authentifizierung für SAS 9.4 AD/LDAP Authentifizierung für SAS Viya SAS Clients (9.4 & Viya) SAS 9.4 SAS Viya SAS Management Console SAS VDMML (Browser) SAS Enterprise Miner SAS VA (Browser) SAS Data Integration Studio SAS VS (Browser) SAS Enterprise Guide Topologie
SAS Viya Solution scenarios Viya Distributed Deploy CAS on MapR data nodes Recommend to increase to 8 cores at least (4x8 cores) Deploy Viya support services on SAS9 node SAS 9.4 M5 on the SAS9 side to leverage seamless integration tools
For the fiscal year ending in January 2017, Walmart s total Revenue was $485.9 Billion
Improve User Experience Develop Global Analytics Platform Increase utilization of investment in SAS Platform Goals & Returns SAS Grid SAS Viya Analytics Enhanced Analytical Innovation Hub Leveraging SAS Open Platform Make it easier for users to access environment Provide relevant tools to work being completed Allow world wide access to environment Simplicity in managing global platform Leverage central IT functions Unrestricted user access Additional organizations and users leverage central environment Known High Impact Analytics Projects Small Format - Store format optimization - Remodel Optimization - Store Clustering Supply Chain - Forecasting Truck demand for 2yr - DC Route Optimization Real Estate - Store layout optimization - Site selection for remodels Energy - Solar energy production - Energy demand by store by hour Logistics and Transportation - Ability to leverage existing projects - U.S. Distribution Center Network Coverage Model - Transportation Optimization Forecasting - Forecast Online Sales - Demand Planning and Forecasting for Stores Omni Channel Merchandising & Marketing - Customer Analytics - Market Basket Analysis - Trade Area Analytics - Marketing Attribution Recommendations Summary Better Best Keep same number of cores (64 Cores) like for like approach with fresh architecture and versions of software Combine Sunnyvale and Grid environment to provide for large scale forecasting License the same software across all grid nodes for ease of deployment and future upgrades Provide for high availability & lower cost by virtualizing & clustering SAS Metadata, Midtier and Grid Manager Add Small environment of Viya to the below which adds the following benefits; Allow for Open Source programming (API s for R, Python, Java, Lua, REST) that teams already have strong skills in Industry leading machine learning, deep learning and natural language processing Governance of and standardization on models and analytics SAS s Open Platform Viya is the next iteration of SAS s industry leading analytics platform that will enable significant future innovation Open Platform Including SAS Viya Controllers Workers
SAS platform strategy & SAS Viya SAS 9 one SAS platform