Copyright © 2012, SAS Institute Inc. All rights reserved. 巨量分析BigData,發掘企業新 契機,讓不可能變可能 SAS巨量分析事業處產品顧問 林輝倫( Allen Lin ) 2012/9/11 http://www.sas.com/offices/asiapacific/taiwan/high- performance-analytics/index.html
Copyright © 2012, SAS Institute Inc. All rights reserved.
巨量分析BigData,發掘企業新契機,讓不可能變可能 SAS巨量分析事業處產品顧問 林輝倫( Allen Lin ) 2012/9/11
http://www.sas.com/offices/asiapacific/taiwan/high-performance-analytics/index.html
2
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS是企業成長最忠實的夥伴
全球企業;在地支援
成立於 1976年
總部:美國 北卡羅來納州 卡麗
全球52個國家有400+個據點
台灣分公司成立於1989年(20年以上)
深耕台灣;國際接軌
全球10,000+位員工
台灣50+位
全球超過55,000個客戶
台灣300+客戶
Fortune 500前100大中有97家採用SAS
BusinessWeek 50 List中有41家採用SAS
2011營收:US $2.75 Billion 研發經費:24 % ($660 million)
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS 致力於進階商業分析已超過35年
商業智慧
進階商業分析
反應型決策模式
主動型決策模式
未來會發生什麼?
過去發生了什麼?
警示
Alert
多維度分析
OLAP
即時性報表
Ad-hoc Report
標準報表
Standard Report
最適化分析
Optimization
預測模型
Predictive Modeling
趨勢分析
Forecasting
統計分析
Statistical Analysis
傳統的商業智慧 Business Intelligence
進階的商業分析 Advanced Analytics
文字分析
Text Analytics
No. 1 World Leader In Business Analytics
SAS leads Advanced Analytics Market by Wide Margin (IDC, June 2011)
Copyright © 2012, SAS Institute Inc. All rights reserved.
資料量 VOLUME
資料種類 VARIETY
資料產生的速度 VELOCITY
資料蘊含的價值 VALUE
現在 未來
資料量大小
趨勢 巨量資料的挑戰
Copyright © 2012, SAS Institute Inc. All rights reserved.
當進階商業分析遇到巨量資料…
巨量資料 Big Data
SAS高效能分析 SAS High Performance
Analytics
進階商業分析 Advanced Analytics
能夠充分運用平行處理資源進行高效能進階分析的廠商
讓分析不需受限於資料種類、樣本大小、變數量、及歷史資料的長短
讓充分的情境模擬分析可於短時間內完成
讓分析人員得以解決更多更複雜的業務問題
讓即時分析、預測、與模擬的結果融入於決策過程中
支援Hadoop, Greenplum與其他資料庫廠商等
+ =
http://www.sas.com/offices/asiapacific/taiwan/high-performance-analytics/index.html
Copyright © 2012, SAS Institute Inc. All rights reserved. 6
Allows in-memory modeling against entire data
sets on a specialized Greenplum appliance
Increases business value of models by
improving model selection
Dramatically accelerates the analytic lifecycle
process for select models
DATA
EXPLORATION
MODEL
DEVELOPMENT
MODEL
DEPLOYMENT
•
PREDICT
IVE
MODELI
NG
•
VARIABL
E
SELECTI
ON
ANALYTIC
LIFECYCLE
Scoring Accelerators – Translate Enterprise Miner models into
database-specific functions to execute
in database.
SAS Visual Analytics– Visualize ALL your data – billions of records
-- to understand the variables that your
datasets contain. Uncover relationships and
correlations in your data.
SAS HPA (the
product) – Develop predictive models in
memory alongside distributed
relational databases. Data is not
physically moved, SAS
processing is brought to the
database appliance. Models can
be built on ALL of the data.
SAS High-Performance Analytics
for Greenplum Offering
7
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics Key Components
8
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS Server
libname GP joe;
proc means data=joe.flights;
var delay;
run;
Approach: Use Access Engines
Appliance
SELECT delay
FROM flights SELECT delay
FROM flights
Master Workers
Access
Engine
SELECT delay
FROM flights Big Data
9
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS Server
libname GP joe;
proc reg data=joe.flights;
model delay=length day;
run;
Inside the Database – SQL and UDF’s
Appliance
UDF that accumulates X’X Aggregator UDF
Master Workers Access
Engine
TKTS
X’X X’X
Copyright © 2012, SAS Institute Inc. All rights reserved.
Data Extracts
Analytical Data Preparation
Modeling Scoring
Scoring
ADS
Data Extracts
Scoring Data Preparation
Modeling
ADS
Model Development Model Deployment
Database /Data Warehouse
Database /Data Warehouse
SAS
M
Model
SAS
Traditional Architecture
Model Development Model Deployment
SAS
M
Model
SAS
Model Translation
In-database Scoring
Analytical Data Preparation
Scoring Data Preparation
Scoring
ADS
Modeling
ADS
Modeling ADS
Modeling
SAS Model
Greenplum
In-Database Architecture
SAS® IN-DATABASE
SAS高效能分析解決方案 – 架構說明
11
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics Key Components
12
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS Server
libname GP joe;
proc hpreg data=joe.flights;
class airline day(split);
model delay=airline day
duration …;
selection method=lasso;
run;
Alongside-the-Database
Appliance
Master Workers
tkgrid
Access
Engine
General Captains
TK TK TK TK TK
MPI
SQL SQL SQL SQL SQL
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics Appliance
SAS高效能分析解決方案運作方式 SAS® IN-MEMORY ANALYTICS
Greenplum Node
14
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics: Architecture
Submit a program from a SAS client session (eg. HPLOGISTIC)
proc hplogistic data=GPlib.sgf_binary;
class A B C;
model y = a b c x1 x2 x3;
performance details host="green1";
run;
15
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics: Architecture
Worker Node 1
Worker Node 2
Worker Node N
Master
Request is sent to the appliance
and received by the Master Node
16
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics: Architecture
Worker Node 1
Worker Node 2
Worker Node N
Master
Analytical Computation and data request sent to the worker nodes
17
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics: Architecture
Worker Node 1
Worker Node 2
Worker Node N
Master
Data request sent to the database, data slice moved into memory
18
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics: Architecture
Worker Node 1
Worker Node 2
Worker Node N
Master
Analytic Processing with internode communication
19
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics: Architecture
Worker Node 1
Worker Node 2
Worker Node N
Master
Worker node results returned to the Master Node, finalize
computation
20
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics: Architecture
Worker Node 1
Worker Node 2
Worker Node N
Root Node
Result returned to the client
Copyright © 2012, SAS Institute Inc. All rights reserved.
高效能資料採礦與SAS EM整合,提供多個高效能運算分析節點
採礦處理流程可進
行自動化處理
與模型比較整合
SAS® IN-MEMORY ANALYTICS
SAS高效能分析解決方案 – SAS HAP與EM整合
22
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS® HPA Server
MODELING RESULTS
Variable Selection Classification and
Prediction Text Mining
Copyright © 2012, SAS Institute Inc. All rights reserved.
Central Entry Point Integration Role-based Views
MOBILE DATA PREPARATION EXPLORER DESIGNER
• Native iOS application
that delivers interactive
reports created in the
designer
• Monitor SAS® LASR™
Analytic server
• Load and join data
• Create calculated
columns
• Perform ad-hoc analysis
and data discovery • Create dashboard style
reports for web
or mobile
SAS® LASR™ ANALYTIC SERVER
(VISUAL ANALYTICS) SAS高效能分析解決方案 SAS® IN-MEMORY ANALYTICS
24
Copyright © 2012, SAS Institute Inc. All rights reserved.
SAS High-Performance Analytics Key Components
Copyright © 2012, SAS Institute Inc. All rights reserved. 25
Allows in-memory modeling against entire data
sets on a specialized Greenplum appliance
Increases business value of models by
improving model selection
Dramatically accelerates the analytic lifecycle
process for select models
DATA
EXPLORATION
MODEL
DEVELOPMENT
MODEL
DEPLOYMENT
•
PREDICT
IVE
MODELI
NG
•
VARIABL
E
SELECTI
ON
ANALYTIC
LIFECYCLE
Scoring Accelerators – Translate Enterprise Miner models into
database-specific functions to execute
in database.
SAS Visual Analytics– Visualize ALL your data – billions of records
-- to understand the variables that your
datasets contain. Uncover relationships and
correlations in your data.
SAS HPA (the
product) – Develop predictive models in
memory alongside distributed
relational databases. Data is not
physically moved, SAS
processing is brought to the
database appliance. Models can
be built on ALL of the data.
SAS High-Performance Analytics
for Greenplum Offering
Copyright © 2012, SAS Institute Inc. All rights reserved.
最高準確度之預測
最佳之執行效能
最大廣度及深度之分析
無與倫比之企業績
效
WHY SAS 巨量分析
• SAS是充分運用平行處理資源進行高效能進階分析的廠商
• 讓分析人員得以解決更多更複雜的業務問題
• 讓即時分析、預測、與模擬的結果融入於決策過程中
• 讓原本很多不可能的服務與應用變可能
http://www.sas.com/offices/asiapacific/taiwan/high-performance-analytics/index.html