Big Data Architecture
Jordan Chung
2015
Agenda
• Big-Data data characteristic
• Big-Data whole process pictures
• Big-Data scenario (EKG example)
• How to build Big-Data server farm architecture
Jordan’s Profile
• Linkedin: https://www.linkedin.com/profile/view?id=5911091
• Facebook: https://www.facebook.com/profile.php?id=548778295
• FB club: https://www.facebook.com/SoftwareABC
Skype: jordan_chung, Line: jordan_chung, Wechat: chung_jordan
-- Business Development
-- Product development life cycle management
-- Product Management
-- Project Management
-- Solution Architect
-- Recruiting
-- Quality Assurance
-- Cloud solution
-- Golf platform http://www.pgggame.com
-- Pre-sales
-- eMarketing
-- eBook, Video, Audio, Karaok digital content solution.
Big-Data Data Characteristic
• Automatically generated by machine (or sensor)
• Typically an entirely new source of data
• Not designed to be friendly ( or not readable)
• May not have much values (價值)
Big-Data Source
Users
Systems
Applications (Desktop / Mobile)
Sensors
Large and growing files
Source Generator
Big-Data Process Picture (I)
Source Generation Web service Storage Reporting
traffic
Big-Data Process Picture (II)
Source Generation Storage ReportingWeb service
More traffic
Big-Data Process Picture (III)
Source Generation Storage ReportingWeb service
More traffic
Big-Data Process Picture (IV)
Source Generation Storage ReportingWeb service
More Data
需要以TB /PB來幾算
More traffic
Big-Data Process Picture (V)
Source Generation Storage ReportingWeb service
More Data
需要以TB /PB來幾算
More traffic
Big-Data Process Picture (VI)
Source Generation Storage Reporting
1. 分析的Power
2. 分析的時間
Web service
More DataMore traffic
需要以TB /PB來幾算
Big-Data Process Picture (VII)
Source Generation Storage Reporting
1. 分析的Computing Power
2. 分析的時間
需要以TB /PB來幾算 分析的方法
Web service
** 紅色的項目 是在Big data 中會出現且不太好解決的問題
More DataMore traffic Analytics
Model
One Tool?
One tool to rule Big
Data Service out
Big-Data Scenario - EKG
• User wear EKG (心電圖) device and post the message to
server every second
• Data format can be data text or number
• One hour will generate EKG data around 1 MB
• If having 10,000 user this EKG service,
then will roughly accumulate around 90 TB one year.
Which means 10,000 connection every seconds
情境示意圖
Sensor
RF/BLE http
(text) (text)
SQL
NoSQL DB
(text)
Server Farm
Client site
How to build Big-Data
Server farm
ELB
EC2 (Web)
Big-Data Server Architecture with
AWS Cloud
S3
EMR
EC2(Web)
MySQL (A/A/A)
定期
EC2(NoSQL)
ELB
EC2 (API/Web)
For APP
For Web
Sensor
RF/BLE
http
APP
(text)
http
Reporting
Take a bit time
Real Time Can be Batch
SQS/Kinese
EC2 ( Consumer Web)
EC2 (Cluster)
Mgr server
Auto -Scale
Auto -Scale
Build server farm architecture with Machines
FW (Active – Standby)
HA: 30 seconds
A S
…
VIP
internet
Web server
Switch
LBS
Switch
DB server
Cache Servers
VIP
Management server
Deployment server
IPS
VNX 5200 (Horizontal storage solution)
Switch
Switch
Storage
Can have single value
但機率很小很小
Switch
有機會合作, 將會了解Super Value

Big-Data Server Farm Architecture

  • 1.
  • 2.
    Agenda • Big-Data datacharacteristic • Big-Data whole process pictures • Big-Data scenario (EKG example) • How to build Big-Data server farm architecture
  • 3.
    Jordan’s Profile • Linkedin:https://www.linkedin.com/profile/view?id=5911091 • Facebook: https://www.facebook.com/profile.php?id=548778295 • FB club: https://www.facebook.com/SoftwareABC Skype: jordan_chung, Line: jordan_chung, Wechat: chung_jordan -- Business Development -- Product development life cycle management -- Product Management -- Project Management -- Solution Architect -- Recruiting -- Quality Assurance -- Cloud solution -- Golf platform http://www.pgggame.com -- Pre-sales -- eMarketing -- eBook, Video, Audio, Karaok digital content solution.
  • 4.
    Big-Data Data Characteristic •Automatically generated by machine (or sensor) • Typically an entirely new source of data • Not designed to be friendly ( or not readable) • May not have much values (價值)
  • 5.
    Big-Data Source Users Systems Applications (Desktop/ Mobile) Sensors Large and growing files Source Generator
  • 6.
    Big-Data Process Picture(I) Source Generation Web service Storage Reporting traffic
  • 7.
    Big-Data Process Picture(II) Source Generation Storage ReportingWeb service More traffic
  • 8.
    Big-Data Process Picture(III) Source Generation Storage ReportingWeb service More traffic
  • 9.
    Big-Data Process Picture(IV) Source Generation Storage ReportingWeb service More Data 需要以TB /PB來幾算 More traffic
  • 10.
    Big-Data Process Picture(V) Source Generation Storage ReportingWeb service More Data 需要以TB /PB來幾算 More traffic
  • 11.
    Big-Data Process Picture(VI) Source Generation Storage Reporting 1. 分析的Power 2. 分析的時間 Web service More DataMore traffic 需要以TB /PB來幾算
  • 12.
    Big-Data Process Picture(VII) Source Generation Storage Reporting 1. 分析的Computing Power 2. 分析的時間 需要以TB /PB來幾算 分析的方法 Web service ** 紅色的項目 是在Big data 中會出現且不太好解決的問題 More DataMore traffic Analytics Model
  • 13.
    One Tool? One toolto rule Big Data Service out
  • 14.
    Big-Data Scenario -EKG • User wear EKG (心電圖) device and post the message to server every second • Data format can be data text or number • One hour will generate EKG data around 1 MB • If having 10,000 user this EKG service, then will roughly accumulate around 90 TB one year. Which means 10,000 connection every seconds
  • 15.
  • 16.
    How to buildBig-Data Server farm
  • 17.
    ELB EC2 (Web) Big-Data ServerArchitecture with AWS Cloud S3 EMR EC2(Web) MySQL (A/A/A) 定期 EC2(NoSQL) ELB EC2 (API/Web) For APP For Web Sensor RF/BLE http APP (text) http Reporting Take a bit time Real Time Can be Batch SQS/Kinese EC2 ( Consumer Web) EC2 (Cluster) Mgr server Auto -Scale Auto -Scale
  • 18.
    Build server farmarchitecture with Machines FW (Active – Standby) HA: 30 seconds A S … VIP internet Web server Switch LBS Switch DB server Cache Servers VIP Management server Deployment server IPS VNX 5200 (Horizontal storage solution) Switch Switch Storage Can have single value 但機率很小很小 Switch
  • 19.