SlideShare a Scribd company logo
Japanese  Startup  Use-‐‑‒Cases  and
Tech  Deep  Dive
Amazon  Data  Services  Japan
Eiji  Shinohara
March  29,  2015
  Name:
  ・Eiji  Shinohara  /  篠原  英治  /  @shinodogg
  Role:
・AWS  Solutions  Architect  for  Startups
・Amazon  CloudSearch  Subject  Matter  Expert
Who  am  I?
Talking  to  Startup  CTOs/Engineers  on  daily  basis
 AWS  Startup  CTO  Night  with  Amazon  CTO
•  We  had  Amazon  CTO  Werner  Vogels
TechCrunch  Tokyo  CTO  Night  powered  by  AWS
•  Startups  pitch  contest  for  “CTO  of  the  year”
 IVS  CTO  Night  &  Day  powered  by  AWS
•  3  days  Over  100  CTOs  gathering
•  w/  Infinity  Ventures  Summit
CTONight  Series  in  2014
Japanese Startup Use-Cases and Tech Deep Dive
 AWS  Startup  CTO  Night  with  Amazon  CTO
TechCrunch  Tokyo  CTO  Night  powered  by  AWS
Contest  for  JP  Startup  CTO  of  the  year!
TechCrunch  Tokyo  CTO  Night  powered  by  AWS
  Pitch  Presenters  
  (Startup  CTOs)
  Judges  
  (Popular  Company  CTOs)
GREE
Cookpad
BizReach
Hatena
CyberAgent
Amazon
TechCrunch  Tokyo  CTO  Night  powered  by  AWS
  IVS  CTO  Night  &  Day  powered  by  AWS
  IVS  CTO  Night  &  Day  powered  by  AWS
  IVS  CTO  Night  &  Day  powered  by  AWS
【Survey  Result】
100%  participant  CTOs  said...
“  WANT  TO  JOIN  THIS  EVENT  AGAIN!!”
AWS  is  empowering  Startups!
Letʼ’s  Meetup  at  CTO  Night  (´́▽`̀)ノ
AWS  Pop-‐‑‒up  Loft  in  San  Francisco
AWS  Pop-‐‑‒up  Loft  in  San  Francisco
I  will  give  a  talk  about  “Japanese  Startups”  at  
SF  Pop-‐‑‒up  Loft  (´́▽`̀)ノ
We  never  disclose  AWS  customersʼ’  info  without  permission.
We  got  agreements  for  all  use-‐‑‒cases  in  this  slide.
   Surefire  Architecture
•  ELB  +  EC2  +  RDS  on  MultiAZ
•  Super  Fast  Distribution  with  S3  +  CloudFront
   Beat  “Bursty  Traffic”
•  How  to  beat  burst
•  Cost  range
   Startup  Pragmatic  Use-‐‑‒Cases
   Amazon  CloudSearch  Startup  Use-‐‑‒Cases
Agenda
   Surefire  Architecture
•  ELB  +  EC2  +  RDS  on  MultiAZ
•  Super  Fast  Distribution  with  S3  +  CloudFront
   Beat  “Bursty  Traffic”
•  How  to  beat  burst
•  Cost  range
   Startup  Pragmatic  Use-‐‑‒Cases
   Amazon  CloudSearch  Startup  Use-‐‑‒Cases
Agenda
Surefire  Architecture
ELB  +  EC2  +  RDS  on  Multi  AZ
EC2
RDS
ELB
Availability  Zone
Web
Availability  Zone
RDS
EC2
WebWeb
BASE  https://thebase.in/
Easy  to  create  online  shopping  site
Variety  of  templates  and  Payment  System
And  itʼ’s  all  free
BASE  https://thebase.in/
EC2 EC2
RDS
(Active)
DB
ELB
AZ① AZ②
RDS
(Standby)
ElastiCache S3
CloudFront
BASE  https://thebase.in/
EC2 EC2
RDS
(Active)
DB
ELB
AZ① AZ②
RDS
(Standby)
ElastiCache S3
CloudFront
BASE  https://thebase.in/
EC2 EC2
RDS
(Active)
DB
ELB
AZ① AZ②
RDS
(Standby)
ElastiCache S3
CloudFront
EC2  on  MultiAZ
Just  increase  EC2  instances  if  needed  ☺
EC2
RDS
ELB
Availability  Zone
Web
Availability  Zone
RDS
EC2
WebWeb
EC2
RDS
ELB
Availability  Zone
Web
Availability  Zone
RDS
Web
EC2
Web
EC2
Web
EC2
Web
Create  EC2  instances  from  AMI
Web
  AMI(Amazon  Machine  Image)
•  Easy  to  Create  Image
•  Able  to  Create  EC2  instances  from  the  Image
AMI
Web
Web
Create  AMI Create  EC2  instances
Dynamic  Scaling  with  Auto  Scaling
  Auto  Scaling
•  EC2  scale  out/in  to  adjust  load  and  traffic
•  e.g.  if  70%  CPU  usage  for  5min,  then  add  2  instances
Auto  Scaling  Group
EC2
Web
EC2
Web
ELB
Auto  Scaling
CloudWatch
monitoring
Dynamic  Scaling  with  Auto  Scaling
  Auto  Scaling
•  EC2  scale  out/in  to  adjust  load  and  traffic
•  e.g.  if  70%  CPU  usage  for  5min,  then  add  2  instances
Auto  Scaling  Group
EC2
Web
EC2
Web
ELB
Auto  Scaling
CloudWatch
monitoring
Web Web
EC2 EC2
create  EC2  Instances
add  to  LoadBalancer
BASE  https://thebase.in/
EC2 EC2
DB
ELB
AZ① AZ②
Sync
ElastiCache RDS
(Active)
RDS
(Standby)
S3
CloudFront
RDS  on  MultiAZ
RDS  Multi-‐‑‒AZ  deployments  provide  
enhanced  availability  and  durability
Scale-‐‑‒Up  if  needed
EC2
RDS
ELB
Availability  Zone
Web
Availability  Zone
RDS
EC2
WebWeb
EC2
RDS
ELB
Availability  Zone
Web
Availability  Zone
RDS
EC2
WebWeb
SnapShot
Auto/Manual
Multi-‐‑‒AZ  makes  you  feel  you  relieved  ☺
Availability  Zone Availability  Zone
RDS RDS
“Sync”  Replication
Automatic  Failover
Log  Flush
5min  interval
Replication  across  data  centers
Restore  from  Snapshots  and  Logs
Easy  to  create  Read  Replicas
Availability  Zone Availability  Zone
RDS RDS
Async  Replication
To  handle  massive  query  requests
   Static  Contents  Distribution
Make  it  Super  fast  with  S3+CloudFront
Often  happen  to  see  like  this
  Copy  from  File  server  to  Web  servers  with  rsync/scp
•  Capacity  Limit?Backup?
•  High  Latency  from  overseas..
Web Web
Shared
File
Server
Amazon  S3
 Static  Contents  with  S3
•  Image,  Movie,  HTML,  JavaScript  ,  CSS  ,,
http://www.slideshare.net/AmazonWebServices/why-‐‑‒scale-‐‑‒matters-‐‑‒v4-‐‑‒rodos
Over  3  Trillion  Objects
in  2014
Amazon  S3
 Static  Contents  Distribution
Upload  HTML/Image  to  S3
Make  Public
Amazon  S3
 Static  Contents  Distribution
Upload  HTML/Image  to  S3
Make  PublicPopular  reasonable  solution  for  Teaser  
site  and  Landing  page
Amazon  S3  –  e.g.)  WordPress  Site
 Using  WordPress  Plugin  like  “StaticPress”  to  make  
contents  static  and  upload  to  S3
 Start  EC2  instance  only  when  you  create  contents
•  Utilize  per  Hour  billing  model
http://en.staticpress.net/
Amazon  S3  –  e.g.  WordPress  Site
 S3  +  CloudFront  can  handle  massive  request
WordPress
CloudFrontS3EC2
Amazon  CloudFront
  Pay-‐‑‒as-‐‑‒you-‐‑‒go  Contents  Distribution  Network
  Route  the  request  to  the  nearest  Edge  Location
  Over  50  Edge  Locations  around  the  world
レスポンス向上 負荷軽減
Cached  
Contents 4.Send  contents  to  
the  Edge  Location
1.Request  to  
the  nearest  
Edge  Location
3.If  no  cache  data,  then  
access  to  Origin
2.Deliver  contents  from  Cache
5.Cache  contents  on  the  Edge  Location
Amazon
CloudFront
(Edge  Location)
OriginClient
iQON  http://www.iqon.jp/
iQON  http://www.iqon.jp/
Web/AppWeb/App
S3
CloudFront
Solr Memcached Redis MySQL MongoDB
Crawler Zabbix
VarnishELB
PC/Mobile API
GW
iQON  http://www.iqon.jp/
Web/AppWeb/App
S3
CloudFront
Solr Memcached Redis MySQL MongoDB
Crawler Zabbix
VarnishELB
PC/Mobile API
GW
Unlimited  Capacity
Donʼ’t  need  to  be  worried  about  backup
Put  Images  to  S3
Web/AppWeb/App
S3
CloudFront
Solr Memcached Redis MySQL MongoDB
Crawler Zabbix
VarnishELB
PC/Mobile⽤用 API⽤用
GW
Super  Fast  and  Reduce  EC2  load
All  the  static  contents  
are  distributed  with  
CloudFront  +  S3
iQON  http://www.iqon.jp/
 Surefire  Architecture
•  EC2  on  Multi-‐‑‒AZ  with  Elastic  Loadbalancer
•  RDS  on  Multi-‐‑‒AZ  enhances  Durability  and  Availability
 Static  Contents  Distribution
•  S3:  Unlimited  Capacity  /  Eleven-‐‑‒Nine  Durability  
•  CloudFront:  Super  Fast  Global  Distribution
 Cool  Japanese  Startups  are  actually  doing!
   Surefire  Architecture
•  ELB  +  EC2  +  RDS  on  MultiAZ
•  Super  Fast  Distribution  with  S3  +  CloudFront
   Beat  “Bursty  Traffic”
•  How  to  beat  burst
•  Cost  range
   Starup  Pragmatic  Use-‐‑‒Cases
   Amazon  CloudSearch  Startup  Use-‐‑‒Cases
Agenda
Access  increasing  through  Internet
 Twitter  buzz,  Pick  up  on  popular  blog,  Tech  
Crunch,  News  site,,,
Traffic  increases  gradually
But  TV  broadcasting…
 NHK  Kōhaku_̲Uta_̲Gassen
•  Most  popular  JP  TV  Program  on  the  last  day  of  the  year
•  Audience  Rating  is  about  50%
•  Second  Screen  -‐‑‒  Smartphone  App
•  Handle  hundreds  thousands  of  simultaneous  requests
[Classmethod  Use-‐‑‒Case]  http://classmethod.jp/cases/nhk/nhk-‐‑‒kouhaku-‐‑‒64/
But  TV  broadcasting…
 NHK  Kōhaku_̲Uta_̲Gassen
•  Most  popular  JP  TV  Program  on  the  last  day  of  the  year
•  Audience    Rating  is  about  50%
•  Second  Screen(Smartphone  App)
•  Handle  hundreds  thousand  of  simultaneous  requests
[Classmethod  Use-‐‑‒Case]  http://classmethod.jp/cases/nhk/nhk-‐‑‒kouhaku-‐‑‒64/
Burst  at  sudden!
 What  can  we  do  for  that  burst?
The  ways  to  take  the  advantage  of  chances!
Basic  Idea..
↑Scale  
UP
RDS
ELB
AZ① AZ②
EC2 EC2EC2
Web
App
Web
App
Web
App
increase  EC2  instances
EC2 EC2
Web
App
Web
App…
EC2 EC2EC2
Web
App
Web
App
Web
App
EC2 EC2
Web
App
Web
App…
RDS
 Question
How  much  is  it  for  Beat  “Bursty  Traffic”?
 1:$10
 2:$100
 3:$1000
・return  EC2  and  RDS  back  after  two  hours
・m3.medium:  1  core  CPU  /  3.7GB  Memory
・m3.xlarge:  4  core  CPU  /  15GB  Memory
For  two  hours
increase  20  EC2(medium)  instances
two  stage  scale  up  RDS  (medium→xlarge)
How  much  is  it  for  Beat  “Bursty  Traffic”?
 1:approx.  $10
 2:$100
 3:$1000
・return  EC2  and  RDS  back  after  two  hours
・m3.medium:  1  core  CPU  /  3.7GB  Memory
・m3.xlarge:  4  core  CPU  /  15GB  Memory
For  two  hours
increase  20  EC2(medium)  instances
two  stage  scale  up  RDS  (medium→xlarge)
○
×
×
Even  20  instances  are  in  this  price  
range.  You  can  run  a  lot  to  be  relieved!
*  Donʼ’t  forget  to  return  it  back!
   Is  ELB(Load  Balancer)  OK…?
Elastic  Load  Balancing
 Handling  capacity  to  meet  the  demands  of  traffic
ELB  can  handle  generally
But  need  a  little  time  to  scale
But  TV  broadcasting…
RDS
ELB
AZ① AZ②
EC2 EC2EC2
Web
App
Web
App
Web
App
EC2 EC2
Web
App
Web
App…
EC2 EC2EC2
Web
App
Web
App
Web
App
EC2 EC2
Web
App
Web
App…
RDS
The  situation  that  ELB  scaling  doesnʼ’t  
make  it  can  be  happen.
 AWS  Support  can  scale  ELB  in  Advance
24/365  Support  (You  can  choose  JAPANESE!)
Basic Developer Business Enterprise
Forum ✔ ✔ ✔ ✔
Access  to  tech  
support
Support  for  Health  
Checks
Email  (local  business  
hours)
Phone,  chat,  email,  
live  screen  sharing  
(24/7)
Phone,  chat,  email,  
live  screen  sharing,  
TAM  (24/7)
Response  Time -‐‑‒ <12  hours <1  hour <15  minutes
Users -‐‑‒ 1
Unlimited
(IAM  supported)
Unlimited
(IAM  supported)
24/365 -‐‑‒ -‐‑‒ ✔ ✔
Trusted  Advisor 4  checks 4checks 41checks 41checks
TAM -‐‑‒ -‐‑‒ -‐‑‒ ✔
Cost  (Monthly) $0 $49
monthly  AWS  usage
  $0~∼$10K:      10%
  $10K~∼$80K:  7%
  $80K~∼$250K:  5%
  $250K~∼:  3%
    (Min  $100)
10%  of  AWS  monthly  
usage
(Min  $15000)
AWS  Support(Business)  can  handle  ELB  Pre-‐‑‒Warming
RDS
ELB
AZ① AZ②
EC2 EC2EC2
Web
App
Web
App
Web
App
EC2 EC2
Web
App
Web
App…
EC2 EC2EC2
Web
App
Web
App
Web
App
EC2 EC2
Web
App
Web
App…
RDS
JMTY  http://jmty.jp/
   Similar  to  craigslist.  Picked  up  frequently  on  TV
“JMTY  Engineers  and  AWS  team  are  handling  massive  
traffic!  Great!”
CEO  Kato-‐‑‒san
Replica  Sets
Images
CloudFront
MongoDB
(Secondary)
MongoDB
(Primary)
MongoDB
(Hidden)
Web/App
Mail(postfix)
Multi AZ
RDS
Redis(Resque)
Batch
Memcached
Replica  Sets
Images
CloudFront
MongoDB
(Secondary)
MongoDB
(Primary)
MongoDB
(Hidden)
Web/App
Mail(postfix)
Multi AZ
RDS
Redis(Resque)
Batch
Memcached
Replica  Sets
Images
CloudFront
MongoDB
(Secondary)
MongoDB
(Primary)
MongoDB
(Hidden)
Web/App
Mail(postfix)
Multi AZ
RDS
Redis(Resque)
Batch
Memcached
Scale  Up
Scale  Out
Pre  
Warming
   Life-‐‑‒long  learning  platform.  Getting  popular!
schoo  http://schoo.jp/
Itʼ’s  me  ☺
 heard  from  schoo  engineer  Iwata-‐‑‒san  
about  their  activity  for  TV  broadcasting
Schoo  engineer  Iwata-‐‑‒san
schoo  http://schoo.jp/
▼改定前価格(2015年年3⽉月31⽇日中ご登録分まで)525円/⽉月額
▼改定後価格(2015年年4⽉月1⽇日以降降ご登録分より)980円/⽉月額
Solutions  For  TV  broadcasting  at  schoo
 Create  tens  of  EC2  instances  from  AMI
• It  took  about  1  hour
 Just  before  and  after  the  TV  program
• Low  Cost  Affordable!
 Set  aside  extra  EC2  instances
• Relieved  and  Didnʼ’t  have  any  problem!!
Solutions  For  TV  broadcasting  at  schoo
 Create  AWS  Support  cases
•  EC2  Instances  Limit  Increase
•  Instance  numbers  is  limited  to  avoid  unnecessary  problem  by  default
•  ELB  Pre-‐‑‒Warming
Solutions  For  TV  broadcasting  at  schoo
 Iwata-‐‑‒san  said
•  Thanks  to  Japanese  Language  Support!!
•  Didnʼ’t  have  any  problems.  It  was  quite  smooth.
   Surefire  Architecture
•  ELB  +  EC2  +  RDS  on  MultiAZ
•  Super  Fast  Distribution  with  S3  +  CloudFront
   Beat  “Bursty  Traffic”
•  How  to  beat  burst
•  Cost  range
   Startup  Pragmatic  Use-‐‑‒Cases
   Amazon  CloudSearch  Startup  Use-‐‑‒Cases
Agenda
For  Couples.  Photo  Album  /  Chat  /  Date  schedule
EC2
(API)
DB
ELB
(MultiAZ)
RDSDynamoDB
ELB
EC2
(Websocket)
EC2
(Worker)
EC2
(Redis)
S3
CloudFront
SES
Pairy  http://pairy.com/
Pairy  http://pairy.com/
EC2
(API)
DB
ELB
MultiAZ
RDSDynamoDB
ELB
EC2
(Websocket)
EC2
(Worker)
EC2
(Redis)
S3
CloudFront
SES
Basic  and  
Transaction  Data
Right  Database  in  the  Right  Place
Chat  and  News  Feed
Handling  Massive  Write
Temporary  Data
Pairy  http://pairy.com/ C2C  Market  Place  like  Etsy
STORES.jp  https://stores.jp/
S3
MongoDB
ELB
DB Payment
Image
Redis
Web
STORES.jp  https://stores.jp/
S3
MongoDB
ELB
DB Payment
Image
Redis
Web
Put  logs  to  S3
Using  Log  collector  “Fluentd”
There  is  an  S3  plugin(fluent-‐‑‒plugin-‐‑‒s3)
Easy  to  gather  log  data
   Surefire  Architecture
•  ELB  +  EC2  +  RDS  on  MultiAZ
•  Super  Fast  Distribution  with  S3  +  CloudFront
   Beat  “Bursty  Traffic”
•  How  to  beat  burst
•  Cost  range
   Startup  Pragmatic  Use-‐‑‒Cases
   Amazon  CloudSearch  Startup  Use-‐‑‒Cases
Agenda
   Japanese  Search  function??
 Amazon  CloudSearch
•  Full  Managed  Cloud-‐‑‒Based  Search  Service
•  Pretty  easy  to  introduce
•  34  languages  support
•  Sophisticated  Functions
• Highlight
• Suggest(AutoComplete)
• Geo  Search
Amazon  CloudSearch
  Suggestions
/suggest?q=ir&suggester=title_̲sug
"suggest":  {"query":  "iro",  "found":  5,
  "suggestions":  [
    {“suggestion”:  “Iron  Man”,…"id":  "tt0371746"},
    {"suggestion":  "Iron  Man  2”,…"id”:"tt1228705"},
        ...
•  Reading  Search
•  Japanese  language  has
        Kanji/Hiragana/Katakana  ,,
e.g.  Nanboku  Line  Subway  Station  Search
Using  Amazon  CloudSearch
  Create  Domain
Using  Amazon  CloudSearch
  Data(Station  name&Line  name)
Station  Code
Station  Name
A  lot  of  stations  are  served  by  multiple  line  in  Tokyo
Using  Amazon  CloudSearch
  Schema  design(Field  definition)
Using  Amazon  CloudSearch
  Search  with  “JR⼭山⼿手線”  (most  popular  circle  line  in  Tokyo)
Using  Amazon  CloudSearch
  Search  with  “⿇麻布”  or  “⼀一丁⽬目”
Knowledge  is  needed  to  manage  Search  service..
Search  Engine
  Find  documents  with  keyword  from  large  amount  of  data
•  Incrementally  like  grep?  It  takes  too  long
•  Need  to  build  index  in  advance(Inverted  Index)
•  TF-‐‑‒IDF  scoring
•  Multiple  Query  Parser  Support
CloudSearch  -‐‑‒  Japanese  Text  Processing
  形態素解析(Morphological  Analysis)
•  彼(名詞-‐‑‒代名詞)/は(助詞-‐‑‒係助詞)/エンジニア(名詞-‐‑‒⼀一般)/だ(助動詞)
  Stemming
•  飲んだ  →  飲ん(動詞-‐‑‒⾃自⽴立立,  baseForm:飲む)/だ(助動詞)  →  飲む
  Synonym  Addition
•  「ベニス」「ベネチア」「ヴェネチア」
•  Alias
•  search  with  pupil  =>  student  is  hit
•  search  with  student  =>  pupil  in  NOT  hit
•  Group
•  1st,  first,  one  =>  you  can  search  with  all  keywords  in  the  group
  Stop  Word  Removing
Automatic  Scaling
  By  Document  size/number  and  Search  Request
Auto  Partitioning
with  EMR
Auto  Scaling
CloudSearch  –  support  variety  of  filed  types
  Field  Types
Double
Date
Signed  Integer Text
Literal
Amazon  CloudSearch  Meetup  in  Tokyo
A9 schoo nanapi ChatWork Cookpad ADSJ A9
CloudSearch  use-‐‑‒case:  ChatWork
ChatWork:  Business  Communication  Tool
•  Over  40  thousand  companies  are  using
•  About  a  half  million  users
  comment  from  ChatWork  CTO  Yamamoto-‐‑‒san
•  “To  handle  about  5  hundred  million  documents,  we  
introduced  CloudSearch.  Thanks  to  AWS  and  A9  
team,  it  took  only  a  few  month.”
CloudSearch  use-‐‑‒case:  ChatWork  Tanaka-‐‑‒san  slide
https://speakerdeck.com/tanakayuki/kai-‐‑‒fa-‐‑‒zhe-‐‑‒karamitacloudsearch  
Almost  maintenance  free
Positive  feedback  from  end-‐‑‒users  about  Low  latency
CloudSearch  use-‐‑‒case:  Engineer  Cross2015(29th  Jan)
•  ChatWork  is  making  CloudSearch  noise  in  Japan
CloudSearch  use-‐‑‒case:  nanapi
CloudSearch  use-‐‑‒case:  nanapi
nanapi  is  a  Life  Recipe  portal
•  About  20  million  per  user  per  Month
•  Over  0.1  million  recipes
  Getting  popular  these  days
CloudSearch  use-‐‑‒case:  nanapi  Kagaya-‐‑‒san  slide
https://speakerdeck.com/violetyk/cloudsearch-‐‑‒nanapi-‐‑‒use-‐‑‒case  
•  Default  setting  works  a  lot
•  Easy  to  have  Japanese  search  function
•  Fully  managed  by  AWS  is  huge  plus
CloudSearch  use-‐‑‒case:  schoo
CloudSearch  use-‐‑‒case:  schoo
schoo
•  Learning  platform  for  hundreds  of  thousands  students
•  Over  1000  students  join  in  popular  class!
  You  can  learn  “AWS”  on  schoo  ☺
CloudSearch  use-‐‑‒case:  schoo  Ito-‐‑‒san  slide
http://www.slideshare.net/hiromitsuito71/20141017-‐‑‒cloud-‐‑‒searchschoo  
It  took  only  1  WEEK  to  introduce.  Itʼ’s  so  easy  and  nice.
Of  course  you  need  to  escape  XSS  stuff
 Japanese  Language  is  not  so  easy
•  Yahoo!  Japan  Search  Engineer  Osuka-‐‑‒san  slide
Hasegawa-‐‑‒san?  
Tanigawa-‐‑‒san??
Need  to  analyze  and  only  the  
user  can  know  the  answer
CloudSearch  Updates
   Indexing  Bigrams
•  Bi-‐‑‒gramでのインデクシング
   Customizing  Japanese  Tokenization
•  形態素解析辞書のカスタマイズ
Tokenization  Dictionary
CloudSearch  use-‐‑‒case:  Lancers
Crowd-‐‑‒sourcing  service
CloudSearch  use-‐‑‒case:  Gochi-‐‑‒Kuru
Bento-‐‑‒Box  delivery  service
http://aws.amazon.com/jp/solutions/case-‐‑‒studies/gunosy/
Thanks  to  AWS:  scalable  and  programmable,  Now  we  can  manage  
4  million  app  downloads  from  only  40  thousand  users  very  smoothly.
Gunosy  is  a  popular  news  curation  app.  
CTO  Ishiba-‐‑‒san  said,
   Utilizing  AWS  full  managed  services
EC2
DB
ELB
MultiAZ
Gunosy  Architecture
MongoDB
Redis Redshift EMR
S3
CloudFront
RDS
Fluentd
Datastore  &  Analyze
Glacier
Web/App
Using  AWS  Managed  Services
 Surefire  Architecture
•  Stable  and  Scalable
•  Suitable  for  booming  Startup  IT  infra
 Beat  “Bursty  Traffic”
•  Take  advantage  of  the  chance
•  No  need  to  take  a  lot  of  time  and  money
 Startup  Pragmatic  use-‐‑‒cases
•  Utilizing  AWS  full  managed  services!
 AWS  has  bunch  of  use-‐‑‒cases  and  knowledge!
Wrap  up
   FYI
AWS  Black  Belt  Tech  Webinar
  Every  Wed  6PM  –  7PM(JST)  Online  Seminar  in  Japanese
http://aws.typepad.com/sajp/  
w/  Adobe  Connect
AWS  Black  Belt  Tech  Webinar
  Deep  dive  product-‐‑‒cut  seminar  by  Solution  Architect
http://aws.typepad.com/sajp/  
Amazon Simple Queue Service (SQS)
  AWS Elastic Beanstalk: Worker Tier
•  SQS + Auto Scalingでスケーラブルなバッチ処理基盤
Sqsd
(deamon)
Elastic Beanstalk
Application
http://localhost:80/xxx
EC2 Instance
Auto Scaling group
CloudWatch
Auto Scaling
AWS  Black  Belt  Tech  Webinar
    Check  #awsblackbelt  hashtag  on  Twitter!
AWS  Tokyo  Region  4  Years  Anniversary
Thank  you!!

More Related Content

PDF
CTO Night & Days 2015 Winter - AWS Mobile Development
PDF
"How to optimize the architecture of your platform" by Julien Simon
PDF
Building serverless apps with Node.js
PDF
Deep Learning on AWS (November 2016)
PDF
Deep Dive on Amazon Elastic Container Service (ECS) | AWS Summit Tel Aviv 2019
PPTX
AWS 101 - Amazon Web Services
PDF
Building a Serverless Pipeline
PDF
Picking the right AWS backend for your Java application (May 2017)
CTO Night & Days 2015 Winter - AWS Mobile Development
"How to optimize the architecture of your platform" by Julien Simon
Building serverless apps with Node.js
Deep Learning on AWS (November 2016)
Deep Dive on Amazon Elastic Container Service (ECS) | AWS Summit Tel Aviv 2019
AWS 101 - Amazon Web Services
Building a Serverless Pipeline
Picking the right AWS backend for your Java application (May 2017)

Viewers also liked (14)

PPTX
Procuretopia
PDF
Blockchain bootcamp for boards
PDF
Block Chain meets Big Data
PPTX
Use Cases
PPTX
Blockchain for First Timers with Use Cases
PDF
Blockchain Use Cases: Think of a "Public" Pub/Sub Queue
PPTX
Blockchain: Real World Use Cases
PDF
Deep Learning Cases: Text and Image Processing
PPTX
Practical Applications of Block Chain Technologies
ODP
CBGTBT - Part 2 - Blockchains 101
PPTX
Blockchain in IoT and Other Considerations by Dinis Guarda
PDF
Deep Learning Use Cases - Data Science Pop-up Seattle
PDF
Deep Learning - The Past, Present and Future of Artificial Intelligence
Procuretopia
Blockchain bootcamp for boards
Block Chain meets Big Data
Use Cases
Blockchain for First Timers with Use Cases
Blockchain Use Cases: Think of a "Public" Pub/Sub Queue
Blockchain: Real World Use Cases
Deep Learning Cases: Text and Image Processing
Practical Applications of Block Chain Technologies
CBGTBT - Part 2 - Blockchains 101
Blockchain in IoT and Other Considerations by Dinis Guarda
Deep Learning Use Cases - Data Science Pop-up Seattle
Deep Learning - The Past, Present and Future of Artificial Intelligence
Ad

Similar to Japanese Startup Use-Cases and Tech Deep Dive (14)

PDF
Scale, baby, scale!
PDF
Escalando hasta sus primeros 10 millones de usuarios
PDF
What is Amazon Web Services & How to Start to deploy your apps ?
PPTX
2011 Introduction to Cloud Computing and Amazon Web Services
PDF
Escalando hasta sus primeros 10 millones de usuarios
PDF
Scale, baby, scale
PPTX
Scaling on AWS to the First 10 Million Users
PDF
일본 시골 개발자의 AWS 활용기 - AWS Summit Seoul 2017
PDF
Scale, baby, scale! (June 2016)
PDF
[판교에서 만나는 아마존웹서비스] Obama for America를 통해서 본 AWS에서의 데이터 분석
PDF
Build an app on aws for your first 10 million users (2)
PDF
Japanese CloudSearch Use-Cases and Tech Deep Dive
PDF
McrUmbMeetup 22 May 14: Umbraco and Amazon
PDF
Presto + Alluxio on steroids a romantic drama on Production with happy end
Scale, baby, scale!
Escalando hasta sus primeros 10 millones de usuarios
What is Amazon Web Services & How to Start to deploy your apps ?
2011 Introduction to Cloud Computing and Amazon Web Services
Escalando hasta sus primeros 10 millones de usuarios
Scale, baby, scale
Scaling on AWS to the First 10 Million Users
일본 시골 개발자의 AWS 활용기 - AWS Summit Seoul 2017
Scale, baby, scale! (June 2016)
[판교에서 만나는 아마존웹서비스] Obama for America를 통해서 본 AWS에서의 데이터 분석
Build an app on aws for your first 10 million users (2)
Japanese CloudSearch Use-Cases and Tech Deep Dive
McrUmbMeetup 22 May 14: Umbraco and Amazon
Presto + Alluxio on steroids a romantic drama on Production with happy end
Ad

More from Eiji Shinohara (20)

PDF
Indexing with Algolia Ruby API Client
PDF
Getting Started Algolia with InstantSearch.js
PDF
Algolia introduction in Kanazawa - July 2019
PDF
Scalable and Cost Effective Systems Architecture on AWS
PDF
#AWSAdTechJP
PDF
Accelerating AdTech on AWS in Japan
PDF
AWS Summit New York 2017 Keynote Recap
PPTX
#CTONight powered by AWS
PDF
SolrCloud on Amazon ECS
PDF
AWS Summit San Francisco 2017 Werner Vogelsによる基調講演を徹底紹介
PDF
Search Solutions on AWS
PDF
Global AWS AdTech use-cases
PDF
IVS CTO Night and Day Recap - #CTONight 2016 Winter
PDF
Tips for getting the most out of AWS re:Invent IN ENGLISH
PDF
検索技術の活用による広告配信Relevance向上
PDF
エンジニアの為のAWS実践講座
PDF
AWS Summit New York 2016 Recap : AWS Application Load Balancer and Amazon ECS
PDF
個人的にAmazon EMR5.0.0でSpark 2.0を使ってZeppelinでSQL集計してみる
PDF
Accelerating AdTech on AWS #AWSAdTechJP
PDF
IVS CTO Night and Day Recap - #CTONight 2016 Spring
Indexing with Algolia Ruby API Client
Getting Started Algolia with InstantSearch.js
Algolia introduction in Kanazawa - July 2019
Scalable and Cost Effective Systems Architecture on AWS
#AWSAdTechJP
Accelerating AdTech on AWS in Japan
AWS Summit New York 2017 Keynote Recap
#CTONight powered by AWS
SolrCloud on Amazon ECS
AWS Summit San Francisco 2017 Werner Vogelsによる基調講演を徹底紹介
Search Solutions on AWS
Global AWS AdTech use-cases
IVS CTO Night and Day Recap - #CTONight 2016 Winter
Tips for getting the most out of AWS re:Invent IN ENGLISH
検索技術の活用による広告配信Relevance向上
エンジニアの為のAWS実践講座
AWS Summit New York 2016 Recap : AWS Application Load Balancer and Amazon ECS
個人的にAmazon EMR5.0.0でSpark 2.0を使ってZeppelinでSQL集計してみる
Accelerating AdTech on AWS #AWSAdTechJP
IVS CTO Night and Day Recap - #CTONight 2016 Spring

Recently uploaded (20)

PPTX
Programs and apps: productivity, graphics, security and other tools
PDF
Encapsulation theory and applications.pdf
PPTX
Effective Security Operations Center (SOC) A Modern, Strategic, and Threat-In...
PDF
Review of recent advances in non-invasive hemoglobin estimation
PDF
Peak of Data & AI Encore- AI for Metadata and Smarter Workflows
PDF
KodekX | Application Modernization Development
PDF
Chapter 3 Spatial Domain Image Processing.pdf
PDF
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
PDF
Build a system with the filesystem maintained by OSTree @ COSCUP 2025
PDF
Dropbox Q2 2025 Financial Results & Investor Presentation
PDF
Advanced methodologies resolving dimensionality complications for autism neur...
PPTX
20250228 LYD VKU AI Blended-Learning.pptx
PDF
Reach Out and Touch Someone: Haptics and Empathic Computing
PDF
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
PPT
Teaching material agriculture food technology
PDF
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
PDF
Approach and Philosophy of On baking technology
PPTX
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
PPTX
VMware vSphere Foundation How to Sell Presentation-Ver1.4-2-14-2024.pptx
PDF
Unlocking AI with Model Context Protocol (MCP)
Programs and apps: productivity, graphics, security and other tools
Encapsulation theory and applications.pdf
Effective Security Operations Center (SOC) A Modern, Strategic, and Threat-In...
Review of recent advances in non-invasive hemoglobin estimation
Peak of Data & AI Encore- AI for Metadata and Smarter Workflows
KodekX | Application Modernization Development
Chapter 3 Spatial Domain Image Processing.pdf
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
Build a system with the filesystem maintained by OSTree @ COSCUP 2025
Dropbox Q2 2025 Financial Results & Investor Presentation
Advanced methodologies resolving dimensionality complications for autism neur...
20250228 LYD VKU AI Blended-Learning.pptx
Reach Out and Touch Someone: Haptics and Empathic Computing
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
Teaching material agriculture food technology
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
Approach and Philosophy of On baking technology
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
VMware vSphere Foundation How to Sell Presentation-Ver1.4-2-14-2024.pptx
Unlocking AI with Model Context Protocol (MCP)

Japanese Startup Use-Cases and Tech Deep Dive