AWS Summitに参加したのでメモ
昨年同様受付は長蛇の列。11000人が申し込んで延べ8000人きたらしい
資料:「AWS Summit Tokyo 2014」開催レポート動画・資料一覧
Day 1: 2014/07/17
Day 1 基調講演 ~No Boundary~
Werner Vogels(CTO, Amazon.com,Inc.) and guests
By using AWS, let’s convert fixed cost to varied cost just like utilizing temporary workers. Then we can cope with business uncertainty, limited resource, and harder competition.
Since 2006, AWS offered 66 times of price down.
1 |
If you wan to increase innovation, you have to lower the cost of failure.(Jo Ito Director of MIT media lab) |
Innovation the Amazon way
Adam FitzGerald(Head of World Wide Developer Marketing, Amazon Web Services,Inc.)
1 2 3 4 5 6 7 |
Key elements of innovation: Culture, Organization, Architecture, Mechanism With 2 pizza team, build the Minimum Viable Product using Agile Development techniques. Continuous Integration & Continuous Deployment Every Wednesday they have strategy MTG with each service owner (checking # of customer, removing friction, setting priorities,...) Important work: PRESS RELEASE, FAQ, USER MANUAL <img class="alignnone wp-image-187" src="http://6109.hidepiy.com/wp-content/uploads/2017/01/20140718innovation.jpg" alt="" width="602" height="67" /> |
ユーザーの趣味嗜好に適した広告配信システムDynalystができるまで
木村 衆平(株式会社サイバーエージェント AdTechSTUDIO, Dynalyst)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
# current trend of online display ad - detailed targeting (user behavior, budged allocation & ROI improvement) - realtime (50msec or die), RTB # what is Dynalyst RealTime User Personalized Dynamic Ad service Provider - RealTime - User Personalized Ad` # Dynalist internal small team (5 or 6) * short term (2, 3 month) hundreds of million IMP / day -> Use managed services as much as possible (DynamoDB & ElasticCache) - Track who, when, which item, look or purchased - Delivery appropriate ad to user - Report timely to clients # ElasticCache (Redis) Detecting each user is valid for targeting or not. Target ratio is less than 20%, so filtering 80% before DynamoDB High performance # DynamoDB Data structure differs for each clients -> schema less High throughput, low latency, and rare operation. Write capacity can be modified align with clients' ad tag implementation. |
NoSQL Strategy on AWS
David Pearson(Business Development Manager, AWS Business Development, Amazon Web Services, Inc.)
Publicity of AWS
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 |
Keys:Performance, Reliability, TCO (Total Cost of Ownership), Time to management Price policy of AWS: Loop of use AWS -> AWS get bigger -> lower infra price -> affords lower price -> more user -> Service policy of AWS: new service -> more use case -> more user -> more challenge -> prioritize requirements -> RDBMS -> Distributed NoSQL -> Cloud managed NoSQL service ElasticCache: in-memory cache, easy deploy, operate & scale supports both protocol of memcached & redis Redshift: Data Warehouse for Analytics S3: Log store DynamoDB: NoSQL assured performance & scalability fast, high availability (five nines), no operation, scalable, robust, low cost performance is flat even if # of requests increase -> consistent performance Write: Continuous persistence Read: Strong Consistency or Eventual Consistency RDS: Complicated query & transaction MySQL, ORACLE, SQL Server, PostgreSQL CloudSearch: Rich search - specialize, select appropriate database for each work road - optimize, minimize cost by scaling - automation (managed service), reducing cost <img class="alignnone wp-image-183" src="http://6109.hidepiy.com/wp-content/uploads/2017/01/20140718DSC_1141.jpg" alt="" width="597" height="336" /> |
【パネルディスカッション】クラウド時代の運用
..many
1 2 3 4 5 6 7 8 9 10 11 |
45-75% is operation cost (nikkeiBP) Issues in operation - High load -> complicated task - Depends on individual -> black box - invisible ROI -> decreasing return Caution in automation - automation is result, not purpose (purpose is to avoid dependency) - bad: maintenance of automation process depends on individual - bad: no one know the internal of automation -> cannot improve the process - bad: cannot recovery |
Day 2: 2014/07/18
【パネルディスカッション】徹底討論!ゲームプラットフォーマーのクラウド活用法
小野篤司(株式会社ディ・エヌ・エー システム本部 IT基盤部 部長)
梶原大輔(グリー株式会社 執行役員 インフラストラクチャ本部長)
清水 勲(株式会社ミクシィ クロスファンクション本部 システム統括室 第2グループ リーダー)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 |
# Ono Financial Sier -> DeNA in charge of Infra writer of Mobageを支える技術 Regarding failover, they think it's hard to keep same quality, then runnig MySQL on EC2 with their own failover func - why AWS introducing games to Facebook Oversea scale Still Onpremise is reasonable price, but diff is decreasing - hybrid not started - expect linkage with other strage vendors currently region <-> region is internet, and other vendor supports direct connect - skill admin -> programmable # Kajidai GREE HybridEnvironment On-Premise openstack(done), docer, KVM cloud AWS - why AWS everyyear checking cost CDN: Akamai -> CloudFront oversea common language, comunity regarding Infra - hybrid under investigation (done by common line) -> try - expect there are a lot of diff in API b/w openstack please lead now Amazon is winner (not AWS) rival # Shimizu Mixi infra Openstack DC & AWS is directly connected by 10G - why AWS speedy - hybrid small latency |