Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
The Rocky Road from Monolithic to Microservice...
Search
cnu
November 25, 2017
Programming
0
1.1k
The Rocky Road from Monolithic to Microservices Architecture
The explanation of our Microservices architecture and the lessons we learnt from it.
cnu
November 25, 2017
Tweet
Share
More Decks by cnu
See All by cnu
Redisconf 2018: Probabilistic Data Structures
cnu
1
980
Probabilistic Data Structures
cnu
0
640
AWS Lambda - Pycon India 2016
cnu
0
510
ZeroMQ - PyCon India 2013
cnu
2
1.5k
Other Decks in Programming
See All in Programming
Ruby Parser progress report 2025
yui_knk
1
460
1から理解するWeb Push
dora1998
7
1.9k
個人開発で徳島大学生60%以上の心を掴んだアプリ、そして手放した話
akidon0000
1
150
知っているようで知らない"rails new"の世界 / The World of "rails new" You Think You Know but Don't
luccafort
PRO
1
190
Kiroで始めるAI-DLC
kaonash
2
620
OSS開発者という働き方
andpad
5
1.7k
Updates on MLS on Ruby (and maybe more)
sylph01
1
180
詳解!defer panic recover のしくみ / Understanding defer, panic, and recover
convto
0
250
「待たせ上手」なスケルトンスクリーン、 そのUXの裏側
teamlab
PRO
0
570
AWS発のAIエディタKiroを使ってみた
iriikeita
1
190
はじめてのMaterial3 Expressive
ym223
2
900
「手軽で便利」に潜む罠。 Popover API を WCAG 2.2の視点で安全に使うには
taitotnk
0
870
Featured
See All Featured
Intergalactic Javascript Robots from Outer Space
tanoku
272
27k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
252
21k
Art, The Web, and Tiny UX
lynnandtonic
303
21k
Docker and Python
trallard
46
3.6k
XXLCSS - How to scale CSS and keep your sanity
sugarenia
248
1.3M
The Invisible Side of Design
smashingmag
301
51k
Cheating the UX When There Is Nothing More to Optimize - PixelPioneers
stephaniewalter
285
14k
Embracing the Ebb and Flow
colly
87
4.8k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
507
140k
Into the Great Unknown - MozCon
thekraken
40
2k
JavaScript: Past, Present, and Future - NDC Porto 2020
reverentgeek
52
5.6k
Fashionably flexible responsive web design (full day workshop)
malarkey
407
66k
Transcript
THE ROCKY ROAD FROM MONOLITHIC TO MICROSERVICES ARCHITECTURE
THE ROCKY ROAD FROM MONOLITHIC TO MICROSERVICES ARCHITECTURE
SRINIVASAN RANGARAJAN Head of Product Engineering
SRINIVASAN RANGARAJAN https://cnu.name Twitter: @cnu Github: @cnu
RETAIL AUTOMATION PRODUCT
Catalog & User Events Processing Recommendation
MONOLITHIC ARCHITECTURE
MONOLITHIC ARCHITECTURE Image Processing API Image Searcher File Storage
MINIMUM TWO SERVERS BEHIND ELB Load Balancer
CLOUD VS BARE METAL
EXPENSIVE
NOT REALTIME
NOT PERSONALIZABLE
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine
GOTHAM
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine
JOKER • Convert client’s catalog into one common MAD Format
• Normalization of fields and metadata • Can process batch and streaming data • Major cause of chaos in the system
GORDON • Routes the product metadata to the right micro
services • Is it a new product? or update to an existing product? • Streaming data from AWS SQS
WONDER WOMAN • Not a microservice, But a tool used
to generate rules for the catalog • Rules are send to the Image Processing microservice • Works on Samples of data and not entire dataset
WATCHTOWER • Central Source of Truth for all metadata •
Backed by an RDBMS Database (Postgresql) • Input via SQS and REST API • Output via REST API
INGESTION Gordon Joker Joker Joker Wonder Woman Watchtower Next Stage
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine
NIGHTWING • Computer Vision and Deep Learning Models • Convert
Image to high dimensional vectors • Tag image with visual attributes • Computer Intensive
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine
BATMAN • Custom very fast Vector Indexer and Search Engine
• Stores everything in memory • Two sub-parts: Indexer and Searcher • Store binary information about image in DynamoDB
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine
SUPERMAN • User behaviour based recommendation • Multiple products like
Collaborative filtering, Cross Product recommendation • Records every user event data and stores in a data warehouse
TWO FACE • Individual User level Personalization • Shows a
different “face” to each user • Dynamic and realtime
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine
FLASH • Very fast data structure storage - redis instance
• User session level history, Product Availability, etc. • Fast access, but non- expirable
GCPD • “Global Cache for Products Digested” • Rough first
level of cache for the results
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine
ROBIN • API Gateway for all our products • Combines
data from other micro services like Batman, Two Face, Watchtower, Superman, etc and returns JSON Response
API Data Store Ingestion Image Processing Image Searcher User Event
Personalization Engine Joker, Gordon, Wonder Woman Nightwing Batman Robin Watchtower, GCPD, Flash Superman, Two Face
LESSONS WE LEARNT
START WITH A MONOLITH. CHIP OFF PIECES AND BUILD THE
MICROSERVICES. Lesson 0
–Melvin Conway “… organizations which design systems ... are constrained
to produce designs which are copies of the communication structures of these organizations."
DEPLOY HETEROGENOUS MICROSERVICES IN A SINGLE SERVER Lesson 1
Compute Optimized Server Memory Optimized Server Nightwing Batman Robin Robin
Robin Joker Joker Joker Gordon Watch tower Watch tower Joker
IMMUTABLE MICROSERVICES Lesson 2
Constable Inspector Assistant Commissioner Commissioner
ASYNCHRONOUS IS BETTER THAN SYNCHRONOUS Lesson 3
None
NOT ALL MICROSERVICES NEED TO BE SERVERS Lesson 4
ADD REQUEST ID OR TRANSACTION ID TO DEBUG EASILY Lesson
5
GIVE A CHARACTER TO YOUR MICROSERVICES Lesson 6
None
THANK YOU
http://cnu.name/talks/ @cnu