Microservice architecture at ASOS - CloudBrew/AZUG.be

© 2016 ||| ASOS Plc@aliostad @dav3green

Microservice Architecture at ASOS

Ali Kheyrollahi


ASOS

Established at 2000

Global Fashion destination for “20-something”.

Hires +2500 staff & +250 in tech (700)

Grown average of 35% YoY

39th biggest global online retailer


In Numbers

A c t i v e C u s t o m e r s → 12 M

N e w P r o d u c t s / w k → 4 k

U n i q u e V i s i t s / m o → 123 M

P a g e V i e w s / d a y → 95 M

P l a t f o r m T e a m s → 40

A z u r e D a t a C e n t r e s → 5

T u r n O v e r 2 0 1 6 → 1.5b


Nick Beighton - CEO from “IT Cascade” slides - Oct 2014

ASOS - A Tech Company


ASOS Stock Price: 2009-2013

More than 2000% growth!


ASOS Stock Price: 2009-2014

Multiple IT project failures

Tech issues disrupting promotions

Overdue features were not delivered


Microservices…

why?


“Can I haz microservicez?”

“It is all a marketing hoax.”

“I dunno…” The sceptic…


Scaling People not the solution

↓Complexity of each service at the cost of ↑overall solution

Frequent and independent Deployments

Decentralising decision centres

Governance and EA is different

Auftragstaktik Doctrine is the conceptual underpinning of HOW to think and operate effectively; teaching leaders WHAT to think is dogma…

Auftragstaktik encourages commanders to exhibit initiative, flexibility and improvisation while in command…

In what may be seen as surprising to some, Auftragstaktik empowers commanders to disobey orders and revise their effect as long as the intent of the commander is maintained…


SOA Microservices

Main Goal Architectual Decoupling Agility

Audience Mainly Architecture Everyone (inc Buisness)

Set out to solve Architectural CouplingScaling People,

Frequent DeploymentImpact on Structure of

OrganisationMinimal Huge

Service Cardinality Usually up to a dozen >40 (Commonly >100)

When to do Always teams > ~5**

Law Conway’s Reverse Conway’s


Principles (ASOS)

* All queries and commands through HTTP API (No ESB-like pseudo-Microservices!)

*Microservices can subscribe to events raised by other Microservices

**

Each Microservice owns a business responsibility and defines a clear boundary for communication (APIs and Events)

They own their data (all access to data through API only)

* Each Microservice is realised in one (sometimes more) physical components


Domain Modelling

Operating ModelPeople

Successful Architecture


ASOS Architecture


ASOS Journey


Two-speed ITEnterprise Domain

Digital Domain

• Predominantly “Buy” • Integrated at top tier • Minimal Engineering • Project-centric

• Predominantly “Build” • Drives sales and customer

touchpoint • Product-centric • Downtime unacceptable


❤ASOS


ASOS Tech Stack


Digital Domains & PlatformsThis is Domain

This is Platform

Logical Services

•“Domains” are ASOS’ organisational structure for managing Platform Teams

• Platform Teams look after collections of aligned services. They are accountable for the full lifecycle management



Operating ModelIndicative Team Numbers

6 Digital Domains

19 Platforms

35 Dev / Scrum teams

24 Solution Architects

700+ people in Technology (2500 in Asos total)

Avg +20 people per month over last year


Governance

Bottom-Up

Top-Down


Microservice Example:

Recommendations


- Recommendations as a destination

- Widgets

- Personalisation



1 x Soft Store (Redis)

5 x Redis Caches

3 x Cloud Services (Microservie)

2 x SQL Databases

5 x Table Storages

1 x External API

1 x Elasticsearch cluster


<SQL> Models

<uservice> Recommendations

API

<uservice> Composition

API

<uservice> Feedback

API

<External API> Product Catalogue

API

<SQL> Signal

<Redis> Realtime

<Elasticsearch> Catalogue

Computation/Retrieval Filtering Faceting

Authentication Composition Pagination


LMA Platform


A P

I

Application Logging

Microservice Component

cidETW

cid

Thread.CurrentThread.SetLogicalData(…) [EventSource.ActivityId does not flow

over async methods]

SLAB Azure Table Storage Sink

Application Code

cid

To other APIs

cid

Raising Events

Listener

EC

cid

IIS Logs


A P

I

PerformanceMicroservice Component

WAD Windows Azure

Monitoring Agent

CPC

CPC: Custom Performance Counters Inst: Instrumentation

Perfit Library

SLAB Azure Table Storage Sink

Application Code

Inst

Inst

Call to Data Stores or other services

CPC


WoodpeckerFor `Pull` (rather than `Push`) metrics

Queue Depth/Size

Azure SQL Diagnostics

Canaries and Health Endpoints


Logsink API

For Native and Web Clients

A P

I

Logsink

/logchannels/<channel>

GET /logchannels/channelOne?a=b&c=d HTTP1.1

200 OK Content-Type: image/gif

POST /logchannels/channelOne?a=b HTTP1.1 {

“c”: “d” }

202 Accepted

POST /logchannels/_bulk HTTP1.1 [

{“channel”: “…”, “payload”: {...} } ]

202 Accepted [

{“status”: 202}, ... ]

Channels Config

EventHub


ConveyorBelt

Performance Counters

ConveyorBelt

Azure WAD logs

ETW Logs

Elasticsearch

Highly Available Headless Cluster shovelling data to ES

Instrumentation Logs

IIS Logs

Woodpecker Outputs (Pull Logs)

Sources Config

Up to 1TB/day


Monitoring (Demo)


AlertingElasticsearch

Platform Team LMA Support

Watch

OAT Spec

EC

1st-2nd line Support

- xxx EC seen more than 10 times over the last minute. Back-off for 15 minutes - 90th percentile of API response > 100ms over

the last hour... Back-off for an hour


Lessons Learnt‣In cloud? Design for failure… “Cloud is a Jungle”

‣Network latency & failures add up: understand and optimise time from

the user to the data

‣Using any cloud technology? Forget the hype and trust no one: test,

measure, adopt/drop, monitor, engage with your provider

‣Managing platform costs can be difficult

‣Distributed computing is hard, making it geo-distributed even harder

‣Expect to roll your sleeves up: maturity in a lot of areas is can be low

‣ (platform, tooling, skills, supporting technologies) but is changing rapidly

‣At this scale, owning your LMA data and process very important


ASOS is Hiring!If you are intrigued by what we do,

we would love to hear from you!

Just get in touch with us in twitter @aliostad @dav3green


Q AThank You!


References and external artwork used

Adrian Cockcroft talk: Simplifying the future

Work in progress: Picture

Rusty Rolls Royce: Picture

Tesla S Blueprint: Picture

Two-Speed (Slow Lane, Fast Lane): Picture

Copyright of the artworks listed here belong the owners specified in the links below

https://www.youtube.com/watch?v=DGK6jjamzfY

http://cucsa.org.uk/wp-content/uploads/2015/10/Work_In_Progress.png

https://farm4.static.flickr.com/3004/2429436972_3947cfd9c6.jpg

https://s-media-cache-ak0.pinimg.com/736x/a6/24/e4/a624e484d3cde58f8133dff4231488dd.jpg

http://i.telegraph.co.uk/multimedia/archive/02886/10772181_2886857b.jpg

Microservice architecture at ASOS - CloudBrew/AZUG.be

Software