Microservice Architectures for Scalability, Agility …eprints.uni-kiel.de/37490/1/ICSA2017presentation.pdfDevOps & Software Architecture ICSA, 05.04.2017 W. Hasselbring & G. Steinacker

Microservice Architectures for Scalability, Agility and Reliability in E-Commerce

Prof. Dr. Wilhelm (Willi) HasselbringSoftware Engineering Group, Kiel University, Germany

http://se.informatik.uni-kiel.de/Competence Cluster Software Systems Engineering

http://kosse-sh.de/

Guido SteinackerOtto GmbH & Co KG

Werner-Otto-Straße 1–7D-22179 Hamburg, Germany

https://dev.otto.de/

ICSA, 05.04.2017 W. Hasselbring & G. Steinacker 1

Context: otto.de

ICSA, 05.04.2017 W. Hasselbring & G. Steinacker 2http://www.ottogroup.com/de/presse/material.php

Web Shop until 2013


https://support.intershop.com/kb/index.php/Display/276B90

Scaling such Web Information Systems

ICSA, 05.04.2017 W. Hasselbring & G. Steinacker

Source: [Abbott & Fisher 2015]

Approaches to Scalability on the database layer:• Big enterprise server• Database replication• Database sharding

4

However, you have to scale everything to scale anything!

Agenda

1. Context2. Microservice Architectures for Scalability3. Microservice Architectures for Agility4. Microservice Architectures for Reliability5. Scalable Development of Microservices6. Takeaways


Polyglot Persistence


Page Assembly

Backend Integration

Prod

uct

Trac

king

Rec

omm

en-

datio

n

Ord

er

Sear

ch

Wide Column

Database

Full-TextDatabase

GraphDatabase

RelationalDatabase

Time Series

Database

6

[Hasselbring 2016]

Project Lhotse 2011-2013• In 2011, Otto started a complete re-implementation of

their ecommerce software from scratch.• The drivers for this decision were diverse, but had

mostly to do with non-functional requirements like scalability, performance and fault tolerance. – Regarding scalability, they were not only thinking about

technical scalability in terms of load or data. – They needed a solution that was scaling with respect to

the number of teams and/or developers working on the software at a given time.

– In addition to that, they planned to practice DevOps including continuous deployment, in order to deliver features quickly to the customer.


https://www.otto.de/unternehmen/de/newsroom/dossiers/lhotse.php

Modernization Strategy

• What they have found was in the first place a little bit unusual, but in the end highly successful:– Instead of setting up a single development team to create a new

platform for the shop, they were actively employing Conway’s Law by starting development with initially four separate teams with four loosely coupled applications (a.k.a. microservices):

• Product, being responsible for products and their presentation.• Order for shopping carts and the order process.• Promotion, serving product recommendations and promotions for assortments,

brands, and so on.• Search and Navigation for search and navigation in the shop.

• In the following years, they founded several more teams and systems.


Verticals at otto.de


Back

offic

e

Shop

page

s

Sear

ch &

Nav

igat

ion

Prod

uct

Prom

otio

n

Ord

er

Use

r

Afte

rSal

es

Auth

Insi

ghts

Link

Han

dler

Trac

king

Page Assembly Proxy

Backend Integration Proxy

Verticals and Microservices at otto.de


Page Assembly Proxy


...

This architecture allows elasticity for horizontal scalability[van Hoorn et al. 2009, von Massow et al. 2011, Hasselbring 2016]

Agenda



DevOps & Software Architecture


“The deployment pipeline is the place where the architectural aspects and the process aspects of DevOps intersect.”

[Bas et al. 2015]

12

Scaling Delivery Pipelines at otto.deContinuous Deployment and DevOps


Goals:• Set up new pipelines quickly and automatically using a templating mechanism.• Being able to build pipelines just like the microservices from tested code

maintained in a Git repository.• Because most of the time, pipelines are simply waiting to be triggered, they aim

at a solution that does not waste hardware resources while not in use.Solution: LambdaCD (internal domain-specific language written in Clojure)

Life Deployments @ Otto.de


Agenda



Deployment Pipeline with Quality Gates


[Breetzmann et al. 2014][Farley 2007]

Automated Quality AssuranceExample: Regression Benchmarking


Integrated into Continuous Integration Setup[Waller et al. 2015]

Should include automated anomaly detection [Marwede et al. 2009, Ehlers et al. 2011]

17

https://build.se.informatik.uni-kiel.de/jenkins/job/kieker-nightly-release/plot/

Life Deployments & Incidents @ Otto.de


Agenda



Conway’s Law“The basic thesis of this article is that organizations which design systems […] are constrained to produce designs which are copies of the communication structures of these organizations”

[Conway 1968]


If the organizational structure is decomposed vertically and according to the microservices structure into cross-functional feature teams, • scaling development capacities according to changing business

requirements is enabled.• The feature teams should be highly independent, having

members of all roles and skills that are required to build and maintain their microservice.

Decoupling teams as relevant as decoupling software modules

Component vs. Middleware Reuse

Example:https://github.com/otto-de/


Shared Component C

Component A Component B

Component A

Component BOpen Source Middleware

Microservice Architectures for Scalability, Agility and

Reliability


Scalability for both, runtime performance and development performance.

Reliability

Backup


Microservice Architecture Variations


“Scalability is managed by each service individually and is included in its SLA in the form of a guaranteed response time given a particular load.”

[Bas et al. 2015, Chapter 4]

“The trade-off between many small components and a few large components must be considered in component and system design.”

[Hasselbring 2002a]

[Steinacker 2014][Bas et al. 2015]

24

[Kraus et al. 2013]

Back

offic

e

Shop

page

s

Sear

ch &

Nav

igat

ion

Prod

uct

Prom

otio

n

Ord

er

Use

r

Afte

rSal

es

Auth

Insi

ghts

Link

Han

dler

Trac

king

Page Assembly Proxy


References[Abbott & Fisher 2015] M.L. Abbott, M.T. Fisher: The Art of Scalability. Addison-Wesley, 2nd Edition, 2015.[Bas et al. 2015] Len Bass, Ingo Weber, Liming Zhu: “DevOps: A Software Architect’s Perspective”, Addison-Wesley 2015.[Breetzmann et al. 2014] R Breetzmann, S. Kraus, C. Stamm: “Null Toleranz für Fehler: Wie wir auf otto.de die Qualität hoch halten“, OBJEKTspektrum

4/2014, 18-23.[Conway 1968] M.E. Conway: How do committees invent? Datamation, 14(4):28-31, April 1968.[Ehlers et al. 2011] J. Ehlers, A. van Hoorn, J. Waller, W. Hasselbring: Self-Adaptive Software System Monitoring for Performance Anomaly Localization. In:

8th IEEE/ACM International Conference on Autonomic Computing (ICAC 2011).[Farley 2007] D. Farley: “The Deployment Pipeline”. ThoughtWorks, Inc. https://continuousdelivery.com/, 2007.[Hasselbring 2000] W. Hasselbring: “Information System Integration” In: Communications of the ACM, 43(6): 32-36, June 2000.[Hasselbring 2002a] W. Hasselbring: “Component-Based Software Engineering” In: Handbook of Software Engineering and Knowledge Engineering. World

Scientific Publishing, Singapore, pp. 289-305, 2002.[Hasselbring 2002b] W. Hasselbring: “Web Data Integration for E-Commerce Applications,” IEEE Multimedia, vol. 9, no. 1, pp. 16–25, 2002.[Hasselbring 2006] W. Hasselbring: “Software-Architektur”. In: Informatik-Spektrum 29(1): 48-52, February 2006.[Hasselbring 2016] W. Hasselbring: “Microservices for scalability: Keynote talk abstract,” In: Proceedings of the 7th ACM/SPEC on International

Conference on Performance Engineering (ICPE 2016). ACM, 2016, pp. 133–134.[Kraus et al. 2013] S. Kraus, G. Steinacker, O. Wegner: “Teile und Herrsche – Kleine Systeme für große Architekturen“, OBJEKTspektrum 5/2013, 8-13.[Marwede et al. 2009] N. Marwede, M. Rohr, A. van Hoorn, W. Hasselbring: “Automatic Failure Diagnosis in Distributed Large-Scale Software Systems

based on Timing Behavior Anomaly Correlation“, In: 13th European Conference on Software Maintenance and Reengineering (CSMR 2009).[Steinacker 2014] G. Steinacker: Scaling with Microservices and Vertical Decomposition. http://dev.otto.de/2014/07/29/scaling-with-microservices-and-

vertical-decomposition/, 2014.[van Hoorn et al. 2009] A. van Hoorn, M. Rohr, I.A. Gul, W. Hasselbring: “An Adaptation Framework Enabling Resource-efficient Operation of Software

Systems” In: 2nd Warm Up Workshop for ACM/IEEE ICSE 2010 (WUP 2009) . [van Hoorn et al. 2012] A. van Hoorn, J. Waller, W. Hasselbring: “Kieker: A Framework for Application Performance Monitoring and Dynamic Software

Analysis”, In: 3rd joint ACM/SPEC International Conference on Performance Engineering (ICPE 2012).[von Massow et al. 2011] R. von Massow, A. van Hoorn, W. Hasselbring: “Performance simulation of runtime reconfigurable component-based software

architectures” In: Proceedings ECSA 2011, 2011, pp. 43–58.[Waller et al. 2015] Jan Waller, Nils Ehmke, Wilhelm Hasselbring: “Including Performance Benchmarks into Continuous Integration to Enable DevOps“, In:

ACM SIGSOFT Software Engineering Notes, 40(2).


Documents

Microservice Architectures for Scalability, Agility …eprints.uni-kiel.de/37490/1/ICSA2017presentation.pdfDevOps & Software Architecture ICSA, 05.04.2017 W. Hasselbring & G. Steinacker