This document discusses improving the performance of IoT application software on a cloud platform using a microservices architecture with Kubernetes as the container orchestrator in a production environment. It proposes load testing the IoT applications running on a Kubernetes cluster when massive loads are applied. It also discusses analyzing performance based on results from load tests involving large-scale concurrent device traffic sending data to the cloud. The document outlines various parameters for configuring and scaling the system horizontally under increasing IoT workloads.

1. PENINGKATANPERFORMASKALABILITASAPLIKASIIOTBERBASISCLOUD
DENGAN ARSITEKTURMICROSERVICEMENGGUNAKANKUBERNETES
CLUSTER SEBAGAI CONTAINER ORCHESTRATOR
PADAPRODUCTIONENVIRONMENT
Alfian Firmansyah 1706985874
“Art Mage”, illustrated by Surya Asmoro
Available at https://www.artstation.com/artwork/w8RQA9
Pembimbing:
Dr. Ruki Harwahyu, S.T., M.T., M.Sc.
SEMINAR PROPOSAL

5. https://cloudcow.com/content/5-possible-ways-5g-will-impact-cloud-computing/

7. ARSITEKTURIOT-CLOUD

8. Combined Powers
Raises a new
HERO
in the world of
Indonesische
. . .

9. Only
The IoTMyth
5G
CLOUD
Was able to hold
Such a massive
powers
“Haam Tir-Ilmn”, illustrated by Surya Asmoro
Available at https://suryaasm.artstation.com/projects/aR8qe2

10. Bagaimana Performa throughput database, latensi pemrosesan server aplikasi, resource
utilization, dan bandwidth utilization dalam uji coba Kubernetes skala besar?
1
Bagaimana respon dan kinerja aplikasi IoT yang berjalan pada cloud kubernetes
cluster ketika massive load test dioperasikan?
2
Apabila mengalami masalah, apakah solusi yang tepat untuk mengatasi masalah
tersebut?
3

11. Bagaimana merancang arsitektur aplikasi berbasis cloud yang sesuai dengan karakteristik
real-time data IoT?
4
Bagaimana melakukan deployment aplikasi IoT secara kontinyu dan memastikan
bahwa proses deployment tersebut merupakan metode yang efektif?
5

12. Final Quest
Perencanaan pengembangan aplikasi IoT berbasis cloud yang
sederhana, adaptif, responsif, interaktif, memiliki sistematika
protokol untuk bisa terkoneksi dengan perangkat IoT beserta real-
time data monitoring-nya, penjabaran solusi komponen sistem
IoT, dan use case yang dapat diterapkan dari hasil aplikasi IoT
tersebut.
Membuktikan skalabilitas dari aplikasi IoT berbasis cloud yang
disebar dengan menggunakan arsitektur microservice dengan
Cloud Kubernetes Cluster.
Menganalisis performa berdasarkan hasil yang didapatkan dari
load tests yang datang dari traffic concurrent perangkat IoT skala
besar yang mengirimkan data ke cloud.
. . .

13. Dungeon Scopes
- Kubernetes AWS EKS
- MEAN Stack
- MQTT & REST
- CI: Github, CD: CodePipeline,CodeBuild, ECR
- SSL/TLS Security
. . .

14. Internet ofThings
Cloud Computing
Arsitektur
Perangkat Lunak
Protokol IoT
Containerization
Kubernetes dan
Arsitekturnya
Framework Aplikasi
Framework
Monitoring
Dasar Teori
”Ice Berg”, illustrated by Surya Asmoro
Available at https://suryaasm.artstation.com/projects/qA1zaR

15. Client D (IoT ACTUATOR/SENSOR)
Client C (GENERAL APP)
VISUAL ML
Service D Service C
DB A DB B DB C
Service B Service A
REST
API
REST API
REST API
REST API
API Gateway (BACKEND)
REQ/RES
RESPONSE HTML
Client A (DESKTOP APP)
REQ/RES
RESPONSE HTML
Client B (MOBILE APP)
REQ/RES
RESPONSE HTML
RESPONSE HTML
CONTENT ROUTER (FRONTEND)

16. OVERLAY NETWORK
API Server
Scheduler
Controller
Manager
etcd
PHYSICAL NETWORK
kubelet
kube-proxy
Container Runtime
Container
1
Container
2
Container
3
Pod
Operating System
Infrastructure
Node 1
kubelet
kube-proxy
Container Runtime
Container
1
Container
2
Container
3
Pod
Operating System
Infrastructure
Node 2
Master Node
. . . Node n
Kubectl

17. SMART DEVICES
Sensors
Public DNS Routes
(DNSSEC enabled)
developer:
local/bastion terminal /
kubectl /
REST API
over HTTP
Internet Gateway
Static IP
Network
Load Balancer
(TCP)
Cloud Gateway
Kubernetes
Control Plane
If kubectl
AZ C
(Autoscaling Group)
VM
Worker Node
Worker
Node Group
AZ A
(Autoscaling Group)
VM
Worker Node
Worker
Node Group
AZ B
(Autoscaling Group)
VM
Worker Node
Worker
Node Group
If kubectl
APPLICATIONS
API
Access
Ingress
Auto Scaling
Group
Auto Scaling
Group
Auto Scaling
Group
Kubernetes
Master Node
Components
Send Alert to Clients
Email Service /
SMS Service Notification
CI tools (Commit)
Rasp Pi
Millions
of
Connected
Clients / Subscribers
JSON Processing
to
Subscribers
CI/CD Pipeline
Tools
Container
Registry (build)
Deploy
Container
Send Response Message/Webhook
MQTT PUBLISHER
MQTT(s)
Authentication &
Communication
Webhook
Access Token
Door Lock
Car
Lightbulb
House
Utility
Receive data
& GUI
Dashboard
Visualization
Send Trigger to Application
Receive Trigger
(ON / OFF / VALUES)
Security Group
Connected
Actuator /
Smart-Devices
Scalable MQTT
Broker
Response Actions
(Success / Failed)
Persistent &
Cache Storage
Internet Gateway
Arsitektur
Keseluruhan

18. REST API
to
Application
MQTT
Publish/Subscribe
Broker to Application
Data Flow

19. MQTT Brokers

20. https://www.bevywise.com/blog/high-availability-mqtt-cluster/

21. CodePipeline CodeBuild
Container Registry
Kubernetes
Production
Git Commit
(CI)
Continuous Integration
Continuous Deployment
Strategi CI/CD

22. Thermostat
Thermostat
Publish
MQTT (JSON)
Thermostat
Auth Server
(application)
MongoDB
Persistent
Storage
Redis Caching
for high IOPS
(temporary)
User
Dashboard
AuthID
Request & Response
Publish
Subscribe
Cloud DB NoSQL
Actuator
Actuator
Actuator
Actuator
Send Trigger Data
to Smart Devices
MQTT (JSON)
Scalable MQTT
Broker
Komunikasi
Data MQTT

23. Sensor node
establishing
connection to broker
Broker establishing
publisher's new
connection
Auth server detecting
authentication ID.
Process existing
request and create
response
publish data with the
specified topic
Detecting publish
request, forward to
subscriber to check
authentication
Receive response,
publish success
Publisher Broker Subscriber /
Auth Server
Sensor node
establishing
connection to broker
Broker establishing
subscriber's new
connection
subscribe data with
the specified topic
Detecting subscribe
request, forward to
publisher's channel to
and check
authentication
Receive data from
publisher's topic
Subscriber /
Auth Server
Broker
(a) (b)
Komunikasi
Data MQTT

24. Thermostat
Thermostat
JSON
REST API
(HTTP)
Thermostat
Auth Server
(application)
MongoDB
Persistent
Storage
Redis Caching
for high IOPS
(temporary)
User
Dashboard
/AuthID
/AuthID_2
/AuthID_3
Cloud DB NoSQL
Actuator
Actuator
Actuator
Actuator
Send Trigger Data
to Smart Devices
JSON
Komunikasi
Data REST

25. Komunikasi
Data REST
Sensor node
establishing
connection to server
server establishing
client's new
connection
Auth server detecting
authentication ID.
Process existing
request and create
response
send JSON data with
REST
Receive response,
success
Client Auth Server

26. Redis DB
Request
Web App
in Browser
Internet
MongoDB
API
Microservice
Response
GET /data
Accept: application/json
Get the response
and putting in Redis
HTTP/1.1 200 OK
Content-Type: application/json
First request, apps checks data is present in redis
if yes, will return data from redis
else call to the web service to get data from mongo
and put the response in redis
so next time response will be return from redis
Strategi Caching pada
Redis

27. EMQX Broker
Redis
IoT Application
Redis
Horizontal
scaling of IoT
Application
over MQTT
Broker
Horizontal
scaling of IoT
Application
over HTTP
IoT Application
IoT Application
IoT Application
Stress Tools
IoT Application
EMQX Broker
Redis
Horizontal
scaling of MQTT
Broker
EMQX Broker
Horizontal
scaling of Redis
over MQTT
Broker
Horizontal
scaling of Redis
over HTTP
IoT Application
Mongo
Redis
Redis
IoT Application
Mongo
Redis
Redis
EMQX Broker
Horizontal
scaling of MongoDB
over MQTT Broker
Horizontal
scaling of MongoDB
over HTTP
IoT Application
Mongo
Redis
IoT Application
Redis
EMQX Broker
Mongo
Mongo
Mongo
Skenario Penskalaan
Horizontal

28. Parameter Configuration
Average Load from the Southbound
1000, 10000, 100000, 1000000
requests per second
Payload size 10 bytes, 100 bytes, 1000 bytes
IoT Protocol Type MQTT, RESTful API over HTTP(s)
IoT Application Pod Resource 300 mCPU; 256 MB
IoT Application Instance To Be Decided
Broker Cluster Instance To Be Decided
mongoDB Instance To Be Decided
Redis Instance To Be Decided
Broker Cluster Instance To Be Decided
Locust.io (stress tool) Instance To Be Decided
MQTT benchmark (stress tool) Instance To Be Decided
Parameter Konfigurasi

29. Model
Spot Instance
vCPU
Memory
(GiB)
Linux/
Unix Usage
Network
Bandwidth
(Gbps)
EBS Bandwidth
(Mbps)
c5.large 2 4 $0.0313 per Hour Up to 10 Up to 4,750
c5.xlarge 4 8 $0.0638 per Hour Up to 10 Up to 4,750
c5.2xlarge 8 16 $0.1271 per Hour Up to 10 Up to 4,750
c5.4xlarge 16 32 $0.2494 per Hour Up to 10 4,750
c5.9xlarge 36 72 $0.5606 per Hour 10 9,500
c5.12xlarge 48 96 $0.7474 per Hour 12 9,500
c5.18xlarge 72 144 $1.1212 per Hour 25 19,000
c5.24xlarge 96 192 $1.4949 per Hour 25 19,000
c5.metal 96 192 $1.4949 per Hour 25 19,000
c5d.large 2 4 $0.0311 per Hour Up to 10 Up to 4,750
c5d.xlarge 4 8 $0.0623 per Hour Up to 10 Up to 4,750
c5d.2xlarge 8 16 $0.1246 per Hour Up to 10 Up to 4,750
c5d.4xlarge 16 32 $0.2556 per Hour Up to 10 4,750
c5d.9xlarge 36 72 $0.5717 per Hour 10 9,500
c5d.12xlarge 48 96 $0.7474 per Hour 12 9,500
c5d.18xlarge 72 144 $1.1212 per Hour 25 19,000
c5d.24xlarge 96 192 $1.4949 per Hour 25 19,000
c5d.metal 96 192 $1.4949 per Hour 25 19,000
- MQTT
Benchmark
- Locust.io
- IoT Apps
- Redis
- MongoDB
- Broker Cluster

30. The Final Myth?
Terima Kasih