Cache first cloud native microservices using Hazelcast and Kubernetes

1. @mesutcelik
Cache First Cloud-Native
Microservices
Mesut Celik,Hazelcast

2. @mesutcelik
Mesut Celik
● Tech Lead @Hazelcast
● Java Developer
● lived in Turkey,Belgium,US
● Twitter: @mesutcelik

3. @mesutcelik
Agenda
● Hazelcast Introduction
● Caching Patterns
● Kubernetes Deployments
● Day 2 Operations

4. @mesutcelik
Hazelcast Products
● Hazelcast IMDG
● Hazelcast Jet
● Hazelcast Cloud
● Management Center
Hazelcast is an Open Source
Company
● Apache 2 License
● Remote friendly
● raised $21M series C
● Always hiring!!!!

5. @mesutcelik
Why Distributed Cache?
● Cache, Memory is faster than storage
● Distributed,Natural fit for Cloud-Native Deployments

6. @mesutcelik
Read-Through Cache

7. @mesutcelik
Read-Through Cache 1,CaryPartition_34
1,CaryBackup_34

8. @mesutcelik
Hazelcast and Kubernetes
Caching Patterns

9. @mesutcelik
Distributed Caching Patterns in Kubernetes
● Embedded
● Sidecar
● Proxy Cache
● Client-Server

10. @mesutcelik
POD
Spring Boot Microservice
Hazelcast
Member
POD
Hazelcast ClusterKubernetesService
Http Request
Embedded
Spring Boot Microservice
Hazelcast
Member
Kubernetes Cluster

11. @mesutcelik
Embedded
● Simple
● Java only
● Data co-located with Microservice JVM

12. @mesutcelik
Python
Microservice
POD
Hazelcast
Member
Python
Microservice
POD
Hazelcast
Member
Hazelcast ClusterKubernetesService
Http Request
SideCar Kubernetes Cluster
hazelcast-client
hazelcast-client

13. @mesutcelik
Sidecar
● Simple
● Multiple Programming Languages
○ Java, Node.js, .NET, Python, Golang
● Data co-located with Microservice POD

14. hazelcast-proxy
sidecar
POD
Spring Boot
Application
POD
Spring Boot
Application
Hazelcast ClusterKubernetesService
Http Request
Http reverse
proxy caching
Hazelcast
Member
hazelcast-proxy
sidecar
Hazelcast
Member
Kubernetes Cluster

15. Http reverse proxy caching
● Injecting Cache into every microservice
● No coupling with any Cache API
● Enabled w/o code change

16. @mesutcelik
POD
Spring Boot Microservice
hazelcast-java-client
Hazelcast ClusterKubernetesService
Http Request
Client Server
POD
Hazelcast
Member
POD
Hazelcast
Member
Kubernetes Cluster
POD
Spring Boot Microservice
hazelcast-java-client
POD
ASP.NET Core
Microservice
hazelcast-dotnet-client
KubernetesService
Http Request

17. @mesutcelik
POD
KubernetesService
Http Request
Hazelcast Cloud
POD
cloud.hazelcast.com
Hazelcast Cluster
Kubernetes Cluster
Spring Boot
Application
hazelcast-client
Spring Boot
Application
hazelcast-client

18. @mesutcelik
Client Server
● CaaS - Cache as a Service
● Multiple Programming Languages
○ Java, nodejs, .NET, Python, Golang
● Separate Hazelcast Cluster Maintenance

19. @mesutcelik
Kubernetes Deployment Options
For Hazelcast

20. @mesutcelik
Hazelcast Helm Chart
$ helm install stable/hazelcast

21. @mesutcelik
Hazelcast Kubernetes Operator
● OperatorHub.io
● Openshift 4 Operator Catalog
● IBM Cloud Private

22. @mesutcelik
Hazelcast Cloud
● Managed Service
● Pay as you go model
● https://cloud.hazelcast.com

23. @mesutcelik
Deployment Best Practices

24. @mesutcelik
Predictable Resource Allocation
● Calculate CPU/Memory Requirements
● Set Resource Request and Limits as the same value

25. @mesutcelik
Pod Anti-Affinity
● Hazelcast per K8S Node
● Large Kubernetes Clusters
● Safer in Node crushes

26. @mesutcelik
Multi-zone Deployment with ZONE_AWARE
hazelcast-member -1
hazelcast-member -2
hazelcast-member -3
Availability-Zone-A
hazelcast-member -4
hazelcast-member -5
hazelcast-member -6
Availability-Zone-B
Multi-zone Hazelcast Cluster
* Multi-zone if your architecture prefers availability over performance

27. @mesutcelik
Day 2 Operations

28. @mesutcelik
Scaling Hazelcast Cluster
● Manual Scaling (kubectl scale or helm upgrade)
● Horizontal Pod Autoscaler (HPA)
● Vertical Pod Autoscaler
● Cluster Autoscaler

29. @mesutcelik
Rolling Upgrade - Why?
● Hazelcast Version Update
● Config Change
● Adding Jar to the classpath

30. @mesutcelik
Rolling Upgrade - How?
● helm upgrade (--set image.tag)
● kubectl set image
● StatefulSet
● Deployment with RollingUpdateStrategy
● Enable Graceful Shutdown
● Optimized Grace Period

31. @mesutcelik
Monitoring
● Hazelcast Management Center
● Metrics via Prometheus
● Alerting

32. @mesutcelik
Thank you!
● https://twitter.com/mesutcelik
● https://github.com/mesutcelik
● https://www.linkedin.com/in/mesutcelik/